Earthquake Engineering Research - HKU Libraries - The University ...

Proceedings of the International Conference on 

Advances and New Challenges in 

Earthquake Engineering Research 

(ICANCEER) 

In honor of the late Professor Liu Huixian 

August 15-20, 2002 Harbin and Hong Kong, China 

Hong Kong Volume 

Edited by 

J.M. Ko 

Y.L. Xu 

Organised by 

The Hong Kong Polytechnic University, Hong Kong 

Sponsored by 

Asian-Pacific Network of Centres for Earthquake Engineering Research (ANCER) 

The Hong Kong Institution of Engineers 

The Environment, Transport and Works Bureau, The Government of the HKSAR

Copyright © 2003 by the Organizing Committee of 

the International Conference at The Hong Kong 

Polytechnic University. 

All statements made or opinions expressed in the 

Proceedings are those of the Authors. The 

Organizing Committee will not be responsible for 

the statements made or for the opinions expressed in 

the Proceedings. 

All Rights Reserved. No part of this publication 

may be reproduced, stored in a retrieval system, or 

transmitted in any form or by means, electronic, 

mechanical, photocopying, or otherwise, without the 

prior permission of Copyright owner. 

Printed by e8print.com Solutions Limited 

ISBN: 962-367-373-6

vw:. 

PREFACE 

Dunng a time when earthquake engineering research is in a rapid process of transformation 

where innovative engineering methods and new emerging technologies toward hazard 

reduction and improved public safety are being increasingly emphasized, a unique 

professional consortium called Asian-Pacific Network of Centres for Earthquake Engineering 

Research (ANGER) has been established and sponsored a major International Conference on 

Advances and New Challenges in Earthquake Engineering Research (ICANCEER) in August 

15-20, 2002. The Conference consisted of two consecutive back-to-back meetings in Harbin 

and Hong Kong. The Harbin meeting was hosted by the Institute of Engineering Mechanics of 

China Seismological Bureau in August 15-17, 2002, where the program focus was on new 

phenomena of earthquake engineering and innovative solution approaches. Volumes 1 and 2 

(Harbin volume) of proceedings were a collection of the Harbin meeting papers. The Hong 

Kong meeting was hosted by The Hong Kong Polytechnic University in August 19-20, 2002, 

in which problems of moderate seismicity and intelligent infrastructure engineering were the 

meeting emphases. The papers presented in the Hong Kong meeting and the panel discussion 

reports on two special topics were published in this volume (Hong Kong volume) of 

proceedings. 

Based on the historical records, Hong Kong is one of the typhoon regions in the world. On 

the other hand, Hong Kong is relatively far away from the active tectonic plate boundary and 

has not experienced large magnitude earthquake yet. Buildings in Hong Kong are therefore 

designed for strong winds but no provision for seismic resistance. However, Hong Kong is 

classified as an area of earthquake intensity 7 (0.15g GPA at bedrock for a return period of 

475 years) by China Seismological Bureau, which is of the same intensity of Shenzhen where 

buildings are required to design for earthquake. This of course alerts the people in Hong 

Kong in particular after the Kobe Earthquake in 1994 which, to a certain extent, is beyond the 

expectation of many earthquake-engineering experts. 

Since 1995, the Hong Kong Government has been taking a leading role to address this 

sensitive issue. In 1995, the Geotechnical Engineering Office of the Civil Engineering 

Department commissioned a project to the University of Hong Kong and the Guangdong 

Seismological Bureau to undertake the Seismic Hazard Analysis of the Hong Kong Region. 

The need for seismic considerations in our building design regulations has been raised in the 

Legislative Council in November 1995. In response to this, the Buildings Department 

reckoned the need to set up a 'Working Group' to study the seismic effects on buildings in 

Hong Kong with the aim of identifying more specifically the need and scope of a code of 

practice on the subject. Recently, the Buildings Department committed a consultancy to 

study the Seismic Effects on Buildings in Hong Kong. 

Academic staff members of the universities of Hong Kong have been actively involved in 

earthquake engineering research in the past several years. Though there are great 

achievements in recent years, earthquake engineering research is relatively new in Hong Kong. 

More exchanges and collaboration need to be established with those regions and countries 

which have a much longer research history in earthquake engineering to ensure ourselves 

researching in the forefront.

IV 

The Hong Kong Polytechnic University is very pleased to have the opportunity to organize 

this International Conference together with the Institute of Engineering Mechanics of China 

Seismological Bureau, which provides a forum for academics and professionals of Hong 

Kong to share experience and exchange views with participants from other parts of the world. 

The Hong Kong volume of proceedings contains 2 panel reports, 6 keynote papers, and 64 

contributed papers covering a wide spectrum of topics: engineering seismology and 

geotechnical engineering, seismic risk and disaster management, smart materials and smart 

structures, structural analysis and design, structural control, system identification, and health 

monitoring and damage detection. The contributions reflect the current state-of-the-art and 

point to future directions of earthquake engineering research. 

The organization of a conference of this magnitude would not have been possible without the 

support and contributions of many individuals and organizations. We sincerely appreciate the 

support from The Hong Kong Institute of Engineers and The Environment, Transport and 

Work Bureau of The Government of the HKSAR. We would like to thank the members of the 

International Advisory Committee, the International Scientific Committee, and the 

Conference Organising Committee. We also wish to express our gratitude to all the 

contributors for their careful preparation of the manuscripts. Thanks are also due to those who 

devoted time and effort to the organization of the Conference and publication of the 

Proceedings, including the secretarial staff and research students of the University. 

J.M. Ko 

Y.L. Xu

OPENING ADDRESS 

By Prof. Yuchen Liu 

Deputy Director of China Seismological Bureau 

Distinguished guests, friends, ladies and gentlemen, 

Good morning. 

Today we are here to continue the Hong Kong part of the conference after the Harbin part of 

the International Conference on Advances and New Challenges in Earthquake Engineering 

Research, which was just successfully closed the day before yesterday. On behalf of China 

Seismological Bureau and in ray personal capacity, I express my sincere congratulations to 

the Hong Kong part of the Conference and warmly welcome all of you. 

As we know, earthquake disaster is one of the most serious natural disasters threatening the 

mankind, and the effort has been tried hard unswervingly to reduce the earthquake disaster. It 

is confirmed that the earthquake disaster could be effectively reduced by active participating 

and appropriate prevention measures in the disaster reduction. 

Mr. Kofe Annan, the United Nations Secretary-General, emphasized in his speech on the 

occasion of the closing of IDNDR in 1999, that "Prevention is not only more humane than 

cure; it is also much cheaper. Disaster reduction and disaster relief are complimentary, and yet 

quite different. Each is vital. Neither should be subsumed by the other." 

It is no doubt that the advancement of earthquake engineering science is playing a very 

important role in earthquake disaster prevention and reduction. China Seismological Bureau 

has been devoting itself to the efforts, paying much attention to the communities and 

cooperation in this field in the world, we are committed to the promotion of the science of 

earthquake engineering cooperation with the international scientific communities, with all of 

you. 

The conference will also contribute to the advancement of earthquake engineering science and 

disasters prevention and reduction in Hong Kong region, China and Asia-Pacific Region. I 

believe, all the participants will hold full discussion and encourage smooth flow of ideas, and 

offer the best for the advancement of earthquake engineering science and the mutual benefit 

of the human being. 

Finally I wish the conference a great success, and I sincerely thank my friends of the Hong 

Kong Polytechnic University for their hard work that ensures successful opening. 

Many thanks again.

VI 

OPENING ADDRESS 

By Ir, Dr. Hon-kwan Cheng 

Chairman of Hong Kong Housing Authority 

Professor Ko, Professor Liu, Distinguished Guests, Ladies and Gentlemen, 

I am very pleased to attend the second leg of the International Conference on Advances and 

New Challenges in Earthquake Engineering Research. And I am greatly honoured to be given 

the opportunity to address such a distinguished audience. In this part of the Conference, we 

will be focusing on the specific challenges for regions of moderate seismicity, and explore 

how emerging technologies and new design philosophies will impact building design in these 

regions. I would like to take this opportunity to briefly discuss the challenges that we face in 

Hong Kong and the recent development of earthquake engineering here. Since our experience 

should be quite typical for places with moderate seismicity, I hope this would serve as a 

useful backdrop for our discussion over the next two days. 

Probably because Hong Kong has never been seriously damaged by earthquake, the structural 

design of buildings and other structures here do not cater for seismicity. At present, there is 

no such requirements in our ordinance or design codes. Up until not too long ago, most 

building designers here believed our risk of being struck by an earthquake is low. However, 

this perception has been under challenge in the past few years. 

Recent studies have shown that seismic intensity for the Hong Kong region with a 10% 

probability of exceedance in 50 years was assessed to be of intensity VII. This is consistent 

with the intensity given for Shenzhen located immediately north across the Hong Kong border. 

In addition, according to the Seismic Ground Motion Parameter Zoning Map 2001, the peak 

acceleration for Hong Kong Island is 0.15g while that for Kowloon and the New Territories is 

0.10 g. The characteristic period for the Hong Kong region is 0.35 second. These classify 

Hong Kong as a region of moderate seismicity. 

This classification is indeed consistent with historical records. In 1991, our Government 

traced past earthquake data as far back as 900 years ago in places within a 350 km radius from 

Hong Kong and found that Hong Kong experienced at least two moderate earthquakes in 

1874 and 1918 respectively. On 16 September 1994, buildings on the reclamation areas of 

Hong Kong experienced substantial vibrations due to a strong far field earthquake. That has 

been the strongest earthquake felt by us since 1918. 

Local engineers also used to believe that given our moderate seismicity, our design provision 

for strong typhoon forces should be sufficient to withstand the earthquake energy. This has 

likewise been put in doubt. Recent studies on comparison between earthquake and wind 

loads in Hong Kong scenarios showed that for low rise buildings, the inter-storey shear and 

overturning moment from earthquake loadings are generally higher than those from wind 

loadings; and for medium to high-rise buildings, there is a strong possibility that the 

earthquake shear and moment effects will exceed those from wind at the upper storeys. 

I am pleased to note that attention is increasingly being paid to the issue of seismicity in Hong 

Kong in recent years. Through the conferences and seminars organised by the local 

professional bodies, we have witnessed a gradual change of minds, with more and more 

practitioners becoming aware of the potential risks. The local academics have also focused 

more research resources in studying sesimicity in the local context, providing us with crucial

Vll 

scientific support in our call for increased preparedness. In 1996, the Buildings Department 

set up a Seismic Working Group to examine the likely effect of earthquakes on buildings in 

Hong Kong. It is understood that this year the Department will commission a consultancy 

study on Seismic Effects on Buildings in Hong Kong. 

Several severe earthquakes that took place in China in the last 40 years occurred in areas once 

classified as low or moderate seismic regions. This discrepancy of zoning intensity and actual 

observed intensity is probably due to the uncertainty of earthquake occurrence and reliability 

of prediction. At the time when Hong Kong last experienced a moderate earthquake in 1918, 

it was still a small village. But today Hong Kong is a major financial centre and one of most 

densely populated cities in the world, and we simply cannot afford to ignore this risk. I 

believe Hong Kong should put some serious thoughts in formulating suitable seismic design 

guidelines and procedures, and the relevant China codes should offer a useful reference. 

More than ever before, we need the collaboration and cooperation of the Government, clients 

and our allied professions who should also need to acquaint themselves as to what 

earthquakes are all about. 

There is a need to change the design philosophy of local designers. We have to think "static 

and dynamic" rather than "static" alone. We have to consider "plastic" in addition to "elastic". 

Inter-storey drift has become an important design criterion. Without a sound knowledge of 

seismic engineering, it would not be possible to work effectively on Mainland projects. There 

are also many important issues in relation to seismic hazard, risk analysis and design practice, 

all waiting for our researchers to explore. In view of the moderate seismicity of Hong Kong, 

more public funding would be required to facilitate our local research. 

Cost will no doubt be a crucial feature in the public debate for the need of seismic 

preparedness. A recent cost-benefit analysis conducted by the China Academy of Building 

Research shows that the additional cost for the seismic fortification of a building in the urban 

area zoned at seismic intensity VI is no more than a few percent of the total structural costs. 

If that is the kind of money we are looking at, I think it is money well-worth spending. 

In closing, I would like to congratulate the Asian-Pacific Network of Centres for Earthquake 

Engineering (ANCER) for sponsoring this wonderful conference. The establishment of the 

Network last year represents a major step forward in fostering cooperation and collaboration 

in earthquake engineering research in the Asia Pacific region, and this International 

Conference is a very good example of the excellent contributions that we can expect from 

ANCER in the years to come. 

I would also like to thank the Hong Kong Polytechnic University for hosting the Hong Kong 

leg of the Conference. Over the years, the University has made valuable contributions to the 

development of seismic engineering in Hong Kong. Apart from being a leading researcher, 

the University has an impressive track record in organising major events to advance our 

knowledge in earthquake engineering. These include the technical visits to Tangshan and 

Beijing in 1996 and the International Workshop on Earthquake Engineering for Regions of 

Moderate Seismicity in 1998, which they jointly organized with other universities in Hong 

Kong and the Mid-America Earthquake Center in USA. 

Ladies and gentlemen, I am sure you are able to advance our knowledge through a lively 

discussion in this Conference. May I wish you all continued success over the next two days. 

Thank you.

Vlll 

Chairman: Lili Xie (China) 

Members: 

INTERNATIONAL SCIENTIFIC COMMITTEE 

Daniel P. ABRAMS (USA) 

William ANDERSON (USA) 

Ian BUCKLE (USA) 

Fabio CASCIATI (Italy) 

Peter CHANG (USA) 

Sungpil CHANG (Korea) 

Houqun CHEN (China) 

Shel CHERRY (Canada) 

Anil CHOPRA (USA) 

Weimin DONG (USA) 

Lichu FAN (China) 

Toshi FUJIMORI (Japan) 

Yozu FUJINO (Japan) 

Polat GULKAN (Turkey) 

Minhong HE (China) 

George W. HOUSNER (USA) 

Yuxian HU (China) 

Hirokazu EEMURA (Japan) 

W. D. IWAN (USA) 

Paul JENNINGS (USA) 

James JIRSA (USA) 

Hiroyuki KAMEDA (Japan) 

Kazuhiko KAWASHIMA (Japan) 

Kazuhiko KASAI (Japan) 

Jan-ming KO (Hong Kong, China) 

Helmut KRAWINKLER (USA) 

George C. LEE (USA) 

OF. LEE (Hong Kong, China) 

Zhenpeng LLAO (China) 

Gao LIN (China) 

S. C. LIU (USA) 

Yuchen LIU (China) 

Chinhsiung LOH (Chinese Taipei) 

Stephen MAHIN (USA) 

Jack P. MOEHLE (USA) 

Joanne NIGG (USA) 

Tsnueo OKADA (Japan) 

Shun OTANI (Japan) 

Joseph PENZffiN (USA) 

Qiaozhai QI (China) 

Jiping RU (China) 

Haresh SHAH (USA) 

Liqin SHAO (China) 

Mete SOZEN (USA) 

B.F. SPENCER (USA) 

Eizaburo TACHIBANA (Japan) 

Wilson TANG (Hong Kong, China) 

Jun'ichi TOHMA (Japan) 

Kenzo TOKI (Japan) 

Chia-ming UANG (USA) 

Akira WADA (Japan) 

Guangyuan WANG (China) 

Huijuan WU (China) 

Hiro YAMANOUCHI (Japan) 

Kazuo YOSHIDA (Japan) 

Heping ZHAO (China) 

Xiyuan ZHOU (China) 

Shilong ZHU (China)

IX 

INTERNATIONAL ADVISORY COMMITTEE 

Chairmen: Qiaozhai QI (China) 

Jan-ming KO (Hong Kong, China) 

Members: 

Sung-pil CHANG(Korea) 

George C. LEE (USA) 

S. C. LIU (USA) 

B.F. SPENCER (USA) 

LOCAL ORGANIZING COMMITTEE 

ICANCEER 2002 - Harbin Conference 

Chairman: 

Qiaozhai QI 

Vice-Chairman: Xing JIN, Jianfa HUANG 

Secretary general: lie GUI, Zhiqiu WANG, Ming ZHAO 

Members: 

Xun GUO, Xingmin HOU, Junqi LIN, Shutao XING, Xuelan ZHANG 

ICANCEER 2002 - Hong Kong Conference 

Chairman: Jan-ming KO 

Vice-Chairman: Y.L. XU 

Secretary general: Eddie S.S. LAM 

Members: 

Kanvtim CHAU, Yuk-lung WONG, Yi-qing NI, Hoat-Joen PAM, 

Chih-chen CHANG, J. Q. LI

CONTENTS 

Hong Kong Volume 

Preface 

Opening Address by Prof. Yuchen Liu 

Deputy Director of China Seismological Bureau 

Opening Address by Ir. Dr. Hon-kwan Cheng 

Chairman of Hong Kong Housing Authority 

International Scientific Committee 

International Advisory Committee 

Local Organizing Committee 

iii 

v 

vi 

viii 

viii 

ix 

Panel Reports 

Performance-based Design and Engineering Seismology 3 

G. C Lee and T. Hutchinson 

Structural Control and Health Monitoring 9 

J.N. Yang andP.C. Chang 

Keynote Papers 

Earthquake Hazard Evaluation and Risk Mitigation in Hong Kong 17 

K.T. Chau, Y.L Xu, J.M. Ko, E.S.S. Lam, Y.L Wong, CM. Lee and Y.Q. Ni 

Parametric Estimation of Ground Motion Attenuation of Turkish Earthquakes 33 

P. Gulkan and E. Kalkan 

Development of JSSI Manual for Passive Control of Buildings 51 

K. Kasai 

A Framework for Developing Performance-based Earthquake Engineering 67 

/.P. Moehle 

Smart Structures Technology Opportunities and Challenges 69 

B.F. Spencer, Jr. 

Earthquake Action Provision in GB50011 -2001 and Three Improvements 85 

XX. Too and S.W. Geng

XI 

Engineering Seismology and Geotechnical Engineering 

Digitization and Data Processing of Strong Motion Earthquake Accelerograms 99 

V.W.Lee 

Evaluation of Dynamic Soil Properties and Liquefaction Potential by Seismic 115 

Piezocone Tests 

T. Liao and P. W. Mayne 

Attenuation Function of Ground Motions for Guangdong Region of Southern 123 

China 

Y.L Wong, S.H. Zheng, J. Liu, Y. Kang, CM Tarn, Y,K. Leung andX.Q. Zhao 

Slope Stability Against Earthquake Considering Three Dimensional Effect 131 

/. Yoshida, T. Kaneto and M. Takao 

Seismic Risk and Disaster Management 

The Mid-America Earthquake Center Research Program Towards Development 139 

of Consequence-Based Seismic Risk Mitigation 

D. Abrams, A. Elnashai and J. Beavers 

Application of Early Damaged Area Estimation System Using DMSP/OLS to 149 

Recent Destructive Earthquakes 

K. Hasegawa, M. Higashida, M. Kohiyama, N. Maki, H. Hayashi, H. W. Kroehl, 

CD. Elvidge and V,R. Hobson 

CCD Camera System Application for Disaster Management 157 

M. Higashida, N. Maki and H. Hayashi 

A Simplified Method to Evaluate Liquefaction of Subsoil of Building 165 

LP.JingandZY.Wu 

A General Approach to Seismic Performance Assessment 173 

H. Krawinkler 

The Real-Time Earthquake Information System in the Northern Kyushu, Japan 181 

with a Small Scale Geoinformation Database for Seismic Intensity Estimation 

H. Narahashi 

Development of Virtual Emergency Response Network and Application 183 

S. Nishimura 

Strong Motion Database and Analysis Method in Mainland China 187 

H.Y. Yu 

Towards Earthquake Hazard Mitigation on Metropolis - a Platform for Risk 195 

Communication 

P. Zhu, M. Abe andJ. Kiyono

Xll 

Smart Materials and Smart Structures 

Critical Factors for Magnetorheological Fluids m Civil Structures 205 

J.D. Carlson 

Implementation of Modal Control for Seismically Excited Structures Using MR 207 

Dampers 

S.W. Cho, B.W. Kim, H.J. Jung and LW. Lee 

Effectiveness of Smart Base Isolation System with MR Dampers in Protecting 215 

Structures in Near-Fault Earthquakes 

S. Nagarajaiah, S. Sahasrabudhe and Y.Q. Mao 

Seismic Protection of A Benchmark Cable-Stayed Bridge Using A Hybrid 223 

Control Strategy 

K.S. Park, HJ. Jung, KM Choi andl.W. Lee 

Structural Vibration Control Using Piezoceramic Patch Actuator 231 

G. Song and B. Xie 

Rheological Behavior of MR Materials in Flow Mode 239 

X. Wang, F. Gordaninejad, G.H. Hitchcock, A. Fuchs, M. Xin and G. Korol 

Innovative Approaches for Structural Health Monitoring of Intelligent 247 

Infrastructure Systems 

R.W. Wolfe, S.F. Masri andJ. Caffrey 

Structural Analysis and Design 

Drift-Based Seismic Assessment of Buildings in Hong Kong 257 

AM Chandler, R.K.L Su andM.N. Sheikh 

Understanding of the Seismic Performance of Asymmetric R/C Building 265 

Structures 

J.W. Dai, Y.L Wong andM.Z. Zhang 

Seismic Resistance of Very High Strength High Rise RC Buildings for Urban 273 

Development 

A.S. Elnashai andB. Laogan 

Performance-Based Seismic Design of Steel Moment Frames Using Target Drift 285 

and Yield Mechanism 

B.C. Goel and S.S. Lee 

A Hybrid Optimization Algorithm: Genetic Algorithm-Simplex 293 

W, Han andZ.P. Liao 

Seismic Simulation of Prestressed Concrete Bridges 301 

C.H. Jeng, Y.L Mo, andT.T.C. Hsu

Xlll 

Progress in the Seismic Design and Retrofit of Bridges in Korea, a Moderate 309 

Seismicity Region 

J.K. Kim, LH. Kim, G.H. Juhn andD.Y. Cho 

Seismic Behavior of Wooden House Using Distinct Element Method 317 

J. Kiyono and A. Furukawa 

The Design of Structural Concrete Regions for Seismic Actions by the Strut-and- 323 

Tie Method 

DA. Kuchma and T.N. Tjhin 

Seismic Response of Arch Dams Including Strain-Rate Effects 331 

G.Lin and S.Y.Xiao 

Nonlinear Static Analysis Method Based on Displacement 339 

W.G. TuandY.S.Zou 

Educational and Code Requirements for Seismic Design in Hong Kong 349 

G. Williams 

Performance-Based Design of Gravity Retaining Wall in Seismic Areas 361 

X. Zeng 

An Optimization Performance-Based Earthquake-Resistant Design 369 

Y.S.ZouandW.G.Tu 

Structural Control 

Adaptive Colony System for Stability Analysis of Slopes in Earthquake Zone 381 

C.F. Chen and X.N. Gong 

Dynamic Analysis of Friction-Damped Structures 389 

LL Chung, LY. Wu and Y.P. Wang 

Behavior of Seismic Protective Devices : A Computational Mechanics Approach 397 

G.F. Dargush, H. Cho andR. Radhakrishnan 

Large-scale Tests on Smart Structures and Semi-Active Control by MR Damper 405 

H. Fujitani, T. Azuhata, K. Morita, T. Hiwatashi, Y. Shiozaki, K. Hata, 

K. Sunakoda and C. Minowa 

Predictive Structural Vibration Control Using Soft Computing 413 

H. Furuta and Y. Nomura 

Damage Control of Hysteretic Structures by Using Instantaneous Optimal 421 

Control Algorithm with A Special Weighing Matrix 

H. Li, J.Y, Peng, Y. Suzuki andA.X. Guo 

An Energy Framework for Decentralized Market-Based Structural Control 429 

J.P. Lynch and K.H. Law

XIV 

Toward the Realization of Seismically Isolated Large-span Spatial Structures: 437 

Model Test and Simulation 

T. Matsui, E. Sugiyama, F. Qiao and T. Hibino 

Application of Semi-Active Devices in Structures Subjected to Earthquakes, 445 

Wind and Vibrations 

0. Nagel 

Active Control Study of Cable-Stayed Ting Kau Bridge under Stochastic 453 

Earthquake Excitation 

7.Q. M, JM. Ko, Z.L Huang andJ.Y. Wang 

Hybrid FRC-Encased Steel Truss Beams for Seismic Upgrading of Reinforced 463 

Concrete Structures 

G. Parra-Montesinos, S.C. Goel andS. Savage 

Seismic Response of Multistory Masonry Building with Restricted Base Sliding 471 

M. Qamaruddin, S. Ahmad, H. Irtaza and S.M. Waseem 

Sliding Mode Semi-Active Base Isolation Using The Switching Hyperplane 481 

Designed by Disturbance-Accommodating Bilinear Control 

N. Santo and K. Yoshida 

Evaluation of Vibration Control Effect on Equipment Using Low-Stiffness and 489 

High-Damping Rubber 

S. Higuchi, K. Okuta, H. Dohi, T. Ozawa and 7. Tsujii 

Fractional Derivative Versus General Linear Models of Viscoelastic Dampers for 495 

Seismic Analysis 

M.P. Singh and T-S. Chang 

Seismic Response Control of A Benchmark Cable-Stayed Bridge 503 

J.N. Yang, S. Lin and F. Jabbari 

Evaluation of Supplemental Energy Dissipation Devices in Protecting Highway 511 

Bridges With Soil-Structure Interaction 

J. Zhang and N. Makris 

Seismic Reliability Analysis of Multistory Isolated Brick Buildings 519 

LX. Zhang, M.Z. Zhang, J.R. Jiang andJ.P. Liu 

System Identification, Monitoring System and Damage Detection 

Structural Health Monitoring for Large Structures Using Ambient Vibrations 529 

JM. Caicedo, E. Clayton, S.J. Dyke andM. Abe 

Simultaneous Estimation of Structural Parameter and Earthquake Excitation from 537 

Measured Structural Response 

J. Chen, J. Li and Y.L Xu

XV 

Crustal Deformation Measurement with Satellite Radar Interfermetry: A Review 545 

X.L Ding, Y.Q. Chen, Z.L Li, G.X. Liu andZ.W. Li 

Vision-Based Sensors for Monitoring Seismic Demands 547 

T.C. Hutchmson and F. Kuester 

Unscented Particle Filter for Time Domain Identification of Nonlinear Structural 555 

Dynamic Systems 

K.Y. KooandC.B. Yun 

Embedding Algorithms in a Wireless Structural Monitoring System 563 

J.P. Lynch, A, Sundararajan, K.H. Law, andA.S. Kiremidjian 

Multifaceted Seismic Evaluation and Retrofit Studies of a Major Viaduct 571 

M. Saiidi, A. Itani, Q. Yang and S. Ladkany 

New System Identification Algorithms Combining Monte Carlo Filter and 579 

Genetic Algorithm 

T. Sato, T. Sakanoue and I. Yoshida 

Real-Time Structural Monitoring System 587 

A.M. Sereci, D. Radulescu and C. Radulescu 

Applying E-monitoring Information System for Highway Bridges Post- 595 

Earthquake Damage Evaulations and Warnings 

G.C. Shiah andS.F. Perng 

Smart Health Monitoring System of a Prestressed Box Girder Bridge 603 

X. Wang and M. L. Wang 

Signature Recognition of Structural Damage: Data Analysis and Model-Based 611 

Validation for Destructive Field Tests 

R.R.C. Zhang, M. MacRostie and Y.L Xu 

Index of Contributors 1-1

PANEL REPORTS


Advances and New Challenges in Earthquake 

Engineering Research, Hong Kong Volume 

PERFORMANCE-BASED DESIGN AND ENGINEERING 

SEISMOLOGY 

Chairperson: G. Lee 1 

Recorder (editor): T. Hutchinson 2 

Panelists: D. Abrams 3 , J.K. Kim 4 , H. Krawinkler 5 , W.K. Pun 6 , H. Yamanouchi 7 

INTRODUCTION 

Panelists and interested researchers representing different areas of seisraiciry around the world 

discussed issues related to performance-based design (PBD). Although the merits of PBD were 

recognized as highly laudable, issues regarding the implementation and realization of PBD in 

practice were raised and the challenges ensued upon the profession highlighted. The importance 

of PBD in highly populated urban areas with low to moderate earthquake hazards (such as Hong 

Kong) and the strategies for seismic retrofit in such areas were emphasized. Summaries of the 

general discussions of the group are provided in the following sections. 

POSITIVE OUTCOMES OF PBD 

Models, such as that being developed by the Pacific Earthquake Engineering (PEER) Center, 

were outlined and the benefits of these and other methodologies to society discussed. The 

objectives of the PEER performance-based earthquake engineering (PBEE) methodology under 

development are (1) to facilitate the decision making on cost-effective risk management of the 

built environment in areas of high seismicity, (2) to facilitate the implementation of 

performance-based design and evaluation by the engineering profession, and (3) to provide a 

foundation on which code-writing bodies can base the development of transparent performancebased 

provisions. Such a methodology is advantageous to the stakeholders (e.g. building owners) 

as they may now invest in mitigation, as they need to, for their own performance requirements. 

As a result, the consequences of an earthquake can be controlled for the prudent stakeholder. 

The concept of PBD is advantageous for seismic retrofit applications as well. Performance levels 

within performance provisions are not mandated, thus it provides a choice for owners desiring to 

retrofit their structure to obtain a selected performance. As such, a building owner must know (or 

believe) that a retrofit to a particular level is needed. The difficulty with this concept is that, 

although the design life of a structure may be in excess of 50 years, the current building owner 

may only own the structure for 7-10 years, a window most owners may not believe an 

earthquake will occur within. 

1 Director, Multidisciplinary Center for Earthquake Engineering Research (MCEER), SUNY-Buffalo, NY, USA 

2 Assistant Professor, University of California, Irvine, CA, USA 

3 Director, Mid-America Earthquake (MAE) Center, University of Illinois Urbana-Champaign, IL, USA 

4 Director, Korea Earthquake Engineering Research Center (KEERC), Seoul National University, Korea 

5 Professor, Stanford University, Stanford, CA, USA 

6 Chief Geotechnical Engineer, Civil Engineering Department, Government of Hong Kong SAR, Hong Kong 

7 Chief Executive, Building Research Institute (BRJ), Tsukuba Science City, Japan.

IMPLEMENTING AND REALIZING PBD 

Several important issues were discussed related to the implementation and realization of PBD in 

practice. These include: (i) the challenges of estimating actual performance, maintaining the 

required design level precision and accounting for the various sources of uncertainty, (ii) 

potential legal impediments, and (iii) the education and knowledge dissemination required. 

Estimation of Actual Performance, Design Level Precision and Uncertainty 

Performance is associated with the behavior of a structural system, however, not directly as 

calculated with present methods. As a result, issues, which may be paramount to performance, 

such as building occupants relating performance to human perception of vibration, or damage to 

nonstructural components and contents, are not incorporated in current procedures. Furthermore, 

the strong economic promise of performance-based design makes the consideration of 

nonstructural and asset (direct and indkect) losses of foremost importance in the implementation 

of PBD. A well-designed structure may reach other performance limit states solely due to 

indirect losses from downtime or other nonstructural damage and this must be incorporated in the 

evaluation process. These issues largely instigated PBD, for example, the nearly $30Billion U.S. 

in losses from the 1994 Northridge earthquake were primarily attributed to nonstructural damage. 

Maintaining the level of precision in the design of structural systems that is required by 

performance-based design provisions is also of concern. In performance-based provisions, design 

will no longer be based on one limit state, but rather multiple limit states, requiring further level 

of detail and precision to be attainable. To compound this issue, previous and current design 

codes provide procedures for proportioning strength, not estimating actual behavior. 

Estimation of actual performance to a reasonable level of precision requires a framework 

accounting for (i) the intensity of the ground motion, (ii) the engineering demand imposed on the 

system, (iii) the corresponding damage generated, (iv) a decision variable based on knowledge of 

this demand and (v) realistic performance targets. Perhaps the most critical of these items are the 

intensity of the ground motion and the performance targets. Neither of these items is directly 

related to the engineers' most common tasks and arguably these may introduce the greatest 

amount of uncertainty in the process. 

This raises important questions regarding our ability (and confidence) in predicting earthquake 

hazard, as well as our ability to discretely characterize the structural system. Various sources of 

uncertainty exist in the estimation of earthquake hazard, including: (i) the occurrence of 

earthquakes in space and time, (ii) the earthquake magnitude, (iii) the attenuation from the source 

to the site, and (iv) the inherent randomness of ground motion time histories. Uncertainties in 

characterizing the structural system include: (i) geometric and material properties and (ii) soilfoundation-structure 

interaction (SFSI), and (iii) modeling uncertainties at both the component 

and system levels. Compounded with these are the errors and randomness of the general 

construction of structures. At the decision state, there are uncertainties associated with (i) the 

consequences of exceeding the limit state (life safety, economic losses, downtime), (ii) economic 

assumptions (e.g. current economic status, discount rate, (iii) the recovery rates (availability of 

finances, state of economy in region).

Legal Impediments 

There are many important legal issues that must be addressed if PBD is to be fully realized. The 

level of precision promised by PBD provides the consumer (the building owner) more leverage 

to attribute responsibility to the engineer if the promised level of performance is not realized. 

Protection mechanisms must be in place to shield designers and builders from excess exposure to 

legal issues. Similarly, owners and society must also be protected from inadequate design and 

construction. 

Education and Knowledge Dissemination 

Currently, there is a lack of education regarding performance-based design amongst the general 

public, students, educators, and engineers. Communication and dissemination over time can 

alleviate these gaps in knowledge, providing awareness and fostering integration of performance 

based provisions in practice. Perhaps the most important educational issue in PBD is the 

education of the stakeholder. Stakeholders must decide what level of performance is acceptable 

to them and the corresponding level of investment they desire. This will require simple indices 

for public understanding of risk (i.e. a 2% in 50 year probability of occurrence is difficult for the 

average stakeholder). The result is that engineering practitioners need to be educators to the 

stakeholders. A reasonable approach to this impediment may be to provide scenario-based 

information in the education of stakeholders, facilitating communication in a quantitative risk 

assessment form. 

As a whole, education of the public is a very significant issue, since practitioners desperately 

require their support, in the implementation of PBD. Such education may occur and exist over a 

short or long term. In the short term, the occurrence of major earthquakes raises public 

awareness, therefore facilitating interest and public education of the potential risks. Long-term 

education is more challenging for the profession and requires frequent awareness programs; such 

as regularly cited demonstration shake table videos or other media intervention. 

PBD IN Low TO MODERATE SEISMIC REGIONS 

Low to moderate seismic regions of the world may have the potential for high intensity seismic 

loading, with very low occurrence. This raises important questions as to the applicability of 

adopting performance-based design provisions in these areas, since earthquakes are a wellknown 

lesser hazard. Design in these regions is generally governed by other factors (e.g. wind), 

thus collapse prevention may control the seismic design process and immediate occupancy and 

life safety levels are controlled by default. The result is that design and associated construction 

costs in low to moderate seismic regions may be exasperated if realistic performance guidelines 

specific to these areas are not addressed. 

As a result of the historically lesser hazard earthquakes pose in low to moderate seismic regions 

ductile design is not routinely practiced, therefore limited ductile behavior of structural systems 

may be anticipated. Furthermore, ductile details are not fully developed in these regions 

therefore engineers will need very simple procedures to adopt these new concepts. However, 

simple procedures may not offer the level of precision required by performance-based design. 

The low occurrence of earthquake events also increases the uncertainty in earthquake hazard 

estimation. Although the characteristics of ground motions in areas of high seismicity may vary

greatly from those regions of the world with low to moderate seismicity, usable ground motion 

records are scarce, thus recordings are often borrowed from high seismic regions. 

PBD FOR SEISMIC RETROFIT 

The implicit inclusion of performance limit states is advantageous to designers trying to evaluate 

the benefits of retrofitting a structure. Furthermore, it assists in providing a reasonable 

perspective to tangible outcomes for owners. However, costs to owners will need to be evaluated 

in terms of the remaining life of the structure. The result is that the need to plan and conduct a 

seismic retrofit program in low to moderate seismic regions becomes a trade-off between the 

structures remaining design life and the anticipated return period of a strong enough earthquake 

(which is generally much longer). Communication to the general public becomes even more 

important in such a case where the lesser occurrence of earthquakes reduces public awareness. 

To convince the public, engineers must benchmark against other kinds of risk, for example, 

earthquake risk compared with risks due to typhoons in Hong Kong. 

In many regions (such as Hong Kong), seismic design procedures do not exist, thus adoption of 

seismic performance-based approaches and evaluation of retrofit plans must be exceedingly 

simple as well as justified in light of other natural hazards (e.g. typhoons). Hong Kong, for 

example, has adopted the approach of first developing a holistic study encompassing all risks to 

structures to evaluate the need for seismic retrofit. In order to convince legislators of this need, a 

quantitative risk assessment is pursued, whereby earthquake risk is placed in the context of other 

risks to structures. In Korea, for example, convincing building owners to retrofit for seismic 

loads has been more challenging, while retrofit programs for bridge structures are more readily 

accepted since bridge structures are controlled by the government and life expectancy is more 

carefully scrutinized. Incremental seismic retrofit programs, which have been adopted in the 

United States, may alleviate large sudden financial burdens. If these programs are put in the 

context of multi-hazard mitigation, this too may be successful in convincing the public. It should 

be emphasized to the stakeholder that conducting a thorough performance-based seismic design 

for a new building requires minimal costs, when compared with the potential costs associated 

with post-construction seismic retrofit. 

PERFORMANCE-BASED DESIGN VERSUS PERFORMANCE-BASED PROVISIONS 

Performance-based design is really nothing new and to many extents has been codified, although 

transparent to designers, for many years. New aspects of PBD include the details and levels of 

performance and the understanding of the implications of such performance. The most important 

outcome of PBD is the development of performance-based provisions to effectively 

communicate these outcomes. Performance-based provisions must carefully be balanced with 

minimal complexity. Transparent physical concepts must be provided that do not make the 

process overly complex. This is particularly important in low to moderate seismic zones where 

earthquake design provisions struggle for adoption. Provisions that are too cumbersome will 

receive little embrace by the community and not be adopted. The recent Japanese 

implementation of performance-based provisions is highlighted as an example of successful 

integration with practice. 

Following the 1995 Kobe earthquake in Japan, a revised building standard was implemented 

(Building Standard Law, 1998). Revisions incorporated performance-based provisions with the

expectation of (i) expanding the freedom in design, (ii) encouraging technical development (e.g. 

new materials, design and construction), (iii) facilitating import of foreign goods and 

technologies, and (iv) activating construction activities. The basic concept of these new 

performance-based provisions is that any material, design method and technique should be 

accepted as effective if the Targeted Performance is realized in the Designed/Constructed 

Structure. The most challenging concept, as observed by the Japanese model code, is the 

evaluation of the actual performance. Performance-based provisions must provide methods 

simple enough for building officials to regulate, while still assuring that a specified level of 

performance has been attained. Building officials in Japan are heavily loaded with numerous 

structural permits each year. As an example, in 1996 alone in Japan over one million building 

applications were submitted to 1800 building officials, resulting in 604 buildings requiring 

regulation per building official. To resolve this contradiction, designers and engineers have to 

explain their design in terms of performance and take responsibility for their design as much as 

possible. 

SUMMARY 

In summary, the panel felt that PBD has positive benefits and clearly it will persevere, however, 

many challenges must be overcome before it can be fully realized. These challenges are related 

to (i) the estimation of actual performance, maintaining the required design level precision and 

accounting for the various sources of uncertainty, (ii) the potential legal impediments, and (iii) 

the education and knowledge dissemination required. The importance of PBD in highly 

populated urban areas with low to moderate earthquake hazards (such as Hong Kong) and the 

strategies for seismic retrofit in such areas were emphasized. Finally, panelists agreed, we have a 

long road ahead of us, but international collaboration will strongly promote the broader interests 

and development of PBEE.




STRUCTURAL CONTROL AND HEALTH MONITORING 

Chairperson: Jann N. Yang 1 

Recorder: Peter C. Chang 2 

Panelists: Hideo Fujitani 3 , Hui Li 4 , Billie F. Spencer, Jr. 5 , You-lin Xu 6 , Fuh-Gwo Yuan 7 

INTRODUCTION 

At the second part of the ICANCEER conference held m Hong Kong in 2002, a panel discussion was 

held on structural control and health monitoring applications to earthquake engineering and its effect on 

civil infrastructures. The fields of structural control and health monitoring have been progressing 

rapidly. However, some research and technological bottlenecks have limited their growth. The 

discussion panel looked into some of these roadblocks; identified the reasons that each research topic 

should be studied, and specified reasons that research in these areas would be beneficial to the 

engineering communities. 

SUMMARY OF DISCUSSION 

The panel responded to questions from the audience regarding issues relevant to structural control and 

structural health monitoring. The discussion started by focusing on the perceived bottlenecks in these 

areas of research. The panel and the audience suggested research topics that could resolve some of these 

problems. Existing technologies and their shortcomings were also discussed. Summaries of the general 

discussion of the panel are given in the following. 

Wireless Sensor Technologies 

It was pointed out that current wireless sensor technologies are not truly wireless. They are typically 

wired from the sensor to a transponder. The wireless link consists of the communication from the 

transponder to the internet or data acquisition equipment. Such "wireless" setups eliminate some of the 

complicated wiring that is in use today, but many of the problems associated with wired network cannot 

be solved, including limited number of channels, expensive wiring, etc. 

Professor, Dept. Civil and Enviro. Engineering, University of California, Irvine, CA, USA 

2 Professor, Dept. Civil and Enviro. Engineering, University of Maryland, College Park, MD, USA. 

3 Chief Researcher, Building Research Institute (BRI), Tsukuba Science City, Japan. 

"Professor, Dept. Civil Engineering, Harbin Institute of Technology, Harbin, China. 

5 Professor, Dept. Civil Engineering, University of Illinois at Urbana-Champaign, IL, USA. 

6 Professor, Dept. Civil & Structural Engineering, The Hong Kong Polytechnic University, Hong Kong. 

7 Professor, Dept. Mechanical & Aerospace Engineering, North Carolina State University, NC, USA.

10 

Some applications of wireless technology exist today. They include the Bluetooth protocol for short 

distance wireless data transfer, laser vibrometers, LIDAR, RFID technology, and different techniques 

for energy scavenging to enable wireless applications. Important issues to be resolved are the need for 

truly tetherless sensors. Two major problems for these sensors to be viable are the need for power supply 

and the reduction of the level of power consumption. Additional benefits of the tetherless sensor 

technology are the possibility of using large and dense arrays of sensors and the application of these 

sensors to existing structures. The table below outlines the critical areas of research. 

Critical problem 

Tetherless Sensors are 

needed 

Reasons this problem is 

important 

Wiring is expensive 

Wired sensors are limited 

to < 1000 

Application to existing 

structures 

Enabling technology to 

solve this problem 

Blue tooth technology 

Mote; smart dust 

RFID 

Laser Vibrometer 

Vibration power 

Research 

Structural health 

monitoring 

ITS 

Ship structure monitoring 

B io/en vironmental 

monitoring 

Damage Identification 

Current damage identification methods are capable of determining structural damages in a controlled 

laboratory environment or computer simulation. The size of the detectable damage using many of the 

existing techniques is so large that the damage could often readily be detected by visual inspection. Yet 

the quantitative damage identification is an important component of management decisions for the 

maintenance and retrofit of structures. One major problem that the current methods cannot adequately 

deal with is the effects of environmental changes, which often create signals of the same order of 

magnitude as those produced by the damage. Current technology includes vibration property detection, 

x-ray and gamma ray techniques, ground penetrating radar, ultrasonic impedance, eddy current, laser 

vibrometers, etc. Two methods to deal with the environmental changes are identified; these are: (i) filter 

the environmental effects, and (ii) detect the novelty in signals after the environmental effects have been 

added. The following table outlines the critical areas of research needed: 


Damage identification 


important 

The damage needs to be 

quantified 

No algorithm is currently 

available for quantifying 

different types of damage 

Provide damage 

assessment and 

maintenance decision 

Enabling technology to solve this 

problem 

MEMS sensors, nano-sensors, microactuators, 

wireless sensors, sensor 

network system, AE sensors, etc. 

Wavelets, Hilbert-Huang transform, data 

mining and data fusion, neural network 

methods, genetic algorithms, etc. 


Autonomous SHM 

systems 

Engines, satellites, 

automobiles 

Concrete strength 

prediction

11 

Damage quantification is Ultrasonic waves, X-ray, gamma ray, 

difficult currently GRP, computer termography, etc. 

Vlaterial 

degradation (e.g. 

corrosion) 

Damage 

quantification 

MEMS Technology 

MEMS technology has reduced the size and cost of sensors for many applications. Current MEMS 

sensors include accelerorneters, temperature sensors, relative humidity sensors, etc. The reduction of the 

size and cost allow MEMS sensors to be applied densely over large areas. An added advantage of these 

small sensors is that they do not interfere with the function of the structure. In aerospace applications, 

MEMS sensors can be placed on the wing without changing the airfoil characteristics. Current MEMS 

sensors are wired with power provided from outside sources. Internally powered MEMS will expand the 

use of these small sensors to more applications. 


MEMS sensors 


important 

Potential low cost 

Small size 

Transparency (low added 

weight and low profile) 



Maturing MEMS 

technology 


Dense sensor arrays 

Reliability and durability 

of sensors 

Semi-Active Control 

The semi-active control technology is maturing to the point that it is used in production automobiles. In 

structural applications, it has been shown to be effective while consuming a small amount of energy. In 

many applications, the energy from small batteries suffices. This is particularly important because 

passive and active controllers have their limitations. The panel recognizes that this technology is being 

applied in Asia for infrastructure control. Its use in USA is limited to one application presently. The 

discussion panel identified the durability of MR fluids used in most semi-active controllers to be an area 

of continuing research. 


Semi-active control 


important 

Passive and active 

controllers have 

limitations. 




More robust MR fluids for In application stage 

a wide range of systems Durability of MR fluid

12 

Smart Materials 

Smart materials, such as piezoelectric material, shape memory alloy, magnetostrive composites, etc., 

have made great strides in recent years. A magneto-shape memory alloy has been produced that can be 

controlled magnetically to a strain of 6%. This technology can be applied to increase the presence of 

structural actuators. The actuators have traditionally been made from piezoelectric materials with a 

maximum strain often less than 1%. Non-magnetic shape memory alloys can have maximum strains of 

2 to 4%, which make them also viable for controllable structures. 


Applications of smart 

materials to structural 

control 


important 

Save power and energy 

Multi-functionality 

Adaptation to random 

loads 



Piezo-electnc ceramic 

actuators and sensors 

Shape memory alloy 

actuators and sensors 

SMA 

MQR. dampers 


PZT actuators 

Magneto-SMA actuators 

Integrating Structural Health Monitoring and Structural Control 

Researchers have shown that controlling member forces can aid in the damage detection of structures, 

and structural control is not currently taking advantages of the myriad of information that a dense array 

of low-cost sensors can provide. The panel recognizes that if the structural control system and the 

sensors are designed together, then it is possible for the control system to use the sensed data more 

effectively. With the same token, by controlling the forces in the structure, it is possible to determine 

damages in the structure more accurately. In this manner, the control system and the sensors are tightly 

coupled. Research in the coupling between these two technologies is needed. 


Integrated SHM and 

structural control (SC) 


important 

SHM and SC are 

independently designed 

and implemented 

Current algorithms for SC 

do not take advantages of 

information that is 

available by SHM 

Current SHM systems do 

not use control actuation 

capabilities to more 

effectively assess the health 

of structures 



Commercially available 

and cost-effective control 

actuators (e.g., MR 

dampers) 

Dense array of low-cost 

sensor technology 

Wireless sensor technology 


Development of new 

algorithms that combine 

sensor information and 

control

13 

Structural Control in Strong Wind and Moderate Seismic Region 

In the past three decades, most of the research efforts in structural control were focused on the 

enhancement of either the safety of buildings and structures in regions of high seismicity or the 

serviceability performance of buildings and structures in regions of strong winds. Less attention was 

paid to the integrated design of structural control systems to satisfy both the safety requirement for 

moderate seismic events and the serviceability requirement for strong winds. Such a task may be 

difficult to be accomplished using the passive control technology; however, the semi-active control 

technology with proper control devices and algorithms can be developed to fulfill this challenge. 


Enhanced performance of 

building in strong wind 

and moderate seismic 

region 


important 

Safety and serviceability 

requirements for a costeffective 

strategy 



Semi-active control 

technology 


Implementation of semiactive 

control devices and 

proper control algorithms. 

Durability and Reliability of Smart and Control Systems 

Reliability of active and semi-active control systems is a principal concern of building owners. Research 

in predicting the performance of smart control systems and their reliability is needed before these 

technologies will see any significant use in USA. The panel points out that in most cases, semi-active 

controllers are used based on their passive control capabilities Gust in case the semi-active control fails 

to work). 


Durability and reliability 

of smart control systems 


important 

[ncrease performance of 

structures against nature 

lazards. 



Base isolation 

Passive control 

Semi-active and hybrid 

systems 


In application stage 

Application according to 

age of structure 

Life cycle cost and 

performance of control 

svstems 

EDUCATION 

A critical issue for the future of smart structures is to create cross-cutting educational programs to 

prepare the next generation of engineers to take advantage of this technology in tackling the challenges 

of today's rapidly evolving society. Educational programs should be developed that integrate academic

14 

research and education in a multi-disciplinary setting, including multi-lateral cooperation. Central to 

such programs are student and young researcher exchange. 



important 




Effective education in 

rapidly evolving 

technology 

Improved illustration, 

improved learning 

experience 

Instructors need to be 

educated. 

Availability of smart 

materials and control 

systems 

Control system design, 

use of this technology to 

improve learning

KEYNOTE PAPERS




EARTHQUAKE HAZARD EVALUATION AND RISK 

MITIGATION IN HONG KONG 

K.T. Chau, Y.L. Xu, J.M. Ko, E. S.S. Lam, Y.L. Wong, C.M. Lee and Y.Q. Ni 

Department of Civil and Structural Engineering, 

Hong Kong Polytechnic University, Hong Kong, China 

ABSTRACT 

Geographically, Hong Kong is not located in a region with frequent attacks from destructive 

earthquakes, and, historically, there is no provision for seismic design of building structures in Hong 

Kong. In recent years, the local academicians and government offices start to realize the potential risk 

or consequences caused by earthquakes. For example, on September 16, 2004, buildings on 

reclamation areas of Hong Kong experienced a strong far field earthquake of magnitude of 7.3 

occurred in the vicinity of Dongsha Island. Thousands of people felt the shaking and left their office 

buildings. This is the strongest earthquake felt in Hong Kong since 1918. Coupled with the recent 

earthquake disasters in Kobe, Taiwan, India and Turkey, the public begins to concern the safety of 

Hong Kong buildings if a major earthquake occurred in the vicinity of Hong Kong. This is not 

unreasonable, considering the fact that the seismicity of Tangshan area is lower than the current hazard 

level of Hong Kong before the hit of the great 1976 Tangshan earthquake. In Hong Kong, there are 

many special issues of earthquake engineering need to be addressed. In the urban areas, many high-rise 

buildings are close to each other and use transfer plate/beam systems to achieve maximum open space 

at lower levels. These buildings are often built in a congested reclaimed land, on hilly topography, in 

the proximity of MTR, or on unstable slopes. Deep basements supported on bored or dnven piles for 

parking, shopping mall or for housing utility facilities are not uncommon in these tall buildings. This 

paper thus discusses the earthquake hazard evaluation and risk mitigation in Hong Kong, together with 

some future issues and challenges lying ahead. 

INTRODUCTION 

Geologically, Hong Kong is located on the South China Sea Plate. The nearest active tectonic plate 

boundary is that of the Philippine Sea Plate and the South China Sea Plate between Taiwan and the 

Philippines. Although this active boundary is relatively far away from Hong Kong, historic records 

have indicated that Hong Kong did experience earthquake shaking of intensity up to VII in the 

Modified Mercalli Scale. For example, in 1874 an earthquake of magnitude 5.75 occurred at Dangan

18 

Island, which is about 35 km southeast of Hong Kong Observatory. This is the largest earthquake 

observed within 100 km of Hong Kong in the past 500 years. Another strongly felt earthquake is the 

1918 Shantou earthquake of magnitude 7.4 which is about 300 km away from Hong Kong. These 

records agree with the "Seismic intensity zonation map of china" published in 1990 by the State 

Seismic Bureau of China (now the China Seismological Bureau). The newest published "Seismic 

ground motion parameter zonation map of China" (GB 18306-2001) predicted that a maximum ground 

shaking of exceeding 0.15g at rock site is expected in Hong Kong at least once in every 475 years 

(Wong et al., 1998a-b). This shaking will further be amplified when topographical and soil effects are 

considered (Wong et al., 2000; Wong et al. 1998a-b; Chau and Lo, 2001). However, when the 1874 

Dangan Island earthquake and the 1918 Shantou earthquake occurred. Hong Kong is still a small 

fishing village (see Fig. 1). However, Hong Kong is now a major financial centre and one of the most 

densely populated cities in the world. Any interruption to critical facilities and business operations may 

have serious social and economical consequences (see Fig. 1). 

(a) 

Figure 1. (a) Central in 1870s; (b) Central in 2000 

For instance, Newcastle city in Australia, an area of low to moderate seismic risk similar to Hong 

Kong, was attacked by a relatively small magnitude earthquake (M=5.6) on December 28, 1989 

causing about 20 billion Hong Kong dollars of damage and 13 deaths (EEFIT, 1991). In total, 50,000 

buildings were damaged, 300 buildings were eventually demolished, and 162 people were hospitalised. 

The Unexpected Catastrophe 

1389 Newcastle tarthtjuaJwj rnrformatHm Resources 

Figure 2. The unexpected Newcastle earthquake on December 28, 1989, Australia.

19 

On September 16, 1994, buildings on reclamation areas of Hong Kong experienced a strong far field 

earthquake of magnitude of 7.3 occurred in the vicinity of Dongsha Island. Thousands of people felt 

the shaking and left their office buildings. This is the strongest earthquake felt in Hong Kong since 

1918. Two separate isoseismical maps for the earthquake are shown in Fig. 3. A simulated isoseismal 

of the earthquake using attenuation law of South China was given in Fig. 4 for comparison; and the 

simulation is plotted using a GIS system developed by a collaboration between the Institute of 

Geophysics and the Hong Kong Polytechnic University. Coupled with the recent earthquake disasters 

in Kobe, Taiwan, India and Turkey, the public begins to concern the safety of Hong Kong buildings if 

a major earthquake occurred in the vicinity of Hong Kong. This has raised an increasing concern by 

local researchers, engineers and the general public on the possibility of being hit by a moderate to large 

earthquake. 

• /^J 

*/ • 

«—1 ***^!&&L< 4«HO*S^S.~ 

^">^» : : 5 a^P v ""> / J« 7^"r » 

*^ 

/ 

. . 

TT 17 *=*•' * «flCtSL ' >• ' JtilHjtJ: \ /)>•*•*•• 

Hong Kong ,.,ig££*-,»*. "V. •&•< • / 

Figure 3. Isoseismal maps for the 1994 Tongsha earthquake 

Figure 4. Simulation Isoseismal map of the 1994 Tongsha earthquake using a GIS system 

In 1996, a Working Group was setup by the Buildings Department of the Hong Kong Government to 

examine the likely effects of earthquakes on buildings in Hong Kong. In 1998. three universities in 

Hong Kong (i.e. the Hong Kong Polytechnic University, the Hong Kong University of Science and 

Technology, and the University of Hong Kong) and the Mid-America Earthquake Center in USA

20 

jointly organized an International Workshop on Earthquake Engineering for Regions of Moderate 

Seismicity. No doubt, earthquake resistant design for structures in Hong Kong is now an important 

issue that needs urgent attention. In the early 2002, the Buildings Department of the Hong Kong SAR 

government has issued a consultancy project CAO K49 on "Seismic Effects on Buildings in Hong 

Kong" worth about 5 millions HKS. More earthquake-related studies for Hong Kong are expected in 

the future. 

In terms of the seismic design code of China (GBJ11-89), Hong Kong is classified as an area of 

Seismic Intensity VII. And according to the newest Chinese seismic code (GB 18306-2001) the peak 

ground acceleration is expected to be 0.15g in a return period of 475 years. It seems that, Hong Kong 

should be regarded as an area with moderate seismic risk. According to the China Seismological 

Bureau, buildings to be constructed in any region with an intensity level of VI or above in 475 years 

should be designed in compliance with the seismic provisions of the Chinese code. Thus, it is expected 

that seismic provisions will eventually be required for designing buildings in Hong Kong in order to 

withstand seismic hazards. Inherited strength of existing structures will have to be assessed and 

strengthened to safeguard our building stock and infrastructures against possible seismic attack. We 

should emphasize that the hazard intensity level at Haicheng, Tangshan, and Yingtai before the 

Haicheng earthquake, Tangshan earthquake and Yingtai Earthquake is only estimated at intensity VI, 

which is lower than the current hazard in Hong Kong. Therefore, the conception that Hong Kong is 

safe from large earthquake is a very misleading concept. 

DANGAN ISLAND—POTENTIAL SEISMIC SOURCE FOR HONG KONG 

As mentioned earlier, on June 23, 1874 an earthquake of magnitude 5.75 occurred at Dangan Island, 

which is about 35 km southeast of Hong Kong Observatory. The earthquake led to an earthquake 

intensity of V-VI in Hong Kong. The isoseismal trended along the ENE-WSW direction as shown in 

Fig. 5. 

Figure 5. The isoseismal of the 1874 Dangan Island earthquake together with those of the Heyuan 

earthquake

21 

Some people were injured while two houses collapsed, boulders were dislodged from slopes and some 

retaining walls collapsed. According to the Daily Press, the ground shaking was felt along NE to SW 

direction. There were also observable ground waves, bells rang and water in the harbour was "rough and 

buoyant" (Lee and Workman, 1996). Thus, Dangan Island area is definitely a potential seismic source 

that will generate earthquakes affecting Hong Kong. A closer look of the geological setting in South 

China reveals that there is a large fault system in the northern South China Sea. This fracture zone is 

called "Binghai Fracture Zone' or "Littoral Fault Zone" (see Fig. 6). Figure 6 is modified from Lin and 

Xie (1996). Along this fracture zone, there had been the 1918 Shantou earthquake of magnitude 7.4 

which is about 300 km away from Hong Kong to the east; and there was also the 1605 Qiongshan 

earthquake of magnitude of 7.5 about 450km to the west of Hong Kong. There is, however, a "largeearthquake-missing 

zone" in the middle around Hong Kong. 

Large-earthquake- \ ^ { 

missing -zone x j 

Earthquake with ji! 

^ 

Figure 6. Large-earthquake missing-zone near Hong Kong of the Binghai Fracture Zone 

Figure 7, Fault system of the Binghai Fracture Zone south of Hong Kong and Dangan Island 

In 1982, in order to understand the basement structure characteristics of oil-bearing basins off the Pearl 

River Delta region, seismic profiling of the South China Sea was conducted by the Institute of 

Geophysics, Institutes of Acoustics and South China Sea Institute of Oceanology (Xia, 1985). A deep

22 

tunnel in the biggest island of the Dangan Island was used to detect the seismic waves generated from the 

explosive ignited from the seabed. The geological structures near Dangan Island area were obtained and 

are illustrated in Fig. 7 (modified from Lai and Langford, 1996). 

Littoral Fault Zone 

^- -"* -*.'; ; -X' Dangan isiS 

-. ,;x^>^-;- x 

v;-'vin-^:;C;^- 

' ' "0 ^ ** 40 

Figure 8. A cross-section cutting through Dangan Island area showing a vertical drop of the rock head 

of up to 3 km across the fault The vertical drop is filled with Quaternary deposit and forming a tensile 

rift zone 

The Dangan Island and Hong Kong area shows a very distinct uplift of the seabed basin through a tensile 

normal fault system (see Fig. 8). The geological setting of such vertical drop of the rock head level is 

similar to those observed in Shantou and Qiongshan, at where large earthquakes of magnitude of 7.5 has 

been generated. Therefore, it is believed that the maximum credible earthquake magnitude at Dangan 

Island area may be up to 7.5. Indeed, the China Seismological Bureau did assign such a value for the 

Dangan Island seismic source during the probabilistic earthquake hazard analysis. Thus, Hong Kong may 

some days subject to near field large earthquake of magnitude up to 7.5. Therefore, there is a genuine 

need to incorporate seismic provisions in the building regulation. 

SEISMIC PERJFORMANCE OF TALL BUILDINGS IN HONG KONG 

In Hong Kong, over 70% of the land of 1,000 square km are hilly and thus most buildings constructed in 

Hong Kong are high-rise in order to fully utilize the land to accommodate 7.8 millions people in such a 

tiny place. As shown in Fig. l(b), tall buildings are everywhere on the Hong Kong Island. Following the 

âllocation of the airport to Lantau Island in 1998, new buildings on Kowloon Peninsula are also 

expected to be high-rise. Figure 9 shows the aerial photographs for both the new and old airport. 

In fact, a lot of tall buildings are being built and planned in Hong Kong. For example, the "Two 

International Finance Center" at the Airport Railway Station at Central is near completion. It will be the

23 

tallest buildings in Hong Kong when it is completed in 2003. The building will be 412m tall with 88 

storeys. Figure 10 shows the construction phase of the building in early 2002, and the sketch of the 

building when it is completed. At the same time, another much taller building is now being built on 

Kowloon. This is the "Union Square" illustrated on Fig. 10; the building will be 480m tall with 102 

storeys. It is expected to complete in 2007. Both of them are much taller than the existing 374m Central 

Plaza and the 369m China Bank.Therefore, the seismic performance of tall buildings designed according 

wind code is of utmost importance to Hong Kong. 

Figure 9. The new and old airport in Hong Kong 

Figure 10. The future tallest buildings in Hong Kong. The "Two International Finance Center" on 

Hong Kong Island and the "Union Square" on Kowloon side

Although according to the Chinese seismic code (GB50011-2001), tall buildings need to be designed by 

time history analysis, probably using nonlinear finite element programs. However, there is a need to 

estimate the seismic risk of those existing tall buildings in Hong Kong. We cannot conduct, both 

financially and practically, numerical analysis for all existing tall buildings. Thus, simplified but reliable 

method should be used to estimate the potential damages and risks of existing buildings. 

Shaking Table Test at IEM 

Since the seismic hazard in Hong Kong has traditionally been considered to be low, reinforced 

concrete structures in Hong Kong have been designed with no seismic provisions. This has led to the 

extensive use of transfer systems in the high-rise buildings in order to achieve maximum open space at 

the lower stories. However, the transfer systems could be vulnerable to possible moderate earthquake 

attack due to the formation of soft stories under the transfer. In a recent joint research study by the 

Hong Kong Polytechnic University and the Institute of Engineering Mechanics ("IEM"), State 

Seismology Bureau in Harbin, a 1:20 scale high-rise building was tested in the shaking table. In view 

of the uncertainty of the seismic behaviour of typical tall buildings in Hong Kong, the results of the 

test is very useful to investigate the vulnerability of Hong Kong buildings. 

Figure 11. Shaking table test for a typical tall building in Hong Kong at IEM. 

DAMAGE ANALYSIS FOR SOME BUILDINGS IN HONG KONG 

Seismic risk and damage analyses were conducted preliminary for Tsim Sha Tsui (TST) East and 

Wanchai as shown in Fig. 12. Over 1,400 buildings were classified into 10 building categories, and a 

total of 13 different buildings have been selected for vulnerability analyses.

25 

Figure 12. Tsim Sha Tsui East and Wanchai have been selected for seismic risk analysis 

Figure 13. Four buildings that have been selected for seismic risk analysis 

Four buildings that have been selected for seismic risk and damage analysis were shown hi Fig. 13. 

They include a 42 storey public housing block with transfer system, a typical 14 storey residential 

block in Mei Foo Sun Chuen, a 6 storey government primary school, and a 14 storey commercial 

building at Tsim Sha Tsui East. 

The results of our vulnerability and risk analysis can be shown in GIS platform. The hazard input for 

our GIS system can be of probabilistic approach or of scenario earthquake approach. Figure 14 shows 

a scenario earthquake analysis of assuming a 7.5 earthquake occurred at Dangan Island, which is 

probably worst scenario that Hong Kong may face. The results are shown in Fig. 15 for both TST East 

and Wanchai. A darker colour indicates a higher possibility of damage.

26 

Seismic vulnerability of typical buildings in Hong Kong has been estimated based on the probable 

seismic demand and the likely building capacity. This is done through the estimation of ductility of the 

building. Five degree building damages can then be estimated as a function of seismic input. A 

number of journal papers have been resulted (Wen et al, 2000a-c); and the full details are referred to 

these references. A GIS system written on ArcView Avenue was developed (Wen et al., 2000d-e). 

Ste £* View Ihane Ijieertcr ^ndow 

13 53 nnmm CDEE 000 

Qngn |2**,4U7B. 3+7.2Z3.35|m Enter* IT 484.65B.1t. 760.434.64!m AWKL 1,1 28.9B5.+5'1.122.33 aim 

Figure 14. A scenario earthquake for Dangan Island with magnitude of 7.5 

Figure 15. The probability of exceeding none damages for scenario earthquake shown in Fig. 14.

27 

OTHER ONGOING WORKS 

Soil-pile-structure interactions 

Some preliminary theoretical works on the soil-pile-structure interactions (SPSI) have been done by 

Koo et al. (2002), and Chau and Yang (2002), by using continuum mechanics and by extending the 

classical work by Novak (1991). More recently, Chau et al. (2002a) performed a series of shaking 

table tests on a soil-pile-structure system using the MTS uniaxial shaking table at The Hong Kong 

Polytechnic University. On main new finding is the observation of pounding between soil and pile 

when a soil-pile-structure model is subject to seismic excitations. More importantly, this pounding 

leads to a very large inertia force at the pile cap level, which is bigger than that at the top of the 

structure. Consequently, cracking in piles is induced. To illustrate this pounding between soil and pile, 

Chau et al. (2002a) applied the finite element method (FEM) to explain this unusual large acceleration 

suffered at the pile cap level. The gap element used in the FEM models can replicate the pile cap and 

structural responses accurately. It is, however, believed that the pounding between soil and pile 

investigated by Chau et al. (2002a) is reported for the first time and its implication to seismic pile 

damages deserves further study. 

Figure 16 shows the elevated view and a photograph of the 216m Hopewell Center together with the 

soil-pile-structure model used in our shaking table test at PolyU. The foundation of Hopewell Center 

is built on a sloping ground, and such differential foundation scheme is not advantageous to seismic 

design. A finite element model is shown in Fig. 17. 

Figure 16. An elevated view and a photograph of the Hopewell center and shaking table test on soilpile-structure 

interaction

28 

Figure 17. A finite element model for the soil-pile-structure interaction 

Pounding of Adjacent Buildings 

Pounding between adjacent structures or between parts of the same structure during major earthquakes 

has often been reported, and it has also been identified as one of the main causes for structural 

damages or for complete collapse of structures (Davis 1992, Chau and Wei 2001). For example, 

poundings between structures have been observed in Alaska Earthquake of 1964, San Fernando 

Earthquake of 1971, Mexico City Earthquake of 1985, Loma Prieta Earthquake of 1989, Kobe 

Earthquake of 1995 and Taiwan Chi-Chi Earthquake of 1999. Hong Kong is a densely populated 

modern city with most buildings built closely to each other and on reclaimed land. The separation 

distances between adjacent buildings in some cases are very small, controlled by static wind load only 

without considering any seismic design requirements. Recently, Chau and Wei (2001) extended the 

model of Davis (1992) to consider poundings as nonlinear impacts between two single-degree-offireedom 

(SDOF) oscillators. To validate this model, Chau et al. (2002b) performed a series of shaking 

table tests on poundings between two steel towers that can be considered approximately as singledegree-of-freedom 

oscillators (see Fig. 18). Both observations on the periodic and chaotic poundings 

agree quantitatively with the prediction.

29 

Figure 18. The pounding experiments at PolyU and the use of dampers to mitigate pounding between 

adjacent buildings 

Dampers have also been used to mitigate the potential pounding between two adjacent buildings of 

different dynamic characteristics, as shown in Fig. 17 (Xu et al., 1999a-b; Zhang and Xu, 1999, 2000). 

Both active, passive and semi-active dampers have been or will be considered. 

FUTURE CHALLENGE 

Though the research on earthquake engineering in Hong Kong has made significant progress within a 

short period of time, there remain, however, some important issues that have not been satisfactorily 

solved or have been not yet addressed. They include the attenuation law for Hong Kong, topographical 

effects for amplification, micro-zonation in urban areas, mitigation techniques, design spectrum at long 

period, behavior of transfer system in tall building, coupling between soil-pile-structure interaction and 

pounding. We have the obligation to assist the Hong Kong SAR Government in the assessment of 

earthquake hazard, and the vulnerability and risk of our building stock and infrastructures. There is 

also an urgent need to develop and implement remedial measures to strengthen or retrofit our 

structures, and to educate the public to understand the meaning of "moderate" seismic hazard and the 

extent of "risk" involved. In recent years, many new technologies have been developed worldwide 

including devices on vibration control, sensor technology, smart materials, global information systems 

and others. Bearing in mind that structures in Hong Kong are unique, application of the new 

technologies will need some degree of modifications. Thus, adaptation of the new technologies to 

improve the seismic performance of our structures is another challenge to be faced by the researchers 

and professionals in Hong Kong.

30 

RESEARCH COLLABORATION 

We have been active in national and international research collaboration since 1995. Academic visits 

have been made to Institute of Engineering Mechanics, the Institute of Geophysics of the China 

Seismological Bureau, the Institute of Earthquake Engineering of China Academy of Building 

Research of Ministry of Construction of China, Tongji University, and Harbin Industrial University. 

More recently, the National Natural Science Foundation of China granted its first kind ever of national 

important project on intelligent vibration control of civil engineering structures, in which the Hong 

Kong Polytechnic University is also involved. 

Internationally, we have linked with the University of California at Berkeley, USA, and the University 

of Turin, University of Notre Dame, University of Illinois, Mid-America Earthquake Center, and the 

newly formed ANCER (Asian-Pacific Network of Centers for Earthquake Engineering Research), 

which comprising of 7 regional earthquake research centers. Figure 19 shows a group photograph 

taken when we participated in the Taiwan Strait Conference on Cities Hazard Mitigation in 2000. 

Figure 19. Seismic group's participant in earthquake conference 

CONCLUSIONS 

The earthquake hazard and risk problems of Hong Kong have been discussed briefly in this paper. 

Some new results and views were presented, especially regarding the potential hazard induced by the 

Dangan Island seismic source. Significant progress has been made in Hong Kong in the field of 

earthquake engineering in the past ten years, but there are still other research topics and urgent tasks 

faced by researchers and other professionals in Hong Kong. National and international collaboration is 

inevitable to speed up the research progress, the development of seismic design standards for building 

structures, and the implementation of new technologies for seismic hazard mitigation. This can be 

realized through exchange of data and information, jointly organized seminars and conferences, 

collaborative research projects, and exchange of personnel.

31 

ACKNOWLEDGMENTS 

The writers are grateful for the financial supports from the Research Grants Council of Hong Kong and 

The Hong Kong Polytechnic University (under A202 and A214). 

REFERENCES 

Chau K.T and Lo, J.H.Y. (2001). Soil amplification in Hong Kong under far field earthquakes. Soft Soil 

Engineering, ed. By Lee CF, Lau CK, Ng CWW, Kwong AKL, PLR Pang, Yin J-H, and Yue ZQ. The 

Third International Conference on Soft Soil Engineering (3 rd ICSSE), Balkema, pp245-250. 

Chau KT and Wei XX (2001). Pounding of structures modeled as nonlinear impacts of two oscillators. 

Earthquake Engineering and Structural Dynamics, Vol. 30, No.5, pp633-651. 

Chau KT, Shen CY, Guo X. (2002a) Pounding between soil and pile under earthquake excitations: 

Shaking table tests and FEM analyses. To be submitted to Journal Geotechnical and Geoenvironmental 

Engineering ASCE. 

Chau KT, Wei XX, Guo X, Shen CY. (2002b) Experimental and theoretical simulations of seismic 

pounding between adjacent structures. Earthquake Engineering and Structural Dynamics, 

(provisionally accepted). 

Chau KT and Yang X (2002). Nonlinear interaction of soil-pile in horizontal vibration. To be 

submitted to Journal of Engineering Mechanics ASCE 

Davis RO (1992). Pounding of buildings modeled by an impact oscillator. Earthquake Engineering 

and Structural Dynamics 21,253-274. 

EEFIT (1991). The Newcastle, Australia Earthquake, Publ. Earthquake Engineering Field 

Investigation Team, Institution of Structural Engineers, London, UK. 

GB 18306-2001 (2001). Seismic ground motion parameter zonation map of China. China, 2001 

GB 50011-2001 (2001). Code for Seismic Design of Buildings National Standard, Beijing, China, 

2001. 

GBJ11-89 (1989). Code for Seismic Design of Buildings, National Standard, Beijing, China, 1994. 

Koo KK, Chau KT, Yang X, Lam, SS and Wong YL (2002). Soil-pile-structure interactions under SH 

waves. Earthquake Engineering and Structural Dynamics, (accepted). 

Lai KW and Langford RL (1996). Spatial and Temporal characteristics of major faults of Hong Kong. 

Seismicity in Eastern Asia, Geological Society of Hong Kong Bulletin No 5, 72-84. 

Lee CM and Workman DR (1996). Earthquakes in the South China region and their impact on Hong 

Kong. Seismicity in Eastern Asia, Geological Society of Hong Kong Bulletin No. 5, 92-103. 

Lin J and Xie M (1996). Time and space distributions of seismicity in Southeastern coastal China. 

Seismicity in Eastern Asia, Geological Society of Hong Kong Bulletin No. 5,12-22. 

Novak M (1991). Piles under dynamic loads. State of the Art Paper. 2nd International Conference on 

Recent Advances in Geotechnical Earthquake Engineering and Soil Dynamics, University of 

Missouri-Rolla, Vol. Ill, pp250-273. 

Wen ZP, Hu YX and Chau KT (2002a). Site effect on vulnerability of high-rise shear-wall-buildings 

under near and far field earthquakes. Submitted to Soil Dynamics and Earthquake Engineering 

Wen ZP, Chau KT and Hu YX (2002b). The effects of site conditions and epicentral distance on damage 

probability matrix of building. To be submitted to Natural Hazards.

Wen ZP, Hu YX and Chau KT (2002c). Site effect on vulnerability of 21 story buildings in Hong Kong 

under near and far field earthquakes. Tenth International Conference On Soil Dynamics and Earthquake 

Engineering, October 7-10,2001, Philadelphia, USA, pp 225. 

Wen ZP, Lu HS, Chau KT and Hu, YX (2002d). A GIS system for seismic damage estimation of Hong 

Kong buildings". Submitted International Conference on Innovation and Sustainable development of 

Civil Engineering m the 21 st Century August 2002, Beijing (accepted). 

Wen ZP, Lu HS, Chau KT and Hu YX (2002e). A GIS system integrated with vulnerability for seismic 

damage and loss estimation of Hong Kong. An International Conference on Advances in Building 

Technology, with The Theme Advances In Building Technology, 4-6 December 2002, Hong Kong 

(accepted). 

Wong YL, Guo X, Yuan Y and Chau KT (2000). Influence of soft sandwich layer on seismic response 

of soil sites in Hong Kong. Journal of Natural Disasters, Vol. 9, No. 1, pp 109-116 (in Chinese). 

Wong YL, Zhao JX, Chau KT and Lee CM (1998). Assessment of seismicity model for Hong Kong 

region. The HK1E Transactions, Vol. 5, No. 1, pp50-62. 

Wong YL, Zhao JX, Lam SEE and Chau KT (1998). Assessing seismic response of soft soil sites in Hong 

Kong using microtremor records. HKIE Transactions, Vol. 5, No.3, pp70-79. 

Xia K (1985). The first sea bottom seismometer test in the South China Sea shallow water area. Proc. Of 

the Seminar on Marine geology of Hong Kong and the Pearl River Mouth. Hong Kong Geological 

Society, HK, pp. 29-37. 

Xu YL, He Q and Ko JM (1999). Dynamic response of damper-connected adjacent buildings under 

earthquake excitation, Engineering Structures, Vol. 2, No. 21, pp 13 5-148. 

Xu YL, Zhan S, Ko JM and Zhang WS (1999). Experimental investigation of adjacent buildings 

connected by fluid dampers. Earthquake Engineering and Structural Dynamics, Vol. 28, pp609-631. 

Zhang WS and Xu YL (1999). Dynamic characteristics and seismic response of adjacent buildings linked 

by discrete dampers. Earthquake Engineering and Structural Dynamics, Vol. 28, ppl 163-1185. 

Zhang WS and Xu YL (2000). Vibration analysis of two buildings linked by Maxwell model-defined fluid 

dampers. Journal of Sound and Vibration, Vol. 233, No. 5, pp775-796. 

32




PARAMETRIC ESTIMATION OF GROUND MOTION 

ATTENUATION OF TURKISH EARTHQUAKES 

Polat GUlkan 

Disaster Management Research Center and Department of Civil Engineering 

Middle East Technical University, 06531 Ankara, Turkey 

Erol Kalkan 

Department of Civil & Environmental Engineering, Rensselaer Polytechnic Institute, 

Troy, 12180 NY, US A 

Abstract: The main result of this study is the development of a consistent set of 

empincal attenuation expressions for predicting free-field horizontal components of peak 

ground acceleration (PGA) and 5 percent damped pseudo acceleration response spectra 

(PSA) from 47 strong ground motion records recorded in Turkey. The relationships for 

Turkey were derived in similar form to those previously developed by Boore et al. (1997) 

for shallow earthquakes in western North America. The used database was compiled for 

earthquakes in Turkey with moment magnitudes (M w ) 5 that occurred between 1976- 

1999, and consisted of horizontal peak ground acceleration and 5 percent damped 

response spectra of accelerograms recorded on three different site conditions classified as 

rock, soil and soft soil The empirical equations for predicting strong ground motion were 

typically fit to the strong motion data set by applying nonlinear regression analysis 

according to both random horizontal components and maximum horizontal components. 

Comparisons of the results shows that ground motion relations for earthquakes in one 

region cannot be simply modified for use in engineering analyses in another region. Our 

results, patterned after the Boore et al. expressions and dominated by the Kocaeli and 

Diizce events in 1999, appear to underestimate predictions based on their curves for up to 

about 15 km. For larger distances the reverse holds. 

Introduction 

Estimation of ground motion, either implicitly through the use of special earthquake 

codes or more specifically from site-specific investigations is essential for the design of 

engineered structures. The development of design criteria requires, as a minimum, a 

strong-motion attenuation relationship to estimate earthquake ground motions from 

specific parameters characterizing the earthquake source, geologic conditions of the site, 

and the length of the propagation path between the source and the site. 

This study describes the best estimates and uncertainties in the ground motion 

parameters predicted in a functional form that can be used in probabilistic hazard studies 

and other earthquake engineering applications. These models and the values of the 

predictor parameters were developed by an extensive analysis of ground motion data and 

its relevant data. This effort was partly motivated by the occurrence of the 1999 (M w = 

7.4) Kocaeli and 1999 (M w = 7.1) Dtizce earthquakes. The Kocaeli earthquake was the 

largest event that occurred in Turkey within the last 50 years, and it is the first wellstudied 

and widely recorded large NAP (North Anatolian Fault) event.

34 

The data includes records from earthquakes of moment magnitude greater than about 

5, and site conditions characterized as soft soil, soil and rock with closest distance less 

than about 150 km. This presents a unique opportunity to study the indigenous 

attenuation characteristics of earthquake ground motions. Also, the study of the effects 

of local site on the attenuation of earthquake ground motions becomes possible since the 

recording stations are fixed and many stations have several records. 

Finally, this paper describes the procedure for estimating ground motion at various 

soil sites by presenting the tables and equations that describe attenuation functions and 

associated measures of uncertainty. One of the major purposes of this paper is to make 

comparisons between the direct use of attenuation relationships developed elsewhere for 

Turkey, and to illuminate the reasons for their differences. 

Database 

Database comprises 93 records from 47 horizontal components of-19 earthquakes 

occurred in Turkey between 1976 and 1999. Details of data library are given in Table 1, 

and listings of the earthquakes and the number of recordings for each of the strong 

motion parameters are presented in Table 2. Station names have not been translated so 

that independent checks may be run. Recordings from small earthquakes were limited to 

the closer distances than large earthquakes depending on the magnitude and the geology 

of the recording site to minimize the influence of regional differences in attenuation and 

to avoid the complex propagation effects coming from longer distances. 

In the data set, earthquake size is characterized by moment magnitude M w , as 

described by Hanks and Kanamori (1979). When original magnitudes were listed in 

other scales, conversion was done according to Wells and Coppersmith (1994). The 

magnitudes are restricted to about M w^ 5.0 to emphasize those ground motions having 

greatest engineering interests, and to limit the analysis to the more reliably recorded 

events. In the regression phase, magnitudes of earthquakes were locked within +/- 0.25 

band intervals centered at halves or full numbers in order to eliminate the errors coming 

from the determination of these magnitude values. Figure 1 shows the distribution of 

these earthquakes in terms of magnitude, station geology (defined below) and source 

distance r c] , defined as the closest horizontal distance between the recording station and a 

point on the horizontal projection of the rupture zone on the earth's surface. However, 

for some of the smaller events, rupture surfaces have not been defined clearly therefore 

epicentral distances are used instead. We believe that use of epicentral distance does not 

introduce significant bias because the dimensions of the rupture area for small 

earthquakes are usually much smaller than the distance to the recording stations. 

Examination of the peak ground motion data from the small number of normal-faulting 

and reverse-faulting earthquakes in the data set showed that they were not significantly 

different from ground motion characteristics of strike-slip earthquakes. Therefore, 

normal, reverse or strike-slip earthquakes were combined into a single fault category. 

Peak horizontal acceleration (PGA) and pseudo response spectral acceleration (PSA) are 

represented considering both maximum and random horizontal components. These are 

explamed below. 

The data used in the analysis constitutes only main shocks of 19 earthquakes. They 

were recorded mostly in small buildings built as meteorological stations up to three 

stories tall because the strong motion stations in Turkey are co-located with institutional 

facilities for ease of access, phone hook-up and security. This causes modified 

acceleration records. This is one of the unavoidable causes of uncertainties in this study,

35 

Table 1. Records Used in the Development of the Attenuation Equations for Peak Horizontal 

Acceleration and Spectral Accelerations 

Date 

dd.mm,w 

1908.1976 

05 10.1977 

16.121977 

18.07 1979 

0507 1983 

05 07 1983 

05 07 1983 

30 10 1983 

29.03.1984 

12.08.1985 

05.05.1986 

06.06.1986 

2004.1988 

13,03.1992 

1303.1992 

06.11.1992 

24.05.1994 

13.11.1994 

01.10.1995 

01.10.1995 

27.06.1998 

27.06.1998 

17.08.1999 

7,08.1999 

7.08.1999 

7.08.1999 

7.08.1999 

7,08.1999 

7.08.1999 

7,08.1999 

7.08.1999 

7.08.1999 

7.08.1999 

7.08. 1999 

7.08.1999 

708.1999 

708.1999 

7.08.1999 

708 1999 

7.08.1999 

7.08.1999 

708.1999 

7.08.1999 

7.08.1999 

2.11.1999 

2.11.1999 

2.11.1999 

Earthquake 

DENZL 

CERKE 

ZMR 

DURSUNBEY 

BGA 

EGA 

EGA 

HORASAN-NARMAN 

BALQCESR 

K I 

MALATYA 

SORGO (MALATYA ) 

MURADYE 

ERZNCAN 

ERZNCAN 

ZMR 

CRT 

KOYCE Z 

DNAR 

DNAR 

ADANA-CEYHAN 

ADANA-CEYHAN 

KOCAEL 

KOCAEL 

KOCAEL 

KOCAEL 

KOCAEL 

KOCAEL 

KOCAEL 

KOCAEL 

KOCAEL 

KOCAEL 

KOCAEL 

KOCAEL 

KOCAEL 

KOCAEL 

KOCAEL 

KOCAEL 

KOCAEL 

KOCAEL 

KOCAEL 

KOCAEL 

KOCAEL 

KOCAEL 

DOZCE 

DOZCE 

DGZCE 

Mw r«, (km) 

-_. 

5.4 

5.5 

5.3 

6.0 

6.1 

6.2 

6.5 

4.5 

4.9 

6.0 

6.0 

5.0 

6.9 

6.9 

6.1 

5.4 

5.2 

6.4 

6.4 

6.3 

6.3 

7.4 

7.4 

7.4 

7.4 

7.4 

74 

7.4 

7.4 

7,4 

7.4 

7.4 

7.4 

7.4 

7.4 

7.4 

7,4 

7.4 

7.4 

7.4 

7.4 

74 

7.4 

7.1 

7.1 

7 1 

15.2 

46.0 

1.2 

10.3 

577 

48.7 

75.0 

25.0 

2.4 

80.8 

296 

34.7 

373 

65.0 

5.0 

41.0 

20.1 

17.4 

3 0 

46.2 

SO.l 

28.0 

55,0 

81.0 

n.o 

116.0 

15.0 

32.0 

49.0 

8.0 

30.0 

140.0 

3.2 

150.0 

17.0 

82.5 

116.0 

72.0 

63.0 

3.3 

63.0 

43.0 

71.0 

81.0 

20.4 

8.2 

30.9 

Recording Station 

Demzli: Mcteoroloji stasyonu 

Cerke Meteoroloji stasyonu 

zmir: Meteoroloji btasyonu 

Dursunbey Kandilli Goziem stasyonu 

Edincik: Kandilli Goziem stasyonu 

Gooen: Meteorolojt stasyoou 

Tekirda Meteoroloji stasyonu 

Horasan- Meteorolqji stasyonu 

Bal kesir Meteoroloji stasyonu 

Ki Meteoroloji stasyonu 

Golba . Devlet Hastanesi 

GSIba Devlet Hastanesi 

Muradiye: Meteoroloji stasyouu 

Refahiye: Kdymakaml k Btnas 

Erztncao: Meteoroloji stasyonu 

Ku adas Meteoroloji stasyonu 

Foca: Gdmruk Mfldiirlu a 

Kdyce iz: Meteoroloji stasyoau 

Dinar: Meteotoloji stasyoau 

Qardak.-Sa Ik Oca 

Mersm: Meteoroloji stasyonu 

Ceyhan: PTT Mad. 

Bursa: Sivtl Sav. Mud. 

Cekraecc: Niikieer Santral Bn. 

DQzce: Meteoroloji stasyonu 

Ere li: Kaymakaml k Bn. 

Gebze. Tiibitak Marmara Ara Mer. 

Go'ynuk. Devtet Hastanesi 

stanbul: Bay ad rl k ve skaa Miid. 

zmit: Meteoroloji stasyonu 

zntk: Karayollar efli i 

Kiitahya: Sivil Savunma MQd. 

Sakarya: Bay nd r! k vc skan Mud. 

Tekirda • Hafctmct Kona 

Dar ca: Arcelik Arge Bn. 

Ambarl : Termik Santral 

M. Ere hsi: Bota Gas Temuaali 

Ye ilfcdy: Havahman 

4. Levent: Yap Kredi Plaza 

Yar mca: Petkira Tesisten 

Fatifa: Fatifa TOrbesi 

Heybeliada; Sanatoryum 

Bursa: Tofa Fab. 

Cekmece: Nukleer Santral Bn. 

3olu: Bay nd rl k ve skan Mud. 

Duzce : Meteoroloji stasyonu 

Mudurnu. KavmakamI k Bmas 

Station 

Coordinates 

37.S140N-29.U20E 

40.8800N-32.9100E 

38.40QON- 27 1900E 

39 6700N- 28.5300E 

403600N-27.S900E 

40.0800N- 27 6800E 

40.9600N-2753QOE 

400400N-42 1700E 

39.6600N- 27.S600E 

39 3400N- 40 2800E 

377810N-376410E 

377810N-37.6410E 

39 0300N- 43.7000E 

39.90 ION- 38. 7690E 

39.7520N- 39 4870E 

37.36 ION- 27 2660E 

38.6400N-26.7700E 

36.9700N- 28.6940E 

38.0600N-30.1500E 

37 8250N- 29 66SOE 

36.8300N- 34.6500E 

370500N35SIOOE 

40.1830N-29.I310E 

40.9700N-28.7000E 

40.8500N-3U7QQE 

40.9800N-27.7900E 

40.8200N- 29.4400E 

40.3850N- 30.7340E 

41.0S80N-29.0130E 

40.7900N- 29.960QE 

4Q.437QN-296910E 

39.4190N-299970E 

40.7370N- 30 3840E 

40.9790N-27.5150E 

40.82360N-293607E 

40.9809N- 28.6926E 

40.99 19N-27.9795E 

40.9823N-28S199E 

4l.08HN-20.OtUE 

40 7639N-29 7620E 

41.0196N-23.9500E 

40.8688N- 29.0S75E 

40.2605N- 29.0680E 

40.9700N- 28.7000E 

40.7450N-31.6100E 

40.S500N-31.1700E 

40.4630N-31.1S20E 

Station 

Site Class 

Soil 

Soft Soil 

Soft Soil 

Rock 

Rock 

Soft Soil 

Rock 

Soft Soil 

Soft Soil 

Soil 

Rock 

Roclc 

Rock 

Sort Soil 

Soil 

Soft Soil 

Rock 

Soft Soil 

Soft Soil 

Soil 

Soft Soil 

Soft Soil 

Soft Soil 

Soil 

Soft Soil 

Soil 

Rock 

Rock 

Rock 

Rock 

Soft Soil 

Soil 

Rock 

Rock 

Sod 

Soft Soil 

Soil 

Soil 

Rock 

Soil 

Soft Soil 

Rock 

Soft Soil 

Soil 

Soft Soil 

Soft Soil 

Soft Soil 

Peak Hor. Ace. (mg) 

_ N-S E-W 

348.53 

36.03 

391.41 

232.29 

5344 

50.11 

2939 

15026 

223.89 

163.06 

11470 

68.54 

4950 

6721 

404.97 

83,49 

36.06 

72.79 

288.30 

65.07 

119.29 

22342 

54.32 

1 18.03 

314.38 

90.36 

26482 

13769 

60.67 

171.17 

91.89 

50.05 

40704 

12979 

211 37 

252.56 

98.88 

90.21 

41.08 

23022 

18939 

56.15 

100.89 

177.31 

739.56 

40769 

120.99 

290.36 

38.94 

125.40 

288.25 

4651 

4677 

3491 

173 30 

128.97 

8909 

7604 

3443 

51.18 

85.93 

470.92 

71.80 

4980 

96.51 

26995 

61.30 

132.12 

273.55 

45.81 

8961 

373.76 

101.36 

14 ( 45 

117.90 

42.66 

224.91 

12332 

59.66 

128.33 

133.68 

186.04 

87 10 

84.47 

35.52 

322.20 

161.87 

1 10.23 

100.04 

132.08 

805.38 

513.78 

58.34 

but there are other attributes that must be mentioned. The first is our omission of 

aftershock data. Most of these come from the two major 1999 events, and contain freefield 

data that we did not wish to cornrningle with the rest of the set. We also omitted 

the few records for which the peak acceleration caused by the main shock is less than 

about 0.04 g. Our entire, non-discriminated ensemble is shown in Figure 2. 

When we consider the effects of geological conditions on ground motion and 

response spectra, the widely accepted method of reflecting these effects is to classify the 

recording stations according to the shear-wave velocity profiles of their substrata. 

Unfortunately, the actual shear-wave velocity and detailed site description are not 

available for most stations in Turkey. For this reason, we estimated the site classification 

by analogy with infonnation in similar geologic materials. The type of geologic material 

underlying each recording site was obtained in a number of ways: consultation with 

geologists at Earthquake Research Division of Ministry of Public Works and Settlement, 

various geologic maps, past earthquake reports and geological references prepared for

36 

Turkey In the hght of this information we divided soil groups for Turkey into three m 

ascending order for shear velocity soft soil, soil, and rock The average shear-wave 

velocities assigned for these groups are 200, 400 and 700m/s, respectively The 

distribution of the records with respect to magnitude and distance plotted by type of 

faulting is shown in Figure 3 

su 

3 

S 

SJD 

5 

a 

CJ 

£ o 

o 

-3 

•5 

I 

a 

o 

8 0 - 

75 - 

70 - 

6 5 - 

60 • 

55 - 

| 55 - 

1 

80 

75 - 

o 

I 70 - 

Sea- 

S 60- 

• * • • • • • • 

• * 

* • • » 

' 

50 • 

• 

ROCK 

45 i 1 ~~T~ 

1 10 100 

o n 

O U T^ 

75 - 

70 - 

65 - 

60 - 

50 - 

4 fi - 

* » •• • • 

• 

- - • 

10 100 

*• 

9 

SOIL 

I 55 ' 

I 50- 

45 - 

40 

10 

Closest Distance (km) 

SOFT SOIL 

1QQ 

Figure 1 Distribution of records in the database in terms of magnitude, distance and 

local geological conditions 

Table 2 Earthquakes Used in the Analysis

37 

Date 

8/19/1976 

10/5/1977 

12/16/1977 

7/18/1979 

7/5/1983 

10/30/1983 

3/29/1984 

8/12/1985 

5/5/1986 

6/6/1986 

4/20/1988 

3/13/1992 

11/6/1992 

5/24/1994 

11/13/1994 

10/1/1995 

6/27/1998 

8/17/1999 

11/12/1999 


DENIZLI 

CERKJES 

IZMIR 

DURSUNBEY 

BIGA 

HORASAN-NARMAN 

BALIKESIR 

KIGI 

MALATYA 

SURGU(MALATYA) 

MURADIYE 

ERZINCAN 

IZMIR 

GIRIT 

KOYCEGIZ 

DINAR 

ADANA-CEYHAN 

KOCAELI 

DUZCE 

Fault Type 

Normal 

Strike-Slip 

Normal 

Strike-Slip 

Reverse 

Strike-Slip 

Strike-Slip 

Strike-Slip 

Strike-Slip 

Strike-Slip 

Strike-Slip 

Strike-Slip 

Normal 

Normal 

Normal 

Normal 

Strike-Slip 

Strike-Slip 

Strike-Slip 

M w 

53 

54 

55 

53 

60 

65 

45 

49 

60 

60 

50 

69 

61 

54 

52 

64 

63 

74 

71 

NUMBER OF RECORDINGS 

Soft Soil Soil Rock 

2 

2 

2 

2 

2 

2 

2 

2 

2 

2 

4 

12 

6 

Total 40 

2 

2 

2 

16 

24 

2 

4 

2 

2 

2 

2 

15 

29 

ALL DATA 

01 - 

* • 

001 

10 100 

Figure 2 Distribution of the larger maximum horizontal acceleration of either component 

versus distance

38 

-) 

1 

*I 

C5 

IS 

"c 

0» 

s 

o 

o u - 

75 - 

45 - 

Q A 

75 - 

70 - 

65 - 

60 - 

55 - 

m * * •* •* •»•••• • i» 

• • • • • 

• • • 

*• 

«h • 

* •• 

STRIKE SLIP 

1 10 100 

_ 

•. — ~ 9 

„ _ __^ 

• » « _ 

50 - 

_ _ _ + 

45 • 

. 

NORMAL 

4 0 - 

1 10 100 

P n - - - - - - - - 

"5 OJ) 

C8 

s s 

o 

s 

50 - 

• • • 

70- 

So 65 - 

1 60 - 

s 

Q 55- 

S 

3 

50- 

75- 

65- 

60- 

55- 

45- 

4 0 - 

REVERSE 

1 10 100 


Figure 3. Distribution of records in the database in terms of magnitude, distance and 

type of faulting 

Attenuation Equation Development 

Attenuation relationships were developed by using the same general form of the 

equation proposed by Boore et al. (1997). The ground motion parameter estimation 

equation is as follows: 

lnY = bi + b 2 (M - 6) + b 3 (M - 6) 2 + b 5 In r + b v In (V s / V A 

(1)

39 

(2) 

Here Y is the ground motion parameter (peak horizontal acceleration (PGA) or 

pseudo spectral acceleration (PSA) in g); M is (moment) magnitude; r ci is closest 

horizontal distance from the station to a site of interest in km; V s is the shear wave 

velocity for the station in m/s; bi, 02, b 3 , bs, h, by, and V A are the parameters to be 

determined. Here h is a fictitious depth, and V A a fictitious velocity that are determined 

by regression. The coefficients in the equations for predicting ground motion were 

determined by using nonlinear regression analysis. Nonlinear regression is a method of 

finding a nonlinear model of the relationship between the dependent variable and a set of 

independent variables. Unlike traditional linear regression, nonlinear regression can 

estimate models with arbitrary relationships between independent and dependent 

variables. This is accomplished using iterative estimation algorithms. This exercise was 

performed separately on PGA and on PSA data at each oscillator period considered 

(total of 46 periods from 0.1 to 2.0s.). 

The procedure that we have used to develop the attenuation curves consists of two 

stages (Joyner and Boore, 1993). In the first, attenuation relationships were developed 

for PGA and spectral acceleration values by selecting the acceleration values in the 

database as maximum horizontal components of each recording station. Then, a 

nonlinear regression analysis was performed. In the next stage, random horizontal 

components were selected for the acceleration values in the database and regression 

analyses were applied. The results were compared for PGA, 0.3 s and 1.0 s PSA cases, 

and it was concluded that selection of maximum, rather than of random, horizontal 

components did not yield improved estimates and smaller error terms. This issue is 

taken up again in the section on comparisons of our results with other relations. 

The coefficients for estimating the maximum horizontal-component pseudoacceleration 

response by Equation (1) are given in Table 3. The resulting parameters can 

be used to produce attenuation relationships that predict response spectra over the full 

range of magnitudes (M w 5 to 7.5) and distances (r d ) up to 150 km. The results were 

used to compute errors for PGA and PSA at individual periods. The standard deviation 

of the residuals, a, expressing the random variability of ground motions, is an important 

input parameter in probabilistic hazard analysis. In this study, the observed value of a 

lies generally within the range of 0.5 to 0.7. The calculated attenuation relationships for 

PGA for rock, soil and soft soil sites are shown in Figures 4 through 6.

40 

Table 3. Attenuation Relationships of Honzontal PGA and Response Spectral 

Accelerations (5% damping) 

ln(Y) = b1 + b2 (M - 6) + 

Period b1 b2 

PGA 

0.10 

0.11 

0.12 

0.13 

0.14 

0.15 

0.16 

0.17 

0.18 

0.19 

0.20 

0.22 

0.24 

0.26 

0.28 

0.30 

0.32 

0.34 

0.36 

0.38 

0.40 

0.42 

0.44 

0.46 

0.48 

0.50 

0.55 

0.60 

0.65 

0.70 

0.75 

0.80 

0.85 

0.90 

0.95 

1.00 

1.10 

1.20 

1.30 

1.40 

1.50 

1.60 

1 70 

1.80 

1.90 

2.00 

-0.682 

-0139 

0.031 

0.123 

0138 

0100 

0.090 

-0.128 

-0.107 

0.045 

0.053 

0.127 

-0.081 

-0.167 

-0.129 

0.140 

0.296 

0.454 

0.422 

0.554 

0.254 

0.231 

0.120 

0.035 

-0.077 

-0.154 

-0078 

-0.169 

-0.387 

-0.583 

-0.681 

-0.717 

-0.763 

-0.778 

-0.837 

-0.957 

-1.112 

-1.459 

-1.437 

-1.321 

-1.212 

-1.340 

-1.353 

-1.420 

-1.465 

-1.500 

-1.452 

0.253 

0.200 

0.235 

0.228 

0.216 

0.186 

0.210 

0.214 

0.187 

0.168 

0.180 

0.192 

0.214 

0.265 

0.345 

0.428 

0.471 

0.476 

0.471 

0.509 

0.499 

0.497 

0.518 

0.544 

0.580 

0.611 

0.638 

0.707 

0.698 

0.689 

0.698 

0.730 

0757 

0810 

0.856 

0.870 

0.904 

0.898 

0.962 

1.000 

1.000 

0.997 

0.999 

0.996 

0.995 

0999 

1 020 

b3 (M - 6) 2 

b3 

0036 

-0.003 

-0.007 

-0.031 

-0.007 

0.014 

-0.013 

0.007 

0.037 

0043 

0.063 

0065 

0.006 

-0.035 

-0.039 

-0.096 

-0.140 

-0.168 

-0.152 

-0.114 

-0.105 

-0.105 

-0.135 

-0.142 

-0.147 

-0.154 

-0.161 

-0.179 

-0.187 

-0.159 

-0.143 

-0.143 

-0.113 

-0.123 

-0.130 

-0.127 

-0.169 

-0.147 

-0.156 

-0.147 

-0.088 

-0.055 

-0.056 

-0.052 

-0.053 

-0.051 

-0.079 

+ b5inr+bVln(V s/V 

b5 

b v 

-0.562 

-0.553 

-0.573 

-0.586 

-0.590 

-0.585 

-0.549 

-0.519 

-0.535 

-0.556 

-0.570 

-0.597 

-0.532 

-0.531 

-0.552 

-0.616 

-0.642 

-0.653 

-0.651 

-0692 

-0.645 

-0.647 

-0.612 

-0.583 

-0.563 

-0.552 

-0.565 

-0.539 

-0.506 

-0.500 

-0.517 

-0.516 

-0.525 

-0.529 

-0.512 

-0.472 

-0.443 

-0.414 

-0.463 

-0.517 

-0.584 

-0.582 

-0.590 

-0.582 

-0.581 

-0.592 

-0.612 

-0.297 

-0.167 

-0.181 

-0.208 

-0.237 

-0.249 

-0.196 

-0.224 

-0.243 

-0.256 

-0.288 

-0.303 

-0.319 

-0.382 

-0.395 

-0369 

-0.346 

-0.290 

-0300 

-0.287 

-0.341 

-0.333 

-0.313 

-0.286 

-0.285 

-0.293 

-0.259 

-0.216 

-0.259 

-0.304 

-0.360 

-0.331 

-0.302 

-0.283 

-0.252 

-0.163 

-0.200 

-0.252 

-0.267 

-0.219 

-0.178 

-0.165 

-0.135 

-0.097 

-0.058 

-0.047 

-0.019 

A ) with r = 

V A 

1381 

1063 

1413 

1501 

1591 

1833 

1810 

2193 

2433 

2041 

2086 

2238 

2198 

2198 

2160 

2179 

2149 

2144 

2083 

2043 

2009 

1968 

1905 

1899 

1863 

1801 

1768 

1724 

1629 

1607 

1530 

1492 

1491 

1438 

1446 

1384 

1391 

1380 

1415 

1429 

1454 

1490 

1513 

1569 

1653 

1707 

1787 

( re! 2 + h 2 ) 1 

h 

4.480 

3.760 

3.890 

4.720 

5.460 

4.980 

2.770 

1.320 

1.670 

2.440 

2.970 

3.480 

1.980 

2.550 

3.450 

4.950 

6.110 

7.380 

8300 

9.180 

9.920 

9.920 

9.090 

9.250 

8980 

8.960 

9.060 

8.290 

8.240 

7.640 

7.760 

7.120 

6.980 

6.570 

7.250 

7.240 

6.630 

6.210 

7.170 

7.660 

9.100 

9.860 

9.940 

9.550 

9.350 

9.490 

9.780 

12 

a 

0.562 

0.621 

0.618 

0.615 

0.634 

0.635 

0.620 

0.627 

0.621 

0.599 

0.601 

0.611 

0.584 

0.569 

0.549 

0.530 

0.540 

0.555 

0.562 

0.563 

0.562 

0.604 

0.634 

0.627 

0.642 

0.653 

0.679 

0.710 

0.707 

0.736 

0.743 

0.740 

0.742 

0.758 

0.754 

0.752 

0.756 

0.792 

0.802 

0.796 

0.790 

0.788 

0787 

0.789 

0.827 

0.864 

0895

41 

1 2 3 4567810 20 3040 SO 100 200 


Figure 4. Peak acceleration versus distance for magnitude 5.5, 6.5 and 7.5 

earthquakes at rock sites 

2 3 4 557810 20 3040 60 100 200 



earthquakes at soil sites

42 

2 3 4 5S7810 20 3040 60 100 200 



earthquakes at soft soil sites 

Comparison with Other Recent Attenuation Relationships 

The estimate equations developed in this study were compared to those recently 

developed by Boore et al. (1997), Campbell (1997), Sadigh et al. (1997), Spudich et al. 

(1997) and finally Ambraseys et al. (1996), The equations of Boore et al. and Ambraseys 

et al. divided site classes into four groups according to shear wave velocities. 

Campbell's equations pertain to alluvium (or firm soil), soft rock and hard rock. Sadigh 

et al, and Spudich et al. state that their equations are applicable for rock and soil sites. 

The attenuation of PGA and PSA at 0.3 and 1.0 s for M w = 7.4 for rock and soil sites 

are compared in Figures 7-9, respectively. The measured database points from the 

Kocaeli event are also marked on these curves to illustrate how well they fit the 

estimates. The differences in the curves are judged to be reasonable because different 

databases, regression models and analysis methods, different definitions for source to 

site distance and magnitude parameters among the relationships are contained m each 

model. 

For some parameters and especially for PGA, there are numerous published 

attenuation equations for use in any particular engineering application. Atkinson and 

Boore (1997) showed the differences between attenuation characteristics in western and 

eastern USA for stable intraplate and interplate regions. Nevertheless, differences among 

attenuation of strong motions from one region to another have not been definitely 

proven. Because of this reason it is preferable to use attenuation equations that are based 

on the records taken from the region in which the estimation equations are to be applied. 

Sensors comprising the national or other strong motion networks in Turkey are 

oriented so that their horizontal axes match the N-S and the E-W directions. Whereas 

Figure 2 illustrates the larger of these two components as a function of distance, it may 

not represent the largest horizontal acceleration that occurred before the cessation of the 

ground motion. The value of the absolute maximum acceleration in whichever direction 

can be determined by monitoring through a simple book-keeping procedure for the size

43 

of the resultant horizontal component, and then resolving all pairs to the direction of that 

largest component once it is known. At variance with the customary practice, we call 

this component the "random" horizontal component. In Figure 10, the difference in the 

predictive power of the regression equations derived from both of these definitions is 

illustrated for M w = 7.4, and compared against the Kocaeli measurements. We believe 

that both sets yield essentially the same results. With the differences between the mean 

or the standard deviation curves substantially less than the value of In (a) itself, an 

improvement in accuracy does not appear to be plausible between the definitions of 

maximum horizontal acceleration. 

1 oo 

ra 

U) 

Q. 

KOCAELI DATA (Max.Hor Comp ) 

Max.HorComp 

Booreetal (1997) 

+/-1 Sigma 

001 

200 

Ambraseysetal (19S6) 

Spudichetat (1997) 

Sadighetai(1997) 

Campbell (1997) 

Closest Distance (km)

44 

100 

(Q 

ra 

a. 

KOCAELI DATA (Max H Comp ) 

Max Hor Comp 


+/ 1 Sigma 

001 

2 3 4567810 20 3040 60 100 200 


Ambraseys et ai (1996) 

Spudich et al (1997) 

Sadigh et al (1997) 

Campbell (1997) 

Figure 7 Peak acceleration versus distance for magnitude 7 4 earthquake at 

rock and soil sites

45 

200 

006 

005 

004 

003 

002 

Rock, Mw = 7 4 

O KOCAELI DATA (Max Hor Camp ) 

Max Hor Camp 


'"•"•"•"•»• +/ 1 Sigma 

Sadighetal(1997) 

1 2 3 4567810 20 3040 60 100 200 


200 

V) 

005 • 

004 • 

003 - 

Soil, Mw = 7 4 

O KOCAEL! DATA (Max Hor Comp ) 

- — Max Hor Comp 

Booreeta! (1997) 

- +/ 1 Sigma 

Sadighatal(1997) 

001 

1 2 3 4567810 20 3040 60 100 200 


Figure 8 Spectral acceleration at T = 0 3 s versus distance for a magnitude- 7 4 earthquake 

at rock and soil sites

46 

100 

080 

070 

060 

050 

040 

030 

020 

010 

008 

007 

006 

005 

004 

003 

002 

Rock, Wlw - 7.4 

O KOCAEL! DATA (Max Her Comp ) 

• Max Hor Comp 


,„....!!„..., +/„ 1 Sigma 

Sadighetal(l997) 

001 

1 2 3 4567810 20 3040 60 100 200 


200 

100 

m060 

050 

040 

030 

020 

flj 

CO 

010 • 

m- 

006 - 

005 

004 

003 

002 

Soil, Mw = 7.4 

O 

KQCAELl DATA (Max Her Comp) 

•• Max Hor Comp 

Baoreetai (1997) 

—"—a +/-1 Sigma 

Sadigh et ai (1997) 

001 

1 2 3 4567810 20 3040 60 100 200 


Figure 9 Spectral acceleration at T = 1.0 s versus distance for a magnitude-7.4 earthquake 

at rock and soil sites

47 

en 

o_ 

KOCAELI DATA (Random Hor Comp) 

003 

002 

Random Hor Comp 


Max Hor Comp 

•*•/-1 Sigma (Ran Hor Comp) 

+/-1 Sigma (Max Hor Comp } 

001 

1 2 3 4567810 20 3040 60 100 200 


Figure 10 Differences caused by using the larger of the two horizontal components or the 

component in the direction of the largest resultant 

Uncertainty and Reliability 

Uncertainty is a condition associated with essentially all aspects of earthquake related 

science and engineering. The principle sources of uncertainty lie in the characterization 

of site geology, calculation of closest distances, determination of seismic shaking 

properties, and in the geotechrucal properties of earthquake motion monitoring sites. The 

regression analysis is based on stochastic analysis method thus the obtained attenuation 

formula contains unavoidable errors. These uncertainties, for the most part stemming 

from the lack of and/or the imperfect reliability of the specific supporting data available, 

affect all analytical methods and procedures applied to the derivation of all 

aforementioned parameters. 

The attenuation relationships presented in this study cannot, and do not, eliminate 

these uncertainties. However through the use of nonlinear regression analysis, it 

provides a more sophisticated and direct approach to address the uncertainties than do 

traditional linear analysis procedures. The results we have presented in tabular and 

graphical form become meaningful only in the context of the error distributions that are 

associated with each variable. In general, our results possess larger deviations in 

comparison with, e.g., Boore et al. (1997). This is plausible because of the smaller 

number of records from which they have been derived. In view of the limited number of 

records utilized in this study it may not be appropriate to expect the distributions to 

conform to the normal distribution. We do this only as a vehicle that permits a direct 

comparison to be made between our results and those of Boore et al. (1997),

48 

Conclusions 

The recommended attenuation relationships presented in detail in this paper through 

Table 3 and illustrated in Figures 4-6 are considered to be appropriate for the estimation 

of horizontal components of peak ground acceleration, and 5 percent damped pseudo 

acceleration response spectra for earthquakes with magnitude in the range M w 5 to 7.5 

and r c!

Atkinson G.M., Boore D.M., Some Comparisons between Recent Ground Motion 

Relations, Seismological Research Letters, Vol. 68 No. 1, pp. 24-40 

January/February, 1997. 

Boore D.M., Joyner W.B., Fumal T.E., Equations for Estimating Horizontal Response 

Spectra and Peak Acceleration from Western North American Earthquakes: A 

Summary of Recent Work, Seismological Research Letters, Vol. 68, No. 1, 

pp.128-153, January/February, 1997. 

Campbell K.W., Empirical Near Source Attenuation Relationships for Horizontal and 

Vertical Components of Peak Ground Acceleration, Peak Ground Velocity, and 

Pseudo-Absolute Acceleration Response Spectra, Seismological Research Letters, 

Vol. 68, No. 1, pp. 154-179, January/February, 1997. 

Hanks T., and Kanamori H., A Moment Magnitude Scale, J.Geophys.Res., Vol. 84, No. 

2, pp. 348-2-350, 1979. 

Joyner B.W., Boore M.D., Methods for Regression Analysis of Strong-Motion Data, 

Bulletin of Seismological Society of America, Vol. 83, No. 2, p. 469-487, April, 

1993. 

Sadigh K., Chang C.Y., Egan J.A., Makdisi F., Youngs R.R., Attenuation Relationships 

for Shallow Crustal Earthquakes Based on California Strong Motion Data, 

Seismological Research Letters, Vol. 68, No. 1, pp. 180-189, January/February, 

1997. 

Spudich P., Fletcher IB., Hellweg M., A New Predictive Relation for Earthquake 

Ground Motions in Extensional Tectonic Regimes, Seismological Research 

Letters, Vol. 68, No. 1, pp. 190-198, January/February, 1997. 

Wells D.L., Coppersmith K.J., New Empirical Formula among Magnitude, Rupture 

Length, Rupture Width, Rupture Area, and Surface Displacement, Bulletin of 

Seismological Society of America, Vol. 84, No. 4, pp. 974-1002, August, 1994. 

49




5 j 

DEVELOPMENT OF JSSI MANUAL 

FOR PASSIVE CONTROL OF BUILDINGS 

(Part 1: 

Background, Scope, and Design Concept) 

Kazuhiko KASAI 

Professor, Structural Engineering Research Center, 

Tokyo Institute of Technology, Yokohama, JAPAN, and 

Chairman, Response Control Cpmrnittee, 

Japan Society of Seismic Isolation (JSSI) 

ABSTRACT 

Japan Society of Seismic Isolation (JSSI), a key organization for promotion and codification for the 

base-isolation scheme in Japan, has decided to initiate a new activity regarding passive control 

scheme. The JSSI Response Control Committee is currently formulating the Manual for Design 

and Construction of Passively-Controlled Buildings. This paper, together with the companion 

paper (Ref. 15), describes such an effort of the committee. In particular, Part 1 addresses 

background and scopes of the manual as well as design methods. 

INTRODUCTION 

Passive control scheme has established its status as a viable means to enhance seismic performance 

of buildings [1-14]. Social desire for adopting this scheme as well as base isolation scheme has 

increased "considerably after the 1995 Kobe earthquake. Up to the year 2002, the number of 

passively controlled buildings has increased to about three hundred in Japan. For the sake of 

further growth in this technology, it is necessary to promote understanding of the passive control 

schemes, as well as to create a uniform basis for assessment of the various stages to be followed 

during the design and construction process. 

Pursuant to this need, the Japan Society of Seismic Isolation (JSSI), a key organization for 

promotion and codification for the base-isolation scheme in Japan, has decided to initiate a new 

activity regarding passive control scheme: the JSSI Response Control Committee is currently 

formulating the Manual for Design and Construction of Passively-Controlled Buildings [3-14]. 

The manual will refer to the mechanism, design, fabrication, testing, quality control, and analytical 

modeling of various types of passive control devices, as well as design, construction, and analysis 

of passively controlled buildings. 

The manual is being developed by researchers, designers, and more than twenty device 

manufacturing companies. This paper, together with a companion paper (Ref. 15) describes such 

an effort. In particular, this paper will address background and scope of the manual, as well as 

design and analysis. Details will be explained in the other papers (Ref. 16-27) that are also to be 

presented in the present and following sessions in this congress.

52 

BACKGROUND AND SCOPES 

In order to justify the use of passive control for producing higher seismic performance, it has 

become necessary to indicate the control effectiveness as well as expected performance to the 

building officials* owners, and/or users. In this regard, it is necessary to develop a common 

standard for design, construction, and quality for this technology. The JSSI manual is intended to 

provide such a standard. 

However, such development requires a great caution, since the history of implementing passive 

control is short. This new system has experienced neither a major earthquake nor frequent minor 

earthquakes, and database for actual performance is poor. In addition, it is not yet exposed to the 

long-term use, and durability of the device as well as frame has not been attested in the actual 

environment. Moreover, analysis and performance prediction are based on the extrapolation of 

limited experimental data that are typically created from testing of reduced-scale devices and 

systems under highly idealized load and boundary conditions. 

The JSSI manual is being developed considering the above-mentioned circumstances. It shall 

clarify the ranges of the device and system performance, as well as limitations of the analysis and 

prediction methods as precisely as possible. Furthermore, the manual will describe broadly the 

important matters to examine in each stage of a design, manufacture, and construction of various 

components of the system. In this manner, the manual is expected to promote mutual 

understanding and common recognition by the structure designer, manufacturer, and builder, which 

could result in better assurance for the stipulated performance of a building,. 

Note also that the manual does not intend to restrain the new idea about passive control, and it aims 

at offering the basis needed to enable flexible and creative thinking on application of the 

technology. 

JSSI RESPONSE CONTROL COMMITTEE 

JSSI established the Response Control Committee in February, 2000, for the purpose of 

investigating and exchanging information on various types of response control schemes such as 

active,"passive, and base isolation systems. Since then, this committee has been growing, and as 

indicated in Table 1, it currently consists of 3 subcommittees and 8 working groups of 88 members. 

The passive control subcommittee, device standard subcommittee, and their 8 working groups are 

developing the manual currently (Table 1). The task is a combined effort by the researchers, 

structural engineers, and device manufacturers. 

Table I. 

JSSI Response Control Committee, Subcommittees, and Working Groups 

Committee 

Response Control 

Subcommittee 

Active Control 

Passive Control 

Device Standard 

Working Group 

— 

Passive Device Analysis 

Passive System Analysis 

System Design Methodology 

Oil Damper 

Viscous Damper 

Viscoelastic Damper 

Steel Damper 

Device Design

53 

CONTENTS OF JSSI MANUAL 

Following summarizes prospective contents of the manual, and so far it is partially completed at the 

present stage: 

Chap. 1 General 

Definitions and Terminology, Classifications of Passive Control Systems, Performance 

Parameters of Control Devices, Applicability of Manual Contents. 

Chap. 2 Target Control Performance 

Earthquake Damage and Target Performance. Performance Grades of Passive Control System. 

Chap. 3 Mechanism of SDQF Passive Control System 

Effects of Supplemental Stiffness and Viscosity, Velocity -, Deformation -, and 

Hybrid-Dependencies, Factors Governing Passive Control Effectiveness, Simplified Response 

Prediction Utilizing Response Spectra, Use of Performance Curves. 

Chap. 4 Design of MDOF Passive Control System 

Mechanism of MDOF Passive Control, Layout of Control Devices, Simplified Response 

Prediction, Static Equivalent Force Considering Elastic and Viscous Forces, Limit Force 

Design and Tri-Service Manual Methods. 

Chap. 5 Dynamic Time History Analysis Methods 

Combined Effect of Device and Supporting Member, Device Modeling Schemes, Algorithms, 

and Accuracy, System Modeling Schemes and Accuracy. 

Chap. 6 Design of Control Devices 

Characteristics and Applicable Range, Test and Evaluation Methods, Limit Condition and 

Design of Device, Limit Condition and Design of Connection Elements. 

Chap. 7 Specifications and Examinations of Device 

Device Performance Specifications, Device Performance Examinations. 

Chap. 8 Quality Control Scheme 

Manufacturing Process and Quality Control Scheme, Items Subjected to Examinations, 

Examinations Prior to Delivery, Installation, and Project Completion. 

Chap. Q Operation Scheme during Construction 

Operational Attentions, Protection and Curing During Operation, Checking Items for 

Connection Elements. 

Chap 10. Maintenance Scheme 

Basic Agreement, Durability of Device and Connection Elements, Maintenance Plans. 

APPLICABILITY OF MANUAL 

Four types of devices are considered in the manual. They are oil dampers, viscous dampers, 

viscoelastic dampers, and steel dampers such as shown in Fig, 1. 

Viscous damper produces the hysteresis loop of combined ellipse and rectangle. The material 

used therein is typically silicon fluid, and its resistance against flow produces the damper force. The 

damper possesses configurations of vertical panel, box, or cylinder. 

Oil damper produces the hysteresis loop of ellipse. The material used therein is oil, and its 

resistance against flow at orifice produces the damper force. The damper possesses the

configurations of cylinder, and it is usually provided with a relief mechanism that prevents increase 

in force. 

Viscoelastic damper produces the hysteresis loop of inclined ellipse. In some material, the 

hysteresis is close to bilinear especially when it is under large deformation. The material used is 

polymer composite of acryl, butadiene, silicon, or others, and resistance is produced from the 

molecular motion when subjected to loading. Typical damper has configurations of vertical panel 

or tube, but it could be designed for many other configurations as well. 

Steel damper produces bi-linear hysteresis. The material is steel, but those using lead or friction 

pad can exhibit similar behavior. These materials produce elasto-plastic resistance due to yielding 

or slipping. Typical damper has configurations of vertical panel or tube, but it could be designed 

for many other configurations as well. This damper is the least expensive among the four types. 

; F=Cif* 

• 

Combined Ellipse mid 

Rectangle Hysteresis 

Silicon Fluid etc. 

Shear Resistance, 

Flow Resistance 

Plane, Box, and 

Tube Shapes 

JF-Ca 

Ellipse Hysteresis 

Oil 

Orifice Flow 

Resistance 

Tube Configuration 

F = K(W it* C (M) '• ii 

Inclined Ellipse 

Hysteresis 

Acryl, Butadiene etc. 

Shear Resistance 

Tube and Plane Shapes 

F=K*f(u) 

Bilinear H y $ te res is 

Steef, Lead, 

Friction Pad, etc. 

Yielding Resistance 

Slipping Resistance 

Tube and Plane Shapes 

Fig. 1 Four Types of Dampers Considered in Manual 

Many framing types are considered in this manual, and they are categorized into directly connected 

system, indirectly connected system, and special system (Fig. 2). 

Directly connected system is wall type, brace type, or shear link type. In such a system, both ends 

of the damper are directly connected to the locations whose differential displacements define the 

story drifts. In other words, the damper is effective in directly controlling the drifts of the frame. 

Indirectly connected system is stud type, bracket type, or connector type. In such a system, both 

ends of the damper are connected to the beams and columns that could deform locally and absorb a 

portion of the deformations that otherwise could be imposed to the damper. Thus, the damper is 

generally less effective than those of the directly connected system mentioned above. However, 

since the system has an advantage of offering greater freedom for architectural planning, it has been 

much favored currently by the structural engineers and architects in Japan. 

Special system considered herein is either column type or beam type. In such a system, the 

damper is inserted into intentionally disconnected zone of a beam or a column, and becomes a part 

of those members. Thus, it does not create any obstacle in the floor plan, but its control 

effectiveness depends on how rigid the rest of the frame is. Similarly to the indirectly connected 

system, the frame must be very stiff such that the deformation takes a place in the damper. Ref. 18 

for instance describes a real application of the column type, which turned out to be very effective in 

controlling both displacements and forces including uplift force of the foundation.

55 

Directly 

Connected 

System 

Wall lype Brace Type Shear Link Type 

Indirectly 

Connected 

System 

Stud Type Bracket Type Connector Type 

Column Type | 

Beam Type 

Special 

System 

Fig. 2 

Framing Types Considered in Manual 

BASIC DESIGN METHOD 

To date, design and performance prediction of passive control systems have typically been based on 

iteration involving extensive response time history analyses or equivalent static analyses using 

various types and sizes of dampers. The analysis methods are also different between the various 

systems; these make direct comparison of the systems difficult. Moreover, they offer limited 

information about the possible range of seismic performance variations and the complex 

interactions between the dampers, their supporting members, frame, seismic input, and response. 

The manual will propose a method to clarify the seismic performance of the various types by 

commonly expressing their response effectiveness as a continuous function of the structural and 

seismic input parameters. The method should promote understanding of the commonalities and 

differences between the various types, which have distinct energy dissipation mechanisms. It 

requires only simple calculations, and its prediction agrees well with the results of the extensive 

multi-degree-of-freedom dynamic analyses performed [1, 2]. The method can also find an 

optimum design solution to control both displacement and acceleration of the systems. The 

method is briefly explained below: 

Fig. 3 shows an example of such a method applied to the multi-story passive system using either a 

viscoelastic damper or elastoplastic damper. The method considers an equivalent 

single-degree-of-freedom system consisting of damper and supporting member such as brace that 

are connected in series, as well as a frame connected in parallel with these components. The 

parameters used to characterize the control effectiveness are: the 1st mode effective mass 

approximately equal to 0.8 times total mass in a case of regular building; elastic stiffness of frame, 

damper, and brace, and; ductility factor and loss factor that quantify the amount of energy 

dissipation by the elastoplastic damper and viscoelastic damper, respectively.

l 

56 

X 

xT / ^ 

V^—Ê-^/i / X\ - 

yX^ | E°0ar,oer i ' 

3r5LCS 

N 

*£-£"• = ' !/ 1. 

PASSIVELY-COTROLLED BUILDING 

EP-SYSTEM 

Kf = Frame Stiffness 

K d = Damper Elastic Stiffness 

K b = Brace Stiffness 

2 "i 

Kif 

f 

ÂVVv 

VE-SYSTEM 

ji d = EP Damper Ductility 

T| d = VTE Damper Loss Factor 

l 

Fig. 3 

Example Passi\e Control Systems and SDOF Representations 

F t j 4 ^:evss the chart created by Kasai's SDOF theory on passive control effectiveness (Refs. 1 and 

2T Tn.s chart can be used to'determine the desired ratio between the above-mentioned stiffness 

parameters as well as loss and ductility factors. For a given earthquake input, the peak displacement 

and base shear of the frame prior to damper installment can be predicted easily from the response 

spectrum. Then using the chart, for desired reduction ratios of the displacement and acceleration, 

one can determine the "necessary stiffness of the damper and brace relative to that of the given frame, 

as uel. as either ductility factor or loss factor of the damper. This chart was created by assuming 

that pseudo-velocity response is constant relati\e to the building period, a typical design assumption 

mace for a moderate to tall building in Japan. The design result for SDOF system can be equally 

applied to sizing of the dampers in the multistory case as well. The accuracy of the method for the 

system using \arious types of dampers has been examined by the System Design Methodology 

\\ jrking Group (Table 1). and the results are briefly explained in Ref. 16. 

O.Z 0.4 O.b OS 

DISP. REDUCTION RATIO 

•1 • | \ _>* U.23 

l«V.4-->" _, "^ i 0.5 

i 2 t|dKd J 'Ki > 

i — i- *——•—, 

i i — 

02 0.4 H6 03 

DISP. REDUCTION RATIO 

1 

-o— rjf-30 

Fig. 4 Example Design Chart for Passive Control Systems 

After the^ simplified procedure outlined abo\e, one can create a muiti-degree-of-freedom model, and 

may perform time-history analyses using a series of earthquake records in order to confirm or make 

modifications in design. The methods 10 create analytical model for the damper of the system are

57 

being developed [19-23] by the Passive Device Analysis Working Group, and Passi\e Svstem 

Analysis Working, Group, respectively (Table 1). Fig. 5 shows summary of the design procedures 

mentioned above. 

Given: Blclg. Config., Damper Type 

( Get 

(c 

\ 

t 

1 

Frame, Stiffness, Period etc. 

1 

1 

T • T ^1 

against EQ. Levels 

4 

Get: Damper ;MZ & ^ 

1 

Perform: Time History Analysis* 

nf Ettitlftinrr 

\ 

Y 

\1f\fla>l 

Evaluate: Building Performance 

l 

.N.-J- 

' j OK* 

| Yes 

END 1 

Fig. 5 

Summary of Damper and System Design Procedures 

CONCLUSIONS 

The JSSI Response Control Committee is currently formulating the Manual for Design and 

Construction of Passively Controlled Buildings. The manual refers to the mechanism, design, 

fabrication, testing, quality control, and analytical modeling of various passive control devices, as 

well as design, construction, and analysis of passively controlled buildings. 

The manual is being developed by researchers, designers, and more than twenty device 

manufacturing companies. This paper has explained background and scope of the manual as weH 

as design concept. A companion paper (Ref. 15) describes characteristics and implementations of 

the control devices. Other papers (Refs. 16-27) also explain some details of the manual contents. 

REFERENCES 

[i] KASAI, K, FU, Y, and WATANABE, A. (1998). Passive Control Systems for Seismic ^Damage 

Mfas&ion, Journal of Structural Engineering, American Society of Civil Engineers, 122:10, 501-512 

[2] FU, V., and KASAi, K. (1998). Comparative Study of Frames Using Viscoelastic and Viscous 

Dampers, Journal oj Structural Engineering, American Society of Civ il Engineers, 122:10, 513-522 

Related Papers by Response Control Committee: 2001 Passive Control Symposium, Structural 

Engineering ReseaVch Center, Tokyo Institute of Technology, JAPAN, Dec. 2001. 

[3] KASA1, K. and KIBAYASHI M., Principles and Current Status of Manual for Design and 

Construction of Passively-Controlled Buildings, 23-32 (in Japanese) 

[4] KIBAYASHI, M., KATO, S., KIKUCHI, M., KIMURA, Y, KOBAYASH1, T., and TSUI, Y, Summary 

of Guidelines for Passive Energy Dissipation Systems in USA, 33-44 (in Japanese) 

[5] TSUYUKI, Y, KAMEI, T., GOHUKU, Y, and IIYAMA, R, Performance and Qualir\ Control of Oil 

Damper, 45-58 (in Japanese) 

[6] KAWAGUCHI, S., SUKAGAWA, M., MASAKL N., SERA, S., iCATO, N.. VrASHIYAMA, \ , 

MITSUSAKA, Y., and FURUKAWA, Y, Performance and Quality Control of Viscous Dampers,

58 

59-74 (in Japanese) 

[7] ISHIKAWA, K., OKUMA, K., OKU, T., SONE, Y, NAKAMURA, K, and MASAKI, N., 

Performance and Quality Control of Viscoelastic Dampers, 75-86 (in Japanese) 

[8] NAKATA, Y., UEMURA, K., IIDA, T., and HIROTA, M., Performance and Quality Control of Steel 

Hysteretic Damper, 87-100 (in Japanese) 

[9] TAKAHASHI, O. and SEKiGUCHI, Y., Constitutive Rule of the Oil Damper with Maxwell Model 

and Source Code for the Analysis Program, 101-108 (in Japanese) 

[10] SEKIGUCHI, Y. and TAKAHASHI, O., Constitutive Rule of the Viscous Damping Wall and Source 

Code for the Analysis Program, 109-116 (in Japanese) 

[11] KASAI, K. and OOHARA, K., Algorithm and Computer Code to Simulate Response of Viscous 

Damper, 117-126 (in Japanese) 

[12] KASAI, K., OOKI, Y., and TOKORO, K., Algorithm and Computer Code to Simulate Response of 

Acrylic Viscoelastic Damper, 127-140 (in Japanese) 

[13] SODA, S. and KAKIMOTO, K., Mechanical Models and Algorithm for Viscoelastic Dampers, 

141-162 (in Japanese) 

[14] ONO, Y, and KANEKO H., Constitutive Rule of the Steel Damper and Source Code for the Analysis 

Programs, 163-170 (in Japanese) 

Companion Paper (Part 2): 

[15] KASAI, K. (2002). Development of JSSI Manual for Passive Control of Buildings (Part2: Device 

Standard and Implementations), Proceedings oflCANCEER 2002, Hong Kong, China, August 19-20, 

2002 

Related SEWC2002 (Structural Engineers World Congress 2002, Yokohama, Japan, October 

9-12, 2002) Papers by Response Control Committee: 

[16] TAKEUCHI, T., KASAI, K., OHARA, K., NAKAJIMA, H., and KIMURA, Y., Performance 

Evaluation and Design of Passively Controlled Buildings Using Equivalent Linearization 

[17] ICHIKAWA, Y., TAKEUCHI, T., MORIMOTO, S., and SUGIYAMA, M., Practical Design of 

High-Rise Structure Using Viscoelastic Dampers and Hysteretic Dampers 

[IS] 

KANADA, M., KASAI, K., and OKUMA, K., Innovative Passive Control Scheme: a Japanese 

12-Story Building with stepping Columns and Viscoelastic Dampers 

[ 19] ONO, Y., KANEKO, H., and KASAI, K., Time-History Analysis Models For Steel Dampers 

[20] OOKI, Y., KASAI, K., and TOKORO, K., Time-History Analysis Models For Linear And Nonlinear 

Viscoelastic Dampers 

[21] OOHARA, K., and KASAI, K., Time-History Analysis Models For Nonlinear Viscous Dampers 

[22] SEKIGUCHI, Y., and TAKAHASHI, 0., Time-History Analysis Models For Nonlinear Viscous 

Damping Wall 

[23] TAKAHASHI, 0. and SEKIGUCHI, Y., Time-History Analysis Models For Nonlinear Oil Dampers 

[24] TSUYUKI, Y., KAMEI, T., GOFUKU, Y., IIYAMA, F., and KOTAKE, Y., Performance And 

Quality Control Of Oil-Damper 

[25] FURUKAWA, Y., KAWAGUCHI, S., SUKAGAWA, M., MASAKI, N., SERA, S., KATO, N., 

WASHIYAMA, Y, and MITSUSAKA, Y., Performance and Quality Control of Viscous Dampers 

[26] OKUMA, K., 1SHIKAWA, K., OKU, T., SONE, Y., NAKAMURA, H., and MASAKI, N., 

Performance and Quality Control of Viscoelastic Dampers 

[27] NAKATA, Y., Performance And Quality Control Of Steel Hysteretic Damper

Proceedings of the International Conference on 59 



DEVELOPMENT OF JSSI MANUAL 

FOR PASSIVE CONTROL OF BUILDINGS 

(Part 2: 

Device Standard and Implementations) 

Kazuhiko KASAI 

Professor, Structural Engineering Research Center, 

Tokyo Institute of Technology, Yokohama, JAPAN, and 

Chairman, Response Control Committee, 

Japan Society of Seismic Isolation (JSSI) 

ABSTRACT 

Part 2 introduces the manual contents regarding analytical modeling, numerical algorithm, and 

example computer codes for load-deformation relationship of each device type. It also explains 

another part of the manual referring to the mechanical, chemical, and environmental characteristics 

as well as acceptable range and quality of each device type. Manual policies on declaration of 

device property, assurance of device quality, and maintenance for long-term use will be explained. 

INTRODUCTION 

The companion paper (Part 1, see Ref.l) explained background and scopes as well as design 

concept of the JSSI Manual for Design and Construction of Passively Controlled Buildings. It 

introduced the tasks of the subcommittees and the working groups, and briefly explained the 

contents of the manual regarding classifications of the seismic energy dissipation devices and 

framing types, as well as comprehensive methods to evaluate effectiveness of the devices and the 

framing schemes [1]. 

The present paper (Part 2) introduces the manual contents regarding analytical modeling, numerical 

algorithm, and example computer codes for load-deformation relationship of each device type, 

which aims at promoting reliable time-history analysis of the passive control system. The contents 

were developed by the researchers and the engineers of software, construction, and design 

companies. Part 2 also explains another portion of the manual referring to the mechanical, 

chemical, and environmental characteristics as well as acceptable range and quality of each device 

type. The contents were created through extensive collaborations among more than twenty device 

manufacturers and their close interactions with the structural engineers and the researchers. 

Finally, manual policies on declaration of device property, assurance of device quality, and 

maintenance for long-term use will be explained. 

ANALYSIS MODELS AND DISSEMINATION OF COMPUTER CODES 

In Japan, significant progress is being made in numerical modeling of the devices (Fig. 1) for 

time-history analysis of passively systems. However, the new models, in spite of enhanced

60 

accuracy and efficiency, have not necessanl> been implemented into the computer programs of 

software companies or construction companies. The manual is intended to accelerate 

implementations, by exhibiting model algorithms and computer codes, which should lead to more 

reliable and fair assessment "of the passive scheme, thereby promoting sound growth in the 

technology. The following briefly describes the models, and detailed information can be found 

from the ""papers by the Device Analysis Working Group (see Rets. 9 to 14, and 19 to 23) 

Oil damper produces the hysteresis loop of ellipse (Fig. 1). The material used therein is oil, and its 

orifice resistance against the flow produces the damper force [5,24]. The damper possesses a 

configuration of cylinder. It can be modeled by a linear dashpot against a small deformation rate. 

However, since the Japanese oil damper typically has the relief mechanism [5, 24], the viscous 

coefficient of the linear dashpot needs to be reset small when subjected to a large deformation rate 

[9,23]. 

Viscous damper produces the hysteresis loop of combined ellipse and rectangle (Fig. 1). The 

material used is typically silicon fluid, and its resistance against flow produces the damper force [6, 

25]. The damper possesses a configuration of vertical panel, box, or cylinder. Unlike the oil 

damper discussed above, its model uses a nonlinear dashpot whose force is a fractional power of 

deformation rate. For some types possessing elastic stiffness, the model considers an in-series 

combination of the spring and the nonlinear dashpot [11, 21]. The elastic stiffness may be a 

nonlinear function of the deformation [10, 22]. Sensitivity against temperature must be modeled 

for some types [10, 22]. 

Viscoelastic damper could be either linear type, softening type, and stiffening type. Hysteresis 

loops of the three types show commonly an inclined ellipse at relatively small deformation, but they 

differ considerably at larger deformation (Fig. 1). The material used is polymer composite of acryl, 

butadiene, silicon, or others, and resistance is produced from the molecular motion caused by 

loading [7, 26]. The damper has a configuration of vertical panel or tube, but it could be designed 

for many other configurations as well. It produces two forces, one proportional to deformation 

and another proportional to deformation rate, and mostly it is sensitive to frequency and 

temperature [7, 26]. In order to simulate these, some models consist of in-series as well as parallel 

combinations of dashpots and springs [13], and another model directly expresses the constitutive 

equation of the damper using fractional time-derivatives of the force and deformation [12, 20]. 

Steel damper produces hysteresis of approximately bi-linear characteristics (Fig, 1). It is a vertical 

panel utilizing shear yielding or a brace utilizing axial yielding of the steel, and can be designed for 

other configurations [8, 27], Analytical model can utilize the constitutive equations of steel 

material readily known from the past research, but the typical Japanese model assumes purely 

bi-linear behavior [14, 19]. The damper using lead or friction pad may be analytically treated in a 

similar manner. Note that the input parameters such as steel yield strength, ultimate strength, and 

strain-hardening modulus are the nominal values, not necessarily the actual ones. The analysis 

results must be cross-referenced to cumulative damage of the damper, since the damper is typically 

designed to yield under the small and frequent seismic loads. Special model is developed for some 

dampers designed to a post-buckled range. 

Fig. 2 shows existing combinations of the above-mentioned device types and framing types that are 

seen from current Japanese practice. The framing types are named as brace, wall, shear link, stud, 

bracket, connector, column, outrigger, and amplified types, respectively. More systems are 

expected to appear in the near future, as better control performance is sought. On the other hand, 

in the typical Japanese practice to-date, the system is converted commonly into an equivalent shear 

beam model for an efficient dynamic time-history analysis. Such modeling, however, has been 

found to cause significant error, and for some system a detailed frame model with 

member-to-member modeling may be inevitable. The manual will try to resolve this situation, by 

suggesting modeling techniques and their accuracy. Analysis methods for device and system must 

be carefully evaluated, especially because actual dampers are being produced precisely with a 

property tolerance of at most 20% relative to the target typically.

61 

Force is Proportional to Velocity 

Orifice Resistance of Oil 

Normally Accompanied by Relief Mechanism 

Force is Proportional to 

Fractional Power i>f Velocity 

ii Fluid i- km R 

Vlscoclastie Material Layers and Steel 

Shearinq Resistance, Various Shanes 

Axial Yklding 

Shear Yielding 

Rigid Plastic Behavior 

riction at Contact Siirta 

Fig. 1 Device Types of Considered in Manual 

(Friction Damper Considered for Analysis Only.)

62 

f 

•c 

J 

— 

c. 

p 

«3 

no 

Fig. 2 Existing Combinations of Various Device Types and Framing Types in Japan 

(a) Systems Using Brace Type, Wall Type, Shear Link Type, or Stud Type

63 

c 

U 

~c 

M Q 

c 

U 

>-„ 

H 

HO 

Fig. 2 Existing Combinations of Various Device Types and Framing Types in Japan 

(b) Systems Using Bracket Type, Connector Type, Column Type, or Other Types

64 

VARIOUS TESTS AND DISSEMINATION OF PROPERTY DATA 

Each of the above device types is designed and produced differently by the manufacturers in Japan. 

And, the Japanese structural engineers are currently making their own search and judgment when 

using the products, relying on the database from each manufacturer. The manual is intended to 

provide broad information for assisting such an effort, as well as a uniform basis for assessment of 

the various products in order to enable fair judgment and better quality control. In the manual the 

property of each damper is described for a common ranges of the loading and environmental 

conditions indicated in Table 1. For the data at the conditions not included in Table 1, the special 

performance check should be made 

Table 1 Design Parameter Ranges 

for Passive Control Devices 

Furthermore, the manual specifies the 

benchmark for the loading and environmental 

conditions. The benchmark conditions are: (1) 

vibration frequencies of 0.3 Hz and 1 Hz, 

typical values for a high-rise building and a 

medium-rise building, respectively; (2) 

temerature of 20°C, a typical value in a room 

where dampers are, and; (3) story drift angle 

of 0.03 rad., a traditionally used deformation 

limit against the so-called level-2 earthquake 

considered in Japan. The benchmark data 

will be also used as a comparative basis, to 

which variations of property and performance 

will be described for the ranges specified in 

Tablel. 

The damper test method for the range 

discussed above can vary according to the 

damper's dynamic characteristics and 

dependencies on loading and environment. 

The test items are listed in Table 2 for each 

Frequency 

Temperature 

Story Drift 

Angle 

Number of 

Cyclic 

Excursions 

Normal Range: 0.2— 3 Hz* 

Normal Range: 10-30t** 

Major Earthquake: 1/100 rad. 

Rare Wind Storm • 1 /200 rad. 

Frequent Wind : 1/1 0,000 rad. 

Major Earthquake: 10 cycles 

Rare Wind Storm : 1 ,000 cycles 

Frequent Wind: 1,000,000 cycles 

Special design consideration will be given for 

frequencies under 0.2 Hz, or over 3~~10 Hz. 

Special design consideration will be given for low 

temperature -IO~--Q 0 C, or high temperature 30~ 

40°C. 

device type. Especially, the loading test data considering the range in Table 1 and the items 

described in Table 2 should be sufficient for creating the proper analysis model of each damper. 

In principle, full-size dampers should be tested, which, however, may not be possible due to the 

limited experimental capabilities. Test methodologies for a reduced-scale damper will be carefully 

determined for each device type. Detailed information regarding the device properties and testing 

methods can be found from the papers (Refs. 5 to 8, 24 to 27) written by the Oil, Viscous, 

Viscoelastic, and Steel Device Working Groups that are defined by Ref. 1. 

POLICIES ON PROPERTY DECLARATION, QUALITY ASSURANCE, AND MAINTENANCE 

It is necessary to specify the performance demand for the damper as well as performance limit of 

the selected damper in a building plan document. The performance demand should reflect the 

items listed in Tables 1 and 2, and could include information such as expected maximum responses 

at the design load level. It is also desirable to indicate in the document and damper itself whether 

or not the damper is to be replaced after a major earthquake. When the damper is intended for a 

long-term use, careful evaluations must be made for the effects of a series of earthquakes that could 

be experienced by the damper. Especially when using a damper that yields and deforms 

permanently, expected consequence must be stated in the document and explained to the building 

owners. 

Post-earthquake investigations into a trace of proper functioning as well as possible damage of the 

damper must be performed as efficiently as possible, and it is desirable to provide the architectural 

detail that makes this task easy. However, in most cases the finish materials covering the damper

65 

need to be destroyed, and such possibility must be declared in the document. Furthermore, when 

two or more earthquakes are experienced, it becomes very important to establish a judgment basis 

for the investigation. The damage in the members transferring the damper force must be carefully 

evaluated, especially in a case of retrofitted building. 

The demands from the society to quality of building and its component has become more severe 

nowadays. As an example of the result, the Japanese law about promotion of quality of a 

residential building _was enacted in July 2000, and the ten-year warranty for the function of main 

structural member is required now. Considering this and the performance superior to general 

earthquake-resistant construction, the quality of the passively controlled building must be assured 

by taking all possible measures. Long-term warranty is highly desirable for the passive devices, 

but realizing it may be difficult due to the following: this new scheme has very little database for 

the actual performance, and the damage of the device may stem from inadequate structural design 

rather than defect of the device itself These obviously make it difficult to establish any warranty 

agreement among the device manufacturer, structural engineer, and building owner. 

Table 2 

Items Considered for Testing of Four Device Types. 

I Oil Damper Viscous Damper Vkcoelastic Damper! Steel Damper 

I; | (Stiffness} 

Elastic Stiffness 



[ BiLsiC 

Performance , 

Dumping Coefficient 

Fust-Relief Damping 

Coefficient 

Stroke 

Failure iMotie 

Dumping Coefficient 

Stiffness Change 

Stroke 

Failure Mode 

Damping Coefficient 

Softening and 

Stiffening 

(.Itimate Deformation 

Failure Mode 

Yield Strength 

Post- Yield Stiffness 

Ultimate Deformation 

Cumulative Plastic 

Deformation 

Failure Mode 

Impact Characteristics 




/ Various j 

1 

Dependencies 

1 Frequency 

Velocity 

Frequency 

Velocity and 

.Deformation 

Temperature 

Frequency 

Velocity and 

Deformation 

Temperature 

Deformation 

1 Low and j 

ji High Cycle | 

Number of Cycles 

Fatigue (Wind Loud) 


Fatigue (Wind Load) 


Fatigue 

(Earthquake Load) 

Fatigue (Wind Load) 


Fatigue 

{Earthquake Load) 

Fatigue {Wind Load) 

Durability 

j 

Aging 

Aging 

Water Resistance 

Aging 

Moisture Resistance 

Aging 

Rust Resistance 

1 Fire 1 

1 Resistance \- 

Firing Point 

Firing Behavior 

Gasification 

Firing Point 


Gasification 

Softening Point 


Gasification 

Softening Point 

Softening Beha\ ior 

j Others { 

Connection Slackness 

Connection Slackness 

Bond Strength 

Weld Quality 

Buckling Control 

Maintenance may be required for some passive devices that are nearly machine products, especially 

when the device warranty is sought. Note that the traditional building members were not 

subjected to maintenance, and the periodical check and repair may be difficult to require. 

However, considering the normal use period of 60 to 100 years for a building, it is strongly felt 

reasonable to enforce maintenance management of the devices that plays a key role in building

66 

response.. The post-earthquake investigation explained earlier could be also considered as a part 

of maintenance In a case of the base-isolated building, the maintenance of the isolators and other 

components, including the post-earthquake investigation, are required now And, the same 

consideration would be necessary for the passively controlled buildings. 

CONCLUSIONS 

Passive control scheme has established its status as a viable means to enhance seismic performance 

of buildings. For the sake of further growth in this technology, it is necessary to promote 

understanding of the passive control schemes, as well as to create a uniform basis for assessment of 

the various stages to be followed during the design and construction process. Pursuant to this, the 

JSSI Response Control Committee is currently formulating the Manual for Design and Construction 

of Passively-Controlled Buildings. The companion paper (Part 1) and this paper (Part 2) described 

the effort of the committee. 

In particular, Part 2 introduced the manual contents regarding analytical modeling, numerical 

algorithm, and example computer codes for load-deformation relationship of each device type. It 

also explained another part of the manual referring to the mechanical, chemical, and environmental 

characteristics as well as acceptable range and quality of each device type. Manual policies on 

declaration of the device property, assurance of the device quality, and maintenance for the 

long-term use were also explained. 

REFERENCES 

[1] KASAI, K (2002). Development of JSSI Manual for Passive Control of Buildings (Parti Background, 

Scope, and Design Concept), Proceedings oj ICANCEER 2002, Hong Kong, China, August 19-20, 

2002 

Note: Refs. [2] to [27] are listed in the above-mentioned companion paper, (i.e., Part 1 paper)




A FRAMEWORK FOR DEVELOPING PERFORMANCE- 

BASED EARTHQUAKE ENGINEERING 

Jack P. Moehle 

Pacific Earthquake Engineering Research Center 

University of California, Berkeley 

ABSTRACT 

Performance-Based Earthquake Engineering seeks to improve seismic risk decision-making 

through assessment and design methods that have a stronger scientific basis and that express 

options in terms that enable stakeholders to make informed decisions. A key feature is the 

definition of performance metrics that are relevant to decision making for seismic risk 

mitigation. In concept, these metrics consist of estimates of losses due to earthquakes, 

including direct losses (repair and restoration costs), loss in functionality (or downtime), and 

casualties. 

The first generation of performance-based earthquake engineering assessment procedures in 

the United States provides approximate relationships between structural response indices 

(interstory drifts, inelastic member deformations and member forces) and performanceoriented 

descriptions such as Immediate Occupancy, Life Safety and Collapse Prevention. 

However, the relationship between the structural indices and performance measures are very 

approximate, determined in part by calibration to expectations of performance intended by 

current building code provisions, and in part by engineering judgment of various experts. 

Moreover, specific expectations of the performance targets are not clearly defined. 

Efforts under way by researchers in Pacific Earthquake Engineering Research (PEER) center 

aim to develop a more robust methodology for performance-based earthquake engineering - 

one that breaks the process into logical elements that can be studied and resolved in a rigorous 

and consistent manner. The process begins with definition of a ground motion Intensity 

Measure, which defines in a probabilistic sense the salient features of the ground motion 

hazard that affect structural response. The next step is to determine Engineering Demand 

Parameters, which describe structural response in terms of deformations, accelerations, or 

other response quantities calculated by simulation of the building to the input ground motions. 

Engineering Demand Parameters are next related to Damage Measures, which describe the 

condition of the structure and its components. Finally, given a detailed probabilistic 

description of damage, the process culminates with calculations of Decision Variables, which 

translate the damage into quantities that enter into risk management decisions. Underlying 

the entire methodology is a consistent framework for representing the inherent uncertainties in 

earthquake performance assessment. 

While full realization of the methodology in professional practice is still years away, 

important advances are being made through research in the Pacific Earthquake Engineering 

Research Center. The discussion will describe the overall approach and some specific 

highlights.




SMART STRUCTURES TECHNOLOGY 

OPPORTUNITIES AND CHALLENGES 

B.F. Spencer, Jr. 

Department of Civil and Environmental Engineering, University of Illinois at Urbana-Champaign 

205 North Matthews Ave., Urbana, IL 61801, USA 

ABSTRACT 

The design, fabrication, and construction of smart structures is one of the ultimate challenges to engineering 

researchers today. In recent years, considerable attention has been paid to research and development 

of efficacious actuation devices for smart structures, with particular emphasis on alleviation of 

wind and seismic response of buildings and bridges. Because of their mechanical simplicity, low power 

requirements, and large force capacity, smart dampers provide one of the most promising new technologies 

for protection of civil infrastructure. Additionally, recent research efforts for advancing smart 

structures technology have centered around innovative sensors and sensor systems, as they form the 

essence of intelligence for a smart structure. In particular, smart sensoring systems offer a new and radical 

departure from the conventional approach to monitoring structures. This paper is divided into two 

parts. The first part focuses on the review two smart damping approaches that have been proposed and 

implemented in full-scale structures. The remainder of the paper then discusses recent advances in 

smart sensing technology, as well as the challenges and opportunities offered. 

INTRODUCTION 

For more than two decades, researchers have investigated the possibility of using smart structures technology 

to improve upon the performance of traditional approaches to reduce structural responses due to 

dynamic loads (Soong 1990; Spencer and Sain 1997; Housner et al 1997; Soong and Spencer 2002). 

The first full-scale application of smart structures technology to a building was accomplished by the 

Kajima Corporation in 1989 (Kobori 1994). The Kyobashi Seiwa building, shown in Fig. 1, is an 11- 

story (33.1 m) building with a total floor area of 423 m 2 . An active mass driver (AMD) system was 

installed, consisting of two AMDs — the primary AMD is used for transverse motion and has a mass of 

4 tons, while the secondary AMD has a mass of 1 ton and is employed to reduce torsional motion. The 

role of the active system is to reduce building vibration under strong winds and moderate earthquake 

excitations and consequently to increase the comfort of occupants of the building. Since that time, 

active/hybrid structural control has been successfully applied in approximately 40 commercial buildings 

and 15 bridges (during construction).

70 

Observation 

-" system 

Figure 1. Kyobashi Seiwa Building with AMD Installation. 

One of the most promising new methods for 

protecting civil infrastructure systems is 

found in smart damping (also known as 

semiactive control) systems. These systems 

offer the reliability of passive devices, yet 

maintain the versatility and adaptability of 

fully active systems without requiring the 

associated large power sources. In fact, 

many can operate on battery power, which is 

critical during seismic events when the main 

power source to the structure may fail. 

According to presently accepted definitions, 

a smart damping device is one which cannot 

inject mechanical energy into the controlled 

structural system (i.e., including the structure 

and the control device), but has properties 

that can be controlled to optimally reduce the responses of the system. Therefore, in contrast to 

active control devices, smart damping devices do not have the potential to destabilize (in the bounded 

input/bounded output sense) the structural system. Studies have shown that appropriately implemented 

smart damping systems perform significantly better than passive devices and have the potential to 

achieve, or even surpass, the performance of fully active systems, thus allowing for the possibility of 

effective response reduction during a wide array of dynamic loading conditions (Dyke et al. 1998; 

Spencer et al. 2000). Examples of such devices include variable-orifice fluid dampers, controllable friction 

devices, variable stiffness devices, adjustable tuned liquid dampers, and controllable fluid dampers 

(Spencer and Sain 1997). 

In addition to being adaptable, smart structures can have the possibility of using their sensors to become 

self-monitoring. The ability to continuously monitor the integrity of structures in real-time can provide 

for increased safety to the public, particularly for the aging structures in widespread use today. Detecting 

damage at an early stage can reduce the costs and down-time associated with repair of critical damage. 

Observing and/or predicting the onset of dangerous structural behavior, such as flutter in bridges, 

can allow for advance warning of such behavior and commencement of mitigating control or removal 

of the structure from service for the protection of human life. In addition to monitoring long-term degradation, 

assessment of structural integrity after catastrophic events, such as earthquakes, hurricanes, tornados, 

or fires, is vital. These assessments can be a significant expense (both in time and money), as 

was seen after the 1994 Northridge earthquake with the shear number of buildings that needed to have 

the moment-resisting connections inspected. Additionally, structures internally, but not obviously, damaged 

in an earthquake may be in great danger of collapse during aftershocks; structural integrity assessment 

can help to identify such structures to enable evacuation of building occupants and contents prior 

to aftershocks. Furthermore, after natural disasters, it is imperative that emergency facilities and evacuation 

routes, including bridges and highways, be assessed for safety. However, a significant number of 

sensors are required to efficaciously monitoring of civil infrastructure systems. 

Recent advances in sensors, wireless communication, MEMS, and information technologies have made the 

dream of inexpensive, powerful, ubiquitous sensing for structural health monitoring a readily achievable 

near-term goal. To assist in dealing with the large amount of data that is generated by such a dense 

array of sensors, on-board processing at the sensor level allows a portion of the computation to be done 

locally on the sensor's embedded microprocessor. Such an approach provides for an adaptable, smart sensor,

71 

with the potential for self-diagnosis and self-calibration capabilities, thus reducing that amount of information 

that needs to be transmitted over the network. Indeed, a National Research Council report recently 

noted that the use of networked systems of embedded computers and sensors throughout society could 

well dwarf all previous milestones in the information revolution. 

This paper is comprised of two parts. Part one focuses on two smart damping strategies that have 

already been implemented in full-scale civil engineering applications. The first, the variable orifice 

damper, has been or is being installed in 11 buildings in Japan. The second smart damping strategy discussed 

herein considers magnetorheological (MR) fluid dampers, one of the most promising of the 

smart devices. These two classes of smart dampers mesh well with application demands and constraints 

to offer an attractive means of protecting civil infrastructure systems against severe earthquake and 

wind loading. The second part of the paper is directed toward the future of smart sensors. A brief review 

of existing systems is provided, followed by an outline of some of the challenges and opportunities 

offered. 

VARIABLE-ORIFICE DAMPERS 

One means of achieving a smart damping device is to 

use a controllable, electromechanical, variable-orifice ^-- 

valve to alter the resistance to flow of a conventional ^» 

hydraulic fluid damper. Such a device, schematically L J 

shown in Fig. 2, typically operates on approximately 

50 watts of power. The concept of applying this type 

Flgure 2> Schematic Of variable-orifice damper 

of variable-damping device to control the motion of bridges experiencing seismic motion was first discussed 

by Feng and Shinozuka (1990), Kawashima and Unjoh (1993) and Kawashima et al. (1992). 

Subsequently, variable-orifice dampers have been studied by Symans et al. (1994) and Symans and 

Constantinou (1996) at the Multidisciplinary Center for Earthquake Engineering Research in Buffalo, 

New York. 

Sack and Patten (1993) conducted experiments in which a hydraulic actuator with a controllable orifice 

was implemented in a single-lane model bridge to dissipate the energy induced bv vehicle traffic. These 

studies were followed by a full-scale experiment conducted on a bridge on interstate highway 1-35 to 

demonstrate this technology (Patten 1998, Patten, et al. 1999; Kuehn et al. 1999) shown in Figs. 3-4. 

Figure 5 shows the effectiveness of the SAVA system. This experiment constitutes the only full-scale 

implementation of structural control in the USA. 

Conceived as a variable-stiffness device, Kobori et al (1993) and Kamagata and Kobori (1994) implemented 

a full-scale variable-orifice damper in a semiactive variable-stiffness system (SAVS) to investigate 

semiactive control at the Kajima Technical Research Institute. The overall system, shown 

schematically in Fig. 6, has SAVS devices installed on both sides of the structure in the longitudinal 

direction. The results of these analytical and experimental studies indicate that this device is effective in 

reducing structural responses. 

More recently, a smart damping system was installed in the Kajima Shizuoka Building in Shizuoka, 

Japan. As seen in Fig. 7, semiactive hydraulic dampers are installed inside the walls on both sides of the 

building to enable it to be used as a disaster relief base in post-earthquake situations (Kobori, 1998:

72 

Figure 3. First full-scale implementation of 

smart damping in the US. 

Figure 4. SAVA-II variable-orifice damper 

'y Truck /Lett Lane 

is Section, Heavy Truck /Right Lane 

r 1500 

1 -2000 

I 

Girder #. Eastern to Western 

Figure 5. Comparison of peak stresses for heavy trucks 

Kurata et al. 1999). Each damper contains a flow control 

valve, a check valve and an accumulator, and can 

develop a maximum damping force of 1000 kN. 

Figure 8 shows a sample of the response analysis 

results based on one of the selected control schemes 

and several earthquake input motions with the scaled 

maximum velocity of 50 cm/sec, together with a simulated 

Tokai wave. Both story shear forces and story 

drifts are seen to be greatly reduced with control activated. 

In the case of the shear forces, they are confined 

within their elastic-limit values (indicated by E-limit); 

without control, they would enter the plastic range. 

The use of the variable-orifice damper has blossomed 

in Japan. Figure 9 shows the construction site in the 

Siodome area in downtown Tokyo. There are 4 buildings 

currently under construction in this area that will 

employ switching semi-active hydraulic dampers for 

Figure 6. Kajima Technical Research Institute 

with the SAVS svstem

73 

Semi-active Hydraulic 

(SHD) 

Control computers and 

Uninterruptible power supply 

Figure 7. Kajima Shizuoka building configured with semiactive hydraulic dampers 

(Story) 

(Story) 

5 

—•— EL50 

--4-TF50 ! 

--D--HA50 I 

-o--Tokai 

v t-^ 1/200 J 

2000 4000 6000 8000 (kN) 

Maximum shear force 

(a) w/ control 

2 4 6 8 10 (cm) 

Maximum story drift 

(b) w/o control 

Figure 8. Maximum responses (El Centro, Taft and Hachinohe waves 

with 50 cm/sec, and assumed Tokai waves) 

structural protection. One of these structures, the Kajima K-Tower, is a 172 m tall, 38-story hotel and 

office complex installed with 88 semi-active dampers and 2 hybrid mass dampers (see Fig. 10). In the 

Roppongi area of Tokyo, the Kajima R-Building, a 54-story building with 356 variable-orifice dampers 

and 192 passive dampers distributed throughout, has recently been completed (see Fig. 11). Altogether, 

the Kajima Corporation is currently constructing or has recently finished 9 buildings in Japan that 

employ semi-active hydraulic dampers for structural protection. Table 1 provides a summary of these 9 

buildings (Kobori 2002). When these projects are completed, a total of more than 777 variable-orifice 

dampers will be installed in building structures in Japan.

74 

Figure 9. View of construction site in the Siodome area in downtown Tokyo 

• ...'.IttttUHMn 

^ 10. Siodome Kajima T( fower 

Figure 11. Roppongi Tower 

CONTROLLABLE FLUID DAMPERS 

Except for controllable fluid dampers, smart dampers employ some electrically controlled valves or 

mechanisms to achieve changes in device characteristics. Such mechanical components can be problematic 

in terms of reliability and maintenance. Another class of smart damping devices uses controllable 

fluids in a fixed-orifice damper. As shown schematically in Fig. 12, the advantage of these 

controllable fluid dampers is their mechanical simplicity; i.e., they contain no moving parts other than 

the damper's piston. 

Two fluids that are viable contenders for development of controllable dampers are: (i) electrorheological 

(ER) fluids and (ii) magnetorheological (MR) fluids. The essential characteristic of these fluids is 

their ability to reversibly change from a free-flowing, linear viscous fluid to a semi-solid with a control-

75 

Table 1 Buildings currently under construction employing semi-active hydraulic dampers 

Name 

Stones 

Height (m) 

# of Semi- active 

Dampers 

Completion Date 

Chuden Gifu Building 

Niigata B-prqject 

Siodome M-Building 

Siodome N-Building 

Siodome K-Tower 

Roppongi Tower 

Siodome T-Building 

S-Hotel 

H-Building 

11 

31 

25 

28 

38 

54 

19 

30 

23 

56 

140.5 

119.9 

136.6 

172 

241.4 

98.9 

104.9 

100.4 

42 

72 

38 

60 

88 

356 

27 

66 

28 

March 2001 

December 2002 

January 2003 

March 2003 

April 2003 

May 2003 

May 2003 

December 2004 

August 2004 

lable yield strength in milliseconds when exposed to an electric (for ER fluids) or magnetic (for MR fluids) 

field. In the absence of an applied field, these fluids flow freely and can be modelled as Newtonian. 

When the field is applied, a Bingham plastic model (Shames and Cozzarelli 1992) is often used to 

describe the fluid behavior. 

Although the discovery of both ER and MR fluids 

Electric or Magnetic Choke 

dates back to the late 1940's (Rabinow 1948; Winslow 

1947, 1949), for many years, research programs 

Controllable Fluid 

concentrated primarily on ER fluids. Nevertheless, X Jlpssi M 

some obstacles remain in the development of commercially 

feasible damping devices using ER fluids. 

\ ^ | 

;••»,» -w» ** * _1 

1 

For example, the best ER fluids currently available 

1 

have a yield stress of only 3.0 to 3.5 kPa and cannot Figure 12. Schematic of controllable fluid damper 

tolerate common impurities (e.g., water) that might 

be introduced during manufacturing or use. In addition, safety, availability and cost of the high voltage 

(e.g., -4000V) power supplies required to control the ER fluids need to be addressed. 

Recently developed MR fluids appear to be an attractive alternative to ER fluids for use in controllable 

fluid dampers (Carlson 1994; Carlson and Weiss 1994; Carlson et al. 1995a,b) (see also: http:// 

www.rheonetic.com/). MR fluids typically consist of micron-sized, magnetically polarizable particles 

dispersed in a carrier medium such as mineral or silicone oil. Carlson and Weiss (1994) indicate that the 

achievable yield stress of an MR fluid is an order of magnitude greater than its ER counterpart and that 

MR fluids can operate at temperatures from -40 to 150°C with only modest variations in the yield 

stress. Moreover, MR fluids are not sensitive to impurities such as are commonly encountered during 

manufacturing and usage, and little particle/carrier fluid separation takes place in MR fluids under common 

flow conditions. The size, shape and performance of a given device is determined by a combination 

of T> and Tîd) . The design equations for most controllable damper geometries indicate that 

minimizing the ratio V têid) is desirable. This ratio for MR fluids ( V 1^) = 5xicr n sec/Pa) is 

three orders of magnitude smaller than the corresponding ratio for today's best ER fluids. Thus, controllable 

devices using MR fluids have the potential of being much smaller than ER devices with similar 

capabilities. Further, the MR fluid can be readily controlled with a low power (e g. 9 less than 50 watts), 

low voltage (e.g., -12-24V), current-driven power supply ourpurting only -1-2 amps. Such power lev-

76 

els can be readily supplied by batteries. Table 2 provides a summary of the key physical characteristics 

of both MR and ER fluids. 

Spencer et al. (1997) and Dyke et al. (1998) (see also: http://www.nd.edu/-quake/) have conducted a 

number of pilot studies to assess the usefulness of MR dampers for seismic response reduction. Dyke et 

al. (1998) have shown through laboratory experiments that the MR damper, used in conjunction with 

recently proposed acceleration feedback control strategies, significantly outperforms comparable passive 

configurations of the damper for seismic response reduction. 

Table 2 Summary of the Properties of Today's MR and ER Fluids (Carlson 1994). 

Property 

Max. Yield Stress i y(fjeld) 

Operable Temp. Range 

Plastic Viscosity, r^ 

Stability 

V 

T y(field) 

Power Supply (typical) 

Response Time 

Particle Sedimentation 

Raw Materials 

MR Fluids 

50-100kPa 

-50 to 150°C 

0.2-1. OPa-s 

Not affected by most impurities 

10- 10 -10- u s/Pa 

2-25V, 1-2 A 

milliseconds 

Little 

nontoxic & environ, safe 

ER Fluids 

2-5 kPa 

+ 10to90°C 

0.2-1. OPa-s 

Cannot tolerate impurities 

l(T 7 -l(r 8 s/Pa 

2000-5000 V, 1-1 OmA 

milliseconds 

Little 

nontoxic & environ, safe 

DEVELOPMENT OF FULL-SCALE MAGNETORHEOLOGICAL DAMPERS 

To prove the scalability of MR fluid technology to smart damping devices of appropriate size for civil 

engineering applications, a 20-ton MR fluid damper has been designed and built (Carlson and Spencer 

1996; Spencer et al. 1997, 1999). Figure 13a shows the schematic of the MR damper tested herein. The 

damper uses a particularly simple geometry in which the outer cylindrical housing is part of the magnetic 

circuit. The effective fluid orifice is the entire annular space between the piston outside diameter 

Figure 13. (a) Schematic of 20-ton MR fluid damper; (b) Experimental setup

77 

and the inside of the damper cylinder housing. The damper has an inside diameter of 20.3 cm and a 

stroke of ±8 cm. The electromagnetic coils are wound in three sections on the piston, resulting in four 

effective valve regions as the fluid flows past the piston. The coils contain a total of about 1.5 km of 

wire. When wired in series, the total coil has an inductance L 0 = 6.6 H and a resistance R Q = 21.9 Q. 

The completed damper is approximately 1 m long and has a mass of 250 kg. The damper contains 

approximately 5 liters of MR fluid. The amount of fluid energized by the magnetic field at any given 

instant is approximately 90 cm 3 . 

Figure 13b shows the experimental setup at the University of Notre Dame for the large-scale 20-ton MR 

fluid damper. The damper was attached to a 7.5 cm thick plate that was grouted to a 2 m thick strong 

floor. The damper is driven by a 560 kN actuator configured with a 57 1pm servo-valve with a bandwidth 

of 30 Hz. A Schenck-Pegasus 5910 servo-hydraulic controller is employed in conjunction with a 

200 MPa, 340 1pm hydraulic pump. So that reliable tests of the dynamic performance could be obtained, 

particular care was taken to minimize compliance in the system. 

Analytical and experimental pseudo-static studies of the damper have been conducted at the Structural 

Dynamics and Control / Earthquake Engineering Laboratory at the University of Notre Dame. Figure 

14a shows the measured force-displacement loops under a commanded 5.4 cm/sec triangular displacement 

at the maximum magnetic field and no magnetic field (off state) respectively. At the maximum 

magnetic field, the output force of the damper is 201 kN, which is within 0.5% of the design specification 

of 200 kN. Moreover, the on/off range of the damper is well over the design specification of 10. 

Fig. 14b also shows the measured force-velocity relationship and its comparison with the results of the 

axisymmetric model and parallel plate model (Spencer et al 1999). Both analytical models compare 

well with the experimental results. The dynamic response of the damper to changes in the commanded 

force has also been shown to be excellent (Yang et al 2000; Yang 2001). 

Recently, Sunakoda et al (2000) have also presented encouraging results regarding design and construction 

of large scale MR dampers. More information regarding MR dampers and their application to 

civil engineering structures can be found at: hitpj/cee muc.edu/sstl/. 

Max Magnetic Reid (2A) 

No Magnetic Field iQA) 

Figure 14. (a) Measured force-displacement loops at 5.4 cm/sec; (b) Comparison of measured and 

predicted force-velocity behavior (Spencer et aL 1999)

78 

FULL-SCALE IMPLEMENTATION OF MR DAMPERS 

In 2001, the first full-scale implementation of MR dampers for civil engineering applications was 

achieved. The Nihon-Kagaku-Miraikan, the Tokyo National Museum of Emerging Science and Innovation, 

shown in Fig. 15, has two 30-ton, MR Fluid dampers installed between the 3rd and 5th floors. The 

dampers were built by Sanwa Tekki using MR fluid from the Lord Corporation. 

The Dongting Lake Bridge in Hunan, China constitutes the first full-scale implementation of MR dampers 

for bridge structures (see Fig. 16). Long steel cables, such as are used in cable-stayed bridges and 

other structures, are prone to vibration induced by the structure to which they are connected and by 

weather conditions, particularly wind combined with rain, that may cause cable galloping. The 

extremely low damping inherent in such cables, typically on the order of a fraction of a percent, is 

insufficient to eliminate this vibration, causing reduced cable and connection life due to fatigue and/or 

breakdown of corrosion protection. Two Lord SD-1005 (www.rheonetic.com) MR dampers have been 

installed on each cable to mitigate cable vibration. The technical support for this engineering project 

was provided through a joint venture between Central South University, The Hong Kong Polytechnic 

University, and the author. Recently, a feasibility study on the applicability of MR fluid dampers for 

cable vibration control of the Stonecutter Bridge, which when construction is complete will be the 

world's longest cable-stayed bridge with a main span of 1018, has been accomplished (Chen et al. 2002, 

Ying et al. 2002). 

While tremendous strides have been made, a number of aspects of the smart damping control problem 

merit additional attention. One particularly important area is system integration. Structural systems are 

complex combinations of individual structural components. Integration of smart damping control strategies 

directly into the basic design of these complex systems can offer the optimal combination of performance 

enhancement versus construction costs and long-term effects. Because of the intrinsically 

nonlinear nature of smart damping control devices, development of output feedback control strategies 

that are practically implementable and can fully utilize the capabilities of these unique devices is 

another important, yet challenging, task. Once the advantages of smart damping control systems are 

fully recognized, a primary task is the development of prototype design standards or specifications 

complementary to existing standards. 

Figure 15. Nihon-Kagaku-Miraikan, Tokyo National Museum of Emerging Science and Innovation

79 

Figure 16. MR damper installation on the Dongting Lake Bridge, Hunan, China 

SMART SENSORS 

Recent research efforts for advancing smart structures technology have centered around innovative sensors 

and sensor systems, as they form the essence of intelligence for a smart structure. In particular, 

smart sensing systems have received many researchers' attention. The essential difference between the 

standard sensor and "smart" sensor, lies in its intelligence capabilities due to the use of an on-board microprocessor. 

This microprocessor is intended for digital signal processing, self-diagnostics, self-identification, 

and self-adaptation (decision making) functions. To date, all smart sensors have also been wireless. 

Some of the first efforts in developing smart sensors for application to civil engineering structures were presented 

by Straser and Kiremidjian (1996, 1998), Straser, et al. (1998), and Kiremidjian et al (1997). This 

research sought to develop a near real-time damage diagnostic and structural health monitoring system. The 

hardware was designed to acquire and manage the data and the software to facilitate damage detection diagnosis. 

The sensor unit consists of a microprocessor, radio modem, data storage, and batteries. To save battery 

life, most of the time the sensor unit is in a sleep mode, periodically checking its hardware interrupts to 

determine if there are external events that require attention. Building on the work of Kiremidjian et al. 

(1997), Lynch et a\. (2001) demonstrated a proof-of-concept wireless sensor that uses standard integrated 

circuit components. This unit consists of an 8-bit Atmel microcontroller with a 4 MHz CPU that can accommodate 

a wide range of analog sensors. The communication between the sensors is done via a direct 

sequence spread spectrum radio. Some units used the ADXL210 accelerometer making use of the duty cycle 

modulator that provides a 14 bit output with an anti-aliased digital signal. In other units, a high performance 

planar accelerometer is used along with a 16 bit A/D converter. The whole system can be accommodated 

within a seal packing unit roughly 5" by 4" by 1" in dimension. The sensor unit was validated through various 

controlled experiments in the laboratory. It was pointed out that pushing data acquisition and computation 

forward is fundamental to the smart sensing and monitoring systems, but represents a radical departure 

from the conventional instrumentation design and computational strategies for monitoring civil structures. 

Maser et al. (1997) proposed the Wireless Global Bridge Evaluation and Monitoring System (WGBEMS) to 

remotely monitor the condition and performance of bridges. This system used small, self-contained, battery 

operated transducers, each containing a sensor, a small radio transponder, and a battery. The complete system 

consisted of a local controller placed off the bridge and several transducers distributed throughout the 

bridge. The data collection at the transducer involves signal conditioning, filtering, sampling, quantization, 

and digital signal processing. The radio link uses a wide band in the 902 to 928 MHz range.

80 

Brooks (1999) emphasized the necessity of migrating some of the computational processing to the sensor 

board, calling them fourth-generation sensors. He indicated that this generation of sensors will be characterized 

by a number of attributes: bi-directional command and data communication, all digital transmission, 

local digital processing, preprogrammed decision algorithms, user-defined algorithms, internal self-verification/diagnosis, 

compensation algorithms, on-board storage, and extensible sensor object models. 

Mitchell et al. (1999) presented a wireless data acquisition system for health monitoring of smart structures. 

They developed a micro sensor that uses an analog multiplexer to allow data from multiple sensors to be 

communicated over a single communication channel. The data is converted to a digital format before transmission 

using an 80C515CO microcontroller. A 900 MHz spread spectrum transceiver system, capable of 

transmitting serial data at the rate of SOKbps, is used to perform the wireless transmission. Mitchell et al. 

(2001) continued this work to extend the cellular communication between the central cluster and the web 

server, allowing web-control of the network. 

Liu et al. (2001) presented a wireless sensor system that includes 5 monitoring stations, each of them with a 

3-axis accelerometer (ADXL05). The stations use an 80C251 microprocessor with a 16 bit A/D converter. 

Because this network is sensing continuously, transmission of data to the base station could present collisions. 

To avoid this problem, a direct sequence spread spectrum radio with long pseudo noise code was used 

to distinguish each substation. Experimental verification was provided. 

While significant strides have been made toward the development of smart sensors, all of the above 

cited systems are proprietary in nature. Moreover, the types of measurements that can be readily made 

are not necessarily suitable for civil infrastructure applications. The Mote platform developed at the 

University of California at Berkeley (see Fig. 17) with funding from the Defense Advanced Research 

Projects Agency (DARPA) offers, for the first time, an open hardware/software environment for broad 

smart sensing research. In addition to the small physical size, low cost, modest power consumption, and 

diversity in design and usage, one of the main advantages of using the Mote is that employing the Berkeley-Mote 

(http://webs.cs.berkeley.edu/) platform allows leveraging of the substantial resources that 

have already been invested by DARPA. Smart sensors based on the Berkeley-Mote platform should 

provide the impetus for the development of the next generation of structural health monitoring and control 

systems. 

Although progress in smart 

sensing technology has made 

substantial progress in recent 

years, a number of impediment 

exist to the realization 

of the vision of massively 

distributed smart sensors for N 

structural health monitoring 

and control. One of the most 

prominent is the lack of a 

computational framework on 

vvhich to build new structural 

health monitoring strategies. 

Relatively complex algo- Figure 17. Berkeley-Mote (Mica) Processor Board. 

rithms for monitoring structures 

and detecting both the extent and location of damage have been developed and implemented in the 

laboratory (Doebling and Farrar 1999). However, these algorithms assume central processing of the 

data; they cannot be implemented readily in the distributed computing environment employed by smart

81 

sensors. New algorithms must be developed that can utilize the smart sensor's on-board processor to 

deal with the large amount of data that will be generated by a monitoring system employing a dense 

array of sensors. Additionally, eliminating redundant information from neighboring sensors may be 

necessary. Such an approach reduces that amount of information that needs to be transmitted over the 

network. To date an appropriate framework on which to build structural health monitoring and control 

strategies does not exist. Other critical issues include errors in synchronization of on-board clocks 

among sensors, delays in transmission due packet collisions (20-30% of the data packets can be lost), 

short transmission range, inability to simultaneously transmit and acquire data, and limited memory 

capabilities. 

CONCLUSIONS 

This paper provides an update of smart structures technology, including both smart dampers and smart 

sensors. Although in their infancy, control strategies based on smart damping devices appear to cornbine 

the best features of both passive and active control systems and to offer a viable means of protecting 

civil engineering structural systems against earthquake and wind loading. In particular, they provide 

the reliability and fail-safe character of passive devices, yet possess the adaptability of fully active 

devices. Full-scale implementations for two smart damping systems were presented. The variable orifice 

damper has been or is being installed in 11 buildings in Japan. Magnetorheological (MR) fluid 

dampers have been installed last year in a building in Japan and on a bridge in China. These two classes 

of smart dampers have been shown to mesh well with application demands and constraints to offer an 

attractive means of protecting civil infrastructure systems against severe earthquake and wind loading. 

To investigate both local and global damage criteria, a dense array of sensors is anticipated to be 

required for large civil engineering structures. Rapid advances in sensor, wireless communication, 

Micro Electro Mechanical Systems (MEMS), and information technologies have the potential to significantly 

impact structural health monitoring. The emerging technology of smart sensors, allows a portion 

of the computation to be done locally on the sensor's embedded microprocessor. This paper presented 

the basic ideas behind smart sensors, some of units developed to date, and indicated certain needs to 

takes advantages of smart sensors for structural health monitoring. The Berkeley-Mote smart sensors 

are an open hardware/software platform that will provide the impetus for the development of the next 

generation of structural health monitoring and control systems. 

ACKNOWLEDGMENT 

The author gratefully acknowledges the partial support of this research by the National Science Foundation 

under grant CMS 99-00234 (Dr. S.C. Liu, Program Director) and the Lord Corporation. The author 

would also like to thank Dr. Takuji Kobori for providing the information and photographs regarding the 

full-scale implementation of semi-active hydraulic dampers by the Kajima Corporation and thank Dr. 

Y.Q. Ni for providing the photographs regarding the full-scale implementation of MR dampers on the 

Dongting Lake Bridge.

82 

REFERENCES 

Brooks T. (1999). Using smart accelerometers and wireless interfaces for condition monitoring. Machine 

Plant & Systems Monitor. May/June issue. 

Carlson J.D. (1994). The promise of controllable fluids. Proc. of Actuator 94 (H. Borgmann and K. 

Lenz, Eds.), AXON Technologic Consult GmbH, 266-270. 

Carlson J.D. and Weiss K.D. (1994), A growing attraction to magnetic fluids. Machine Design, August, 

61-64. 

Carlson J.D., Catanzarite D.M., and St. Clair K.A. (1995). Commercial magneto-rheological fluid devices. 

Proc., 5th Int. Conf on ER Fluids, MR Fluids and Assoc. Tech., Sheffield, UK. 

Carlson J.D., Chrzan M.J., and James P.O. (1995). Magnetorheological fluid devices, U.S. Patent 

#5,398,917. 

Carlson J.D. and Spencer Jr. B.F. (1996). Magneto-rheological fluid dampers for semi-active seismic 

control. Proc, 3rd Int Conf on Motion and Vib. Control, III, 35-40. 

Chen, Y., Ni, Y.Q., and Ko, J.M. (2002). Applicability of passive and active devices for vibration control 

of long-span bridge cables. Research Report, Department of Civil and Structural Engineering, 

The Hong Kong Polytechnic University, Hong Kong. 

Doebling A.W. and Farrar C.R. (1999). The state of the art in structural identification of constructed facilities. 

Los Alamos National Laboratory. 

Dyke S.J., Spencer Jr. B.F., Sain M.K., and Carlson J.D. (1998). An experimental study of MR dampers 

for seismic protection. Smart Mat. and Struct. 7, 693—703 

Feng M.Q. and Shinozuka M. (1990). Use of a variable damper for hybrid control of bridge response 

under earthquake. Proc., U.S. National Workshop on Struct. Control Research. USC Publication no. 

CE-9013. 

Gavin H.P., Hanson R.D., and Filisko F.E. (1996). Electrorheological dampers, part 1: analysis and design. 

J. Applied Mechanics, 63:9, 669-675. 

Hill J. (2000) A software architecture supporting networked sensors. Research Project to obtain the degree 

of Master of Science. University of California at Berkeley. 

Hill J., Szewczyk R., Woo A., Hollar S., Culler D., and Pister, K. (2000). System architecture directions 

for networked sensors. International Conference on Architectural Support for Programming Languages 

and Operating Systems - ASPLOS, 9th International Conference Architectural Support for 

Programming Languages and Operating Systems (ASPLOS-LX), Cambridge, MA, pp. 93-104. 

Hollar S. (2000). COTS dust Research project to obtain the degree of Master of Science. University of 

California at Berkeley. 

Housner G.W., Bergman L.A., Caughey T.K., Chassiakos A.G, Claus R.O., Masri S.F., Skelton R.E., 

Soong T.T., Spencer B.F., and Yao J.T.P. (1997). Structural control: past, present and future," J. of 

Engrg Mech , ASCE, 123(9), 897-971. 

Kamagata S. and Kobori T. (1994). Autonomous adaptive control of active variable stiffness systems 

for seismic ground motion", Proc., 1 st World Conf. on Struct. Control, TA4:33-42, Pasadena, CA. 

Kawashima K. and Unjoh S. (1993). Variable dampers and variable stiffness for seismic control of 

bridges. Proc, Int. Workshop on Struct. Control, Honolulu, Hawaii, 283-297. 

Kawashima K., Unjoh S., and Shimizu K. (1992). Experiments on dynamics characteristics of variable 

damper. Proc., Japan Nat. Symp. on Struct. Resp. Control, Tokyo, Japan, 121. 

Kiremidjian A.S., Straser E.G, Meng T.H., Law K., and Soon, H. (1997). Structural damage monitoring 

for civil structures. International Workshop - Structural Health Monitoring, pp. 371-382. 

Kobori T. (1994). Future direction on research and development of seismic-response-controlled structure. 

Proc. 1st World Conf on Struct. Control, Panel: 19-31.

83 

Kobon T. (1998). Mission and perspective towards future structural control research. Proc, 2 nd World 

Conf on Struct Control, 1, 25-34, Kyoto, Japan. 

Kobon T. (2002). Past, present and future in seismic response control in civil engineering structures. 

Proc , 3rd World Conf on Struct Control, Como, Italy, 

Kobon T., Takahashi M, Nasu T., Niwa N., and Ogasawara, K. (1993). Seismic response controlled 

structure with active variable stiffness system. Earthquake Engrg. and Struct. Dyn., 22, 925-941. 

Kuehn J., Song G, and Sun J. (1999). Experimental verification of a non-protruding intelligent stiffener 

for bridges (ISB). Proc , Int. Post-SMiRT Conf Seminar on Seismic Isolation, Passive Energy, Dissipation 

and Active Control of Vibrations of Structures, Cheju, Korea, August 23-25, 1999. 

Kurata N., Kobori T., Takahashi M., NiwaN., and Midorikawa H. (1999). Actual seismic response controlled 

building with semi-active damper system. Earthquake Engrg and Struct. Dyn., 28, 1427- 

1447. 

Liu R.C., Zhou L, Chen X., and Mau S.T. (2001). Wireless sensors for structural monitoring. Strong 

Motion Instrumentation for Civil Engineering Structures, pp. 253-266. 

Lynch J.P., Law K.H., Kiremidjian A.S., Kenny T.W., Carryer E., and Partridge A. (2001). The design 

of a wireless sensing unit for structural health monitoring." 3rd International Workshop on Structural 

Health Monitoring, Stanford, CA, pp. 1041-1050. 

Maser K., Egri, R., Lichtenstein A., and Chase S. (1997). Development of a wireless global bridge evaluation 

and monitoring system (WGBEMS). Proceedings of the Specialty Conference on Infrastructure 

Condition Assessment- Art, Science, Practice, pp. 91-100. 

Mitchell K., Sana S., Balakrishnan VS., Rao V., and Pottinger, H.J. (1999). Micro sensors for health 

monitoring of smart structures. SPIE Conference on Smart Electronics and MEMS, 3673, pp. 351- 

358. 

Mitchell K., Dang N., Liu P., Rao V., and Pottinger H.J. (2001). Web-controlled wireless network sensors 

for structural health monitoring. Proceedings - SPIE the International Society for Optical Engineering, 

4334, pp. 234-243. 

Patten W.N. (1998). The 1-35 Walnut Creek bridge: an intelligent highway bridge via semi-active structural 

control. Proc, 2 nd World Conf on Struct. Control, 1,427^36, Kyoto, Japan. 

Patten W., Sun J., Li G, Kuehn J., and Song G. (1999). Field test of an intelligent stiffener for bridges at 

the 1-35 Walnut Creek bridge. Earthquake Engrg. and Struct. Dyn., 28(2), 109-126. 

Rabinow J. (1948). The magnetic fluid clutch. A1EE Transactions, 67, 1308-1315. 

Sack R.L. and Patten W. (1993). Semiactive hydraulic structural control, Proc., Int, Workshop on Struct. 

Control, USC Publication No. CE-9311, 417-431, Honolulu, Hawaii, 1993. 

Shames I.H., and Cozzarelli F.A. (1992). Elastic and inelastic stress analysis, Prentice Hall, Englewood 

Cliffs, New Jersey. 

Soong T.T. (1990). Active structural control: theory and practice, Longman Scientific and Technical, 

Essex, England, 

Soong T.T. and Spencer Jr. B.F. (2002). Supplementary energy dissipation: state-of-the-art and state-ofthe-practice. 

Engineering Structures, 24, 243-259. 

Spencer Jr. B.F., Dyke S.J., Sain M.K., and Carlson J.D. (1997). Phenomenological model ofamagnetorheological 

damper. J Engrg Mech., ASCE, 123:3,230-238. 

Spencer Jr. B.F. and Sain M.K. (1997). Controlling buildings: a new frontier in feedback. IEEE Control 

Sys. Mag 17:6, 19-35. 

Spencer Jr. B.F., Yang G, Carlson J.D., and Sain M.K. (1999). 'Smart' dampers for seismic protection 

of structures: a full-scale study. Proc., 2nd World Conf. Struct. Control, 1, 417-426. 

Spencer Jr. B.F., Johnson E.A., and Ramallo J. C. (2000). 'Smart 1 isolation for seismic control. JSME 

Int J. Special Issue on Frontiers of Motion and Vibration Control, C, 43(3), 704-711.

Straser E.G. and Kiremidjian A.S. (1996). A modular visual approach to damage monitoring for civil 

structures. Proceedings ofSPIE, Smart Structures and Materials, 2719, pp. 112-122. 

Straser E.G. and Kiremidjian A.S. (1998). A modular, wireless damage monitoring system for structures. 

The John A. Blume Earthquake Engineering Center, Report No. 128. 

Straser E.G., Kiremidjian A.S., Meng T.H., and Redlefsen, L. (1998). A modular, wireless network platform 

for monitoring structures. Proceedings - SPIE The International Society for Optical Engineering, 

issue 3243, 1, pp. 450-456. 

Sunakoda K., Sodeyama H., Iwata N., Fujitani H., and Soda S. (2000). Dynamic characteristics of magneto-rheological 

fluid damper Proc, SPIE Smart Struct and Mat Conf, 194-203. 

Symans M.D., Constantmou M.C., Taylor D.P., and Garjost K.D. (1994). Semi-active fluid viscous 

dampers for seismic response control. Proc , 1st World Conf on Struct Control, Los Angeles, California, 

FA4:3-12, August, 1994. 

Symans M D and Constantmou M.C. (1996). Experimental study of seismic response of structures with 

semi-active damping control systems. Proc, Structures Congress XIV, Chicago, IL, April 15-18, 

1996 

Winslow W.M. (1947). Method and means for translating electrical impulses into mechanical force, US 

Patent No. 2,417,850. 

Winslow W.M. (1949). Induced fibration of suspensions. J ofAppL Phys , 20, 1137-1140. 

Yang G, Ramallo J.C., Spencer Jr. B.F., Carlson J.D., and Sain M.K. (2000). Dynamic performance of 

large-scale MR fluid dampers. Proc, 14th Engrg. Mech. Spec, Conf, 6 pages. 

Yang G (2001). Large-scale magnetorheological fluid damper for vibration mitigation, modeling, testing 

and control. Ph.D dissertation, University of Notre Dame, IN. 

Ying, Z.G, Ni, Y.Q., and Ko, J.M. (2002). Analysis and control of dynamic stability of bridge stay cables 

induced by parametrically excited oscillation. Research Report, Department of Civil and Structural 

Engineering, The Hong Kong Polytechnic University, Hong Kong. 

84




EARTHQUAKE ACTION PROVISION IN GB50011-2001 

AND THREE IMPROVEMENTS 

Tao Xiaxin 1 and Geng Shuwei 3 ' 2 

'institute of Engineering Mechanics, Harbin 150080, China 

2 Harbin Institute of Technology, Harbin 150001, China 

ABSTRACT 

In the code for seismic design of buildings GB 50011-2001 of China, many new methods and 

techniques have been adopted, and quite large improvements achieved. For earthquake action 

provision, the two main improvements are summarized in this paper. (1) The long period band of the 

design spectrum is extended from 3 seconds to 6 seconds, while the low limit of the long period design 

spectrum is removed, and a slow line of amplitude decline is adopted instead; (2) Two additional 

intervals of a max, the maximum value of the design response spectrum, are added one for each 

Intensity VII and VI respectively, and T & the characteristic period, is determined based on both the 

seismic zoning map and the site condition, and thus, it is not necessary to take into account the 

"far-source" effect. 

Three further improvements are also suggested in this paper, as follows: a max to be adjusted 

according to site condition, the long period band of the design spectrum to be extended to 10 second, 

the declining rate of the design spectrum at long period range to be -1.0, and the ratio of design spectra 

between vertical and horizontal components to be taken as a function of the period, instead of a 

constant value of 0.65 for soil site. 

INTRODUCTION 

The new code for seismic design of buildings GB 50011-2001 of China, was issued by the Ministry of 

Construction and the State General Administration for Quality Supervision and Inspection and 

Quarantine last year, and was put into practice at the beginning of this year. Its earthquake action

86 

provisions have been improved, and are summarized and discussed in detail in this paper. There are 

still some new methods and suggestions that have not been included in the revision, since they had not 

been validated adequately at that time. As the end of a revision could be considered as also the 

beginning of the next, those methods and suggestions need to be considered further. Some other 

seismic design codes, for example, the Code for seismic design of highway engineering (JTJ004-89) 

and the Design code for anti seismic special structures (GB 50191-93), are ongoing or going to be 

revised in China. Therefore, it is necessary to probe into new improvement(s) as past of an overview of 

the state of the art and future trends in seismic design codes. 

THE IMPROVEMENTS IN EARTHQUAKE ACTION PROVISION IN GB 50011-2001 

In the revision of the code for seismic design of buildings of China, many new methods and techniques 

have been adopted, and quite large improvements achieved. For earthquake action provision, there are 

two main improvements. 

Period band of design spectrum extended 

With the development of construction and urbanization, buildings are getting larger and higher, and the 

fundamental natural periods of buildings are getting much longer than the 3.0 second that is the longest 

period of the design response spectrum (so called seismic effect coefficient) in the former edition of 

code for seismic design of buildings GBJ11-89. The long period of the design response spectrum band 

must be extended to meet the demand of construction. Of course, this broadening should be based on 

the reliability of the data at long period range. Since many strong ground motion records have been 

observed by analog accelerographs, the data reliability firstly depends on the period band of the 

instruments. Some correction techniques have been adopted to broaden the instrument 

amplitude-frequency feature, originally from O.SHz to 25Hz. The Ormsby digital filter with low cut-off 

frequency 0.15Hz could broaden the frequency band to 0.15Hz, i.e. the longest period is 6.6 second. 

Furthermore, another correction has been made to remove the noise in the digitizing process, which 

consists mainly of long period components. The long cut-off period of the filter was taken as 26cm/v, 

where v is the running speed of the record paper in cm/'s. Since the paper speed was generally 4.8 cm/s, 

the longest period must be 

5.4 second. Therefore, the 

long period design spectrum 

band in the new edition of the 

code is extended from 3.0 

second to 6.0 second. 

o. 

The low limit of the long 

period design spectrum in 

GBJ11-89, was acceptable 

Figure l ê design spectrum in code GB50011-2001

87 

for a period range less than 3.0 second, but it must be deleted now, because of the band broadening 

mentioned above. A slow declining line has been adopted instead in the new code, as shown in figure 1, 

and the amplitude of the design response spectrum a is described as a function of period T as 

follows 

5T 

s n 

CD 

where a max is the maximum value of the design response spectrum, T g is the characteristic period, 

7 is the descending rate of the design spectrum in period range from T g to 5 T g , and depends on the 

damping ratio as 

0.5 + 5C 

and 

77, is the descending rate of the design spectrum in period range from 5T g to 6.0 second, given 

as 

ty =0.02 + (0.05-£)/8 (3) 

and taking a value of 0.0, if the value is negative from the above equation. 

77., is the damping 

adjustment factor, given as 

(4) 

taking a value of 0.55, if the value calculated from the above equation is less than 0.55. 

Provision for 

a max and T g 

The improvement in design spectrum provision arises mainly from the revision of the China Seismic 

Zoning Map. A new national seismic zoning map of China, also the state standard GB 18306-2001, 

was issued by the State General Administration for Quality Supervision and Inspection and Quarantine 

at the beginning of 2001, and was put into practice in August, 2001. The zoning map consists of two 

sheets, one for ground acceleration, the other for the characteristic period T g . On the acceleration map,

the zones, divided by contours are in 7 intervals,

89 

tens or more than a hundred accelerograms recorded at each event, and mostly recorded by digital 

accelerographs. These instruments possess a wide dynamic range, broad frequency band, high 

sampling rate and pre-event memory. The low frequency errors in the records of these instruments are 

mainly due to instrument noise and site background noise. The total noise level could be estimated by 

comparing the records with the pre-event recordings. A result based on limited data showed that the 

longest period of the effective frequency band could reach a dozen seconds (Xie Lili etal., 2000). By 

checking if the long period drift could be removed completely in the displacement time history by a 

double integration of the acceleration record, a high pass cut-off frequency was selected as 

0.067Hz~0.083Hz. 

In this paper, 132 free field records from the 1999 Chi-Chi earthquake are used for comparing 

systematically the Fourier spectra of the pre-event recordings (i.e. noise) with those of the strong 

ground motion records. In general, the Fourier spectrum amplitudes of noise show a smooth 

distribution hi the band of most engineering interest, while they are getting larger at periods of about 

10 seconds. The spectrum amplitudes of strong ground motion are generally much higher than those of 

the noise, but descend gradually at very short and long periods, becoming similar to those of the noise. 

All results from the 132 records show that the signal to noise spectral ratio is greater than 90 at period 

of 10.24 second. It is higher than the 

ratio at high frequency range of more 

than 30 Hz. The facts show that the 

digital records of strong ground motion 

are reliable at least up to 10 second. The 

seismic effect coefficient at periods in Mtf 

the 6.0-10.0 second range could be 

stipulated in the seismic design code 1 icf 

from these records. An example of a 

Fourier spectra comparison of noise and 

strong ground motion recorded by the 

same accelerograph at the same position in 

two time periods with just tens seconds in 

between, is shown in figure 2. 

; ia 2 ia 3 ia 2 ia 1 icP io 1 10 2 

Lgf(mHz) 

Fig.2 Comparison of Fourier spectra of noise and 

ground motion 

As mentioned above, the amplitude of the design spectrum is depressed obviously by the improvement 

in the new code. However, it is still conservative if the band of the spectrum extends up to 10 seconds. 

In the seismic design codes of many earthquake countries (Euro code 8, 1993), the descending rates 

are given as T" 1 and T" 2 in moderate and long period ranges respectively. The reliability and safety 

redundancy of a design spectrum descending with T 1 at long period range are validated on the basis of 

760 strong ground motion records. The numbers of records in groups of magnitudes (M) and distances 

(R) are listed in the following table. 

Table 2 

Strong ground motion data in groups 

Rock site 

^^^--JQistance 

MaojiituSs^v^ R

90 

4

91 

In order to validate the conclusion above, 

the ground motion recorded at soft soil site 

in Chi-Chi earthquake are processed in the 

same way. The result is shown in figure 4 

for the whole penod band, where the black 

curve is the average spectrum with 

descending rate of T 1 , the grey ones are the 

spectra of the original records. It is also 

clear that the reliability of the average 

spectrum at period range longer than 6 

seconds is higher than that at short period 

range. 

0 5 10 15 

T(sec.) 

Fig 4 Comparison of the average spectrum with those of the 

original records on soft soil sites from Chi-Chi earthquake 

The maximum value of design spectrum adjusted from the site condition 

Before and during the revision of the code for the seismic design of buildings in China, an 

improvement had been adopted in some seismic design codes of other countries in that the maximum 

values of design spectra determined from the zoning maps were further adjusted from site condition. 

For example, in the NEHRP recommended provisions, the maximum value of the design spectrum 

equals the value from the zoning map for site category B (soft rock site) multiplied by a site factor. 

This improvement is consistent with the experience that earthquake damage on soil site is generally 

more serious than on rock site, and is supported by some observed data (TLDobry et al., 2000). The 

values of the adjusting coefficients mainly depend on statistical data for acceleration 0.1 g and less on 

site in category B, and on a lot of numerical results from site response analyses for stronger 

accelerations, as shown in table 3. 

Site 

categories 

A 

B 

C 

D 

E 

Table 3 The adjusting coefficients of various sites 

Acceleration from the zoning map for site category B 

92 

Site 

categories 

A 

B 

C 

D 

Table 4 The adjusting coefficients suggested by Li Xiaojun (2001) 

Acceleration from the zoning map for site category A 

0.05g O.lg 0.15g 0.20g 0.30g 0.40g 

1.0 1.0 1.0 1.0 1.0 1.0 

1.5 1.45 1.40 1.33 1.25 1.18 

1.10 LOO 0.9 0.8 0.7 0.6 

0.8 0.7 0.6 0.55 0.50 0.45 

The values in the above two tables commonly describe the amplifying effect of ground soil layers, and 

the reduction effect of soil nonlinearity when acceleration is strong enough. Meanwhile one can find 

there is quite a large difference between the values in the two tables, especially the trend of varying 

with soil from stiff to soft and from thin to thick. Should it increase or decrease The most convincing 

answer must come from the observed data. The maximum values of normalized response spectra of 

760 free field ground motion records, and the corresponding magnitudes, distances and site conditions, 

are summarized to work out the coefficient values in this paper. 

To deal with the relation between the maximum values of normalized spectra and site condition, the 

key point is to compare them at the same magnitude and distance. The biggest difficulty is that the data 

sample is not large enough, and the data distribution in magnitude and distance is not homogeneous. 

One can rarely find sets of data of the same magnitude (ideally the same earthquake) and exact same 

distance for various site conditions. The qualification has to be relaxed to those of similar magnitude 

and distance. The authors compare the mean maximum values in each magnitude and distance group 

directly rather than compare the attenuation relationships for various site categories. The ratios of 

mean values of maximum spectrum amplitudes for various sites to the value for rock site in every 

group are shown hi table 5, where the number in bracket after the ratio is the sample size (i.e. number 

of records. The records just on the boundaries between two neighboring groups are applied twice). The 

term Ace. is for acceleration of maximum spectral amplitude on rock site divided by 2.5 

Table 5 The ratios of mean values of maximum spectral amplitudes of various sites to one of rock site 

Distance ikm) 

' -— A CC - (§) 

Site Categories----^^_ 

1 

2 

3 

4 

Distance (km) 

' " — "•—--— .Âcc. (3) 

Site Categories'——-^ 

1 

2 

3 

4 

Magnitude ^5.75 

0-10 10-50 

0.2543 

1.00 (9) 

0.1365 

LOO (10) 

139 (6) 

0.79 (84) 0.79 (81) 

Magnitude 5.75 - 6.5 

0-30 30-50 

0.199 

1.00 (18) 

1.47 (8) 

1.96 (8) 

1.35 (91) 

0.1271 

1.00 (6) 

1.38 (6) 

1.10 (4) 

1.25 (112) 

50-100 

0.028 

1.00 (4) 

1.23 (22) 

50-100 

0.0291 

1.00 (6) 

1.53 (4) 

2.21 (38) 

100-200 

0.0303 

1.00 (2) 

100-200 

0.0357 

1.00 (2) 

0.68 (20)

93 

Distance (km) 

""""--Âcc. (g) 

Magnitude "*""~-~--—_ 

1 

2 

3 

4 

Magnitude 6.5 - 7.5 |g 

10-30 30-50 

0.2813 

1.00 (10) 

0.74 (2) 

1.16 (64) 

0.1355 

1.00 (6) 

1.51 (10) 

0.78 (2) 

1.18 (116) 

50-100 

0.1147 

1.00 (6) 

1.11 (12) 

1.78 (16) 

0.98 (72) 

100-200 

0.036 

1.00 (2) 

1.79 (2) 

1.81 (14) 

One can see from the table that the ratios vary with site condition. The ratio values are further 

summarized into four groups according to the Ace. values Ô.OSg, 0.08~0.15g, 0.15~0.25g, or^ 

0.25g, the mean values of these four groups are listed in table 6 with the representative values O.OSg, 

O.lg, 0.2g and 0.3g. From that table, one can find that, in general, soil site will amplify the ratio (the 

maximum to 1.96 times), the rate of amplification may decreases smoothly as the soil gets softer and 

the thicker the soil layers. The rate of amplification decreases obviously until less than 1.0 when the 

Ace. is greater than 0.3g from the nonlinearity effect of soil. It is true that some values in the table vary 

in unstable fashion, even jumpily, because of the insufficient and inhomogeneous data sample. Putting 

the ratios for the soil sites together, a general trend of the ratio decreasing smoothly with the Ace. can 

be recognized. From the above table and some other references, the suggested values of adjustment 

coefficient are listed in table 7, where the values for Ace. 0.15g and 0.3g are estimated by 

interpolation. 

Table 6 The mean ratios of four Ace, groups 

^^^"^•Â-cc (g) 

Site Categories"--^^^ 

1 

2 

3 

4 

0.05 

LOO 

1.66 

1.59 

0.1 

1.00 

1.30 

1.41 ^ 

1.06 

0.2 

1.00 

1.47 

1.96 

1.35 

0.3 

LOO 

0.77 

0.95 

""^^-^Âcc.Cg) 

Site Categon&s--^^^^ 

1 

2 

3 

4 

Table 7 The suggested values of site coefficient 

0.05 

1.0 

1.65 

1.65 

1.60 

0.1 

1.0 

1.5 

1.5 

1.4 

0.15 

1.0 

1.5 

1.5 

1.4 

0.2 

1.0 

1.5 

1.5 

1.3 

0,25 

1.0 

1.3 

1.2 

LI 

0.3 

1.0 

1.2 

LI 

1.0 

0.4 

1.0 

1.0 

1.0 

0.9 

Ratio of vertical to horizontal component varying with period 

As the accumulation of near field records has increased, some quite large vertical components of 

ground motions have been reported, which are even larger than the horizontal components. To examine 

if it is necessary to stipulate the ratio of vertical maximum amplitude to the horizontal one in the 

design spectra for different distances, 1449 ratios of vertical to horizontal parrs of records were

94 

calculated at 91 periods from 0.04 second to 15.0 seconds. The mean value of the ratios obtained over 

all period ranges is 0.65, which indicates the provision of the ratio in GB 50011-2001 is relevant in 

general. The results from a further group rag into nine are listed in table 8. The ratio varies smoothly, 

and reaches a value of 0.5 in far field from a large earthquake. 

Table 8 The V/H spectral ratio in groups 

^^^î^^de 

Distance^--^^ 80 

0.64 (429) 

0.66 (64) 

0.68 (12) 

0.65 (291) 

0.63 (402) 

0.63 (60) 

>6.75 

0.71 (33) 

0.63 (104) 

0.50 (52) 

Examining the variability of the ratio over the whole period range, one can find quite a large difference 

between the ratio values for different period ranges. A gross mean is shown in figure 5, and the general 

variation trend can be recognized as providing a peak with a value up to 1.0 over short period, which 

then descends rapidly to a trough with a value less than 0.5 as the period increases, and gradually 

increases to a value of 0.6 at a period 2.5 second and keeps that value at longer period. From a detailed 

grouping analysis for various magnitudes, distances and sites, the following can be concluded: the ratio 

on rock site varies smoothly, the peak in the short period range is obvious and may be greater than 1.0 

only at the near field from large earthquake; the peak ratio reaches 1.0 in short period range on soil site, 

and gets higher at the near field from a large earthquake. 

Since the peak ground acceleration relates only to the very short period component of ground motion, 

the conclusion above is consistent with some reports. The gross mean value of the ratio, however 

relates mainly to the stable value of 0.6 at long periods as well as the higher contribution over a very 

short period. In a seismic design code, in order to take into account the fact that the ratio of vertical to 

horizontal component varies with period, the authors suggest the ratio R should be expressed as a 

function of period Tfor soil site. A preliminary formula is as follows 

0.9 

1.0-7* 

0.475 + 0.057* 

0.6 

o

95 

CONCLUSION 

In this paper the improvements in earthquake action provision in the China code for seismic design of 

buildings GB 50011-2001 are summarized as (1) The long period band of design spectrum has been 

extended from 3 second to 6 second, while the low limit of the design spectrum at long period has been 

removed, and a slow decline line is adopted instead; (2) Two additional intervals of a ^^ are added , 

one each for Intensity W and VI respectively, and T g9 the characteristic period, is determined based 

on both the seismic zoning map and the site condition. Thus, the "far-source" effect does not need to 

be taken into account once again. 

Three further improvements are also suggested for future consideration for inclusion into the code, as 

follows: Q max to be adjusted according to site condition, the long period band of the design spectrum 

to be extended to 10 second, and the declining rate of the design spectrum at long period range to be 

-1.0, the design spectra ratio of vertical and horizontal components to be taken as a function of period 

for soil sites, instead of a constant value of 0.65. 

REFERENCES 

National Standard of People's Republic of China (2001). Code for seismic design of buildings 

GB50011-2001. 

CALTRAKS (1999). Seismic design criteria, Version 1.1. 

Eurocode 8 (1993). Design provision for earthquake resistance of structures. 

FEMA(1997). NEHRP recommended provisions for seismic regulations for new buildings. 

Xie Lili, Li Shabai and Qian Qukang (1981). Study on the instrument correction of accelerograms 

recorded by accelerograph coupled with galvanometers. Earthquake Engineering and Engineering 

Vibration 1:1, 106-116. 

Xie Lili, Qian Qukang and Li Shabai (1982). The effect of digitization errors on analysis of strong 

ground motion records and their elimination. ACTA Seismologica Sinca 4:4,412-421. 

Xie Lili, Zhou Yongnian, Hu Chengxiang, Yu Haiying(1990). Characteristics of response spectra of 

long period earthquake ground Motion. Earthquake Engineering and Engineering Vibration 10:1, 

1-20. 

W. H. K. Lee, T. C. Shin, K. W. Kuo, K. C. Chen, and C. F. Wu(2001). CWS free-field strong motion 

data from the 21 September Chi-Chi, Taiwan, earthquake. BSSA 91:5, 1370-1376. 

Xie Lili, Zhou Yongnian et.al. (2000). Near field strong ground motion and long period component 

Research Report, Institute of Engineering Mechinics, 1-5. 

Wang Guoxin and Tao Xiaxin (2001). Extraction and study of response spectrum parameters. World 

Information on Earthquake Engineering 17:2,73-78. 

Dobry, R., R. D. Borcherdt, C. B. Grouse, I. M. Idriss, W. B. Joyner, G R. Martin, M. S. Power, E. E.

Rinne, and R. B. Seed (2000). New site coefficients and site classification system used in recent 

building seismic code provisions. Earthquake Spectra 16:1, 41-67. 

C. B. Grouse and J.W. McGuire (1996). Site response studies for purpose of revising NEHRP seismic 

provisions. Earthquake Spectra 12:3,407-438. 

Bo Jingshan (1998). Study on site classification and design response spectra adjusting method, 

Post-doctor Research Report. Institute of Engineering Mechanics. 

Dou Lijun (2001). Site condition and design ground motion. Thesis for Doctor Degree, Institute of 

Engineering Mechanics. 

Li Xiaojun and Peng Qing (2001). Calculation and analysis of earthquake ground motion parameters 

for different site categories. Earthquake Engineering and Engineering Vibration 21:1, 29-36. 

Li Xiaojun, et al. (2001). Consideration of site effects for determination of design earthquake ground 

motion parameters. World Information on Earthquake Engineering 17:4, 34-41. 

96

ENGINEERING SEISMOLOGY AND 

GEOTECHNICAL ENGINEERING



Engineering Research, Hong Kane Volume 

DIGITIZATION AND DATA PROCESSING OF STRONG MOTION 

EARTHQUAKE ACCELEROGRAMS 

V. W. Lee 

Associate Professor, Dept. of Civil Engineering 

University of Southern California, Los Angeles, CA 90089-2531 

Keywords: 

Digitization, Data Processing, Earthquake Accelerorams 

ABSTRACT 

The modern techniques for digitization of strong-motion accelerograms and subsequent data 

processing were first developed in the late 1960s and early 1970s, at which time the only 

available digitization technology was semi-automatic hand digitization. Following the 

development of image processing technologies in the early 1970s, an automatic digitization and 

data processing system, driven by a mini-computer, was developed by Trifunac and Lee (1979). 

The currently used digitization technology is based on personal computers and flat bed digital 

scanners, which appeared respectively in the early and late 1980s. The data processing part has 

been continuously evolving, improving the performance of the entire system. This paper presents 

a review of the developments in hardware and software for digitization of strong-motion 

accelerograms, data processing and data dissemination. In the 1980s, digital strong motion 

accelerographs became commercially available and currently all new deployments are digital. 

However, it takes time to record a large number of strong motion accelerograms, and as of now, 

by far, most of the strong motion data has been recorded in analog form. 

1. INTRODUCTION 

The first strong motion accelerographs were analog, and were built and installed at free-field 

sites and in buildings in the early 1930s. Soon after, strong motion was first recorded, on March 

10, 1933, during the Long Beach, California earthquake. In the 1980s, digital strong motion 

accelerographs became commercially available (Trifunac and Todorovska, 200la), and at present 

essentially all new installations are digit 1 al. Yet, so far the largest number of strong motion 

records has been recorded in analog form, on light sensitive paper or on film. For analysis, 

analog strong motion records must be digitized and processed in a form suitable for subsequent

100 

analyses. This paper reviews the developments in digitization, processing (Trifunac et al., 

1999a,b), and dissemination of strong motion accelerograms(Lee and Trifunac, 1982). 

The concept of response spectrum as a tool in earthquake resistant design was formulated in 

1932, one year before the first recording of strong ground motion (Biot 1932; 33). Before the 

1960s, spectra had to be computed with the aid of analog mechanical devices (e.g. torsional 

pendulum; Biot, 1941; 42) or electrical analog computers (Caughy et al, 1960). The modern era 

of digitization and processing of strong motion accelerograms started in the 1960s (Fig. 1), with 

the availability of digital computers and semi-automatic digitizing machines (Hudson, 1979). 

In the late 1960s and early 1970s, the only available technology for digitization of strong motion 

accelerograms was a semi-automatic, hand digitization system (Hudson, 1979; Trifunac and Lee, 

1973). The semi-automatic digitization system was operating from 1967 to about 1975, It was 

slow (it took about four days to digitize, check and load data onto magnetic tapes), accurate, and 

its noise characteristics were analyzed and documented (Trifunac et al., 1971; 1973a,b). It was 

used to generate all the data published in the so-called Caltech "blue book data reports" (Hudson 

et al., 1969; 1971; 1972a,b), which became the "standard" source for strong motion data all over 

the world. The digitized data of many famous older accelerograms in the blue book reports, such 

as March 10, 1933 Long Beach (Heck et ah, 1936), May 18, 1940, El Centra (Trifunac and 

Brune, 1970), July 21, 1952, Taft (Hudson et al., 1969), and February 9, 1971, Pacoima Dam 

(Trifunac and Hudson, 1971), were all redigitized by Trifunac using this semi-automatic 

digitization system, because the previously digitized versions had inadequate sampling rate and 

occasionally missed or misinterpreted peaks in the strong motion part of the records. 

2. DIGITIZATION 

2.1 The Automatic Digitization System 

The first automatic system for digitization of strong-motion accelerograms was developed by 

Trifunac and Lee (1979), using the commercially available hardware for image processing 

(Photoscan P-1000 Optronics Photodensitometer) and a Data General NOVA-3 minicomputer. In 

the 1980s, it was replaced by the inexpensive high-speed personal computers (PCs) and desktop 

digital scanners that appeared on the market and reduced the cost of the system to several 

thousand dollars (Fig. 1). Our new software package "LeAuto," consisting of LeFilm, LeTrace, 

LeTV and LeScribe computer programs, now runs on a Pentiuum PC with 128 MB or more of 

RAM, running Windows98 or higher operating system, and with a fast Super VGA graphics 

monitor. The software has been continuously updated and improved to be compatible with the 

new scanners that have appeared on the market. By 2000, the cost of the hardware dropped to 

less than US $2,000 (Fig. 1). The current resolution of digitization is >200 points/s (assuming 

film speed of Icm/s), with gray levels resolved by 8, 10 or 12 bits (256, 1024 or 4098 gray levels) 

(Lee and Trifunac, 1990; Trifunac et al., 1999a,b). 

2.2 Hardware Components 

The elements of the new automatic digitization system are shown schematically in Fig. 2. The 

reading of an 8.5x11 inch film positive is performed by a Hewlett-Packard desktop digital 

scanner ("ScanJet"), interfaced with the PC. The PC needs to have a large hard disk drive and a

101 

minimum of 2 MB random access memory (RAM), while most modem PCs have 64, 128, 256 

MB or larger RAM. In the early 1990s, the automatic digitization system ran on operating 

system DOS version 5.00 or higher (preferably DOS 6.2), and currently (early 2000s) on 

Microsoft Windows 98 or higher operating system. The PC is interfaced with a Super VGA 

graphics display monitor, and a LaserJet printer, while other Windows supported (e.g. DeskJets 

and InkJet) printers can also be used. 

A typical strong-motion accelerogram is usually several tens of seconds long, but may be one 

minute or longer, if the complete length of useful recording is considered. Then, considering a 

typical SMA-1 70 mm wide film record, one would have to digitize a rectangular area 70 mm 

(2.75 in) wide and up to say 80 s (80 cm, 31.5 in) long. ScanJets are designed to scan a 

rectangular area up to 8.5x11 inches or 8.5x14 inches. At present, each 11-inch long segment 

("page") is scanned by LeFilm and written onto consecutive disk files. Program. LeTrace then 

processes (reads the scanned data and identifies the traces) each "page" sequentially. Using the 

program LeTV, the operator can edit each "page" one by one. The program LeScribe then reads 

the trace segments for each "page" and assembles them. This procedure thus enables one to 

digitize very long records. At present, LeFilm can also scan a 12-inch wide film records used by 

central recording system (CR-1), with 13 or more acceleration traces. 

2.3 Noise Characteristics of the New System 

The Response and Fourier amplitude spectra of the digitization noise for the overall automatic 

digitization process are evaluated when the system is first assembled, and when it is upgraded. 

These spectra are used to determine the frequency range in which the signal to noise ratio for a 

digitized accelerogram trace is greater than unity, and it represents accurately the recorded 

motions. In general, the amplitudes of this noise depend on the scanner resolution, but also and 

mostly on the thickness of the trace and on the record length. Comparing records from 

acceleration and displacement transducers, Trifunac and Lee (1974) found that, for the old semiautomatic 

digitization, the typical digitization noise, after integration, might result in 

displacements of up to several centimeters. 

While for the old semi-automatic hand digitization system (Trifunac and Lee, 1973) the 

horizontal resolution (number of digitized points per unit length of the record) was a major factor 

determining the noise characteristics of the system. For the automatic system with a flat bed 

scanner used today, the typical resolution is 300, 600 or 1200 points/inch (dpi). Then, for 1 cm/s 

film speed and 600-dpi scan, the resolution is 236 points/s implying Nyquist frequency of 118 

Hz-more than sufficient for most recorded accelerograms. 

The "noise acceleration traces" were created as follows. For several SMA-1 accelerograms, the 

pair of baselines was digitized and processed (scaled, instrument and baseline corrected, and 

band pass filtered between 0.07 and 25, Hz) considering one of them as a "zero" acceleration 

trace and the other one as a zero baseline, and Response and Fourier amplitude spectra were 

finally calculated. Figures 3a and b (Lee and Trifunac, 1990) show smoothed average spectral 

amplitudes for five damping values (0, 2, 5, 10 and 20 percent of critical) for 45 and 90 s long 

records. The overall spectrum amplitudes are similar to those for the automatic digitization 

system with an Optronics scanner (Trifunac and Lee, 1979).

102 

A comparison of spectral amplitudes of noise of various recorders with those typical of recorded 

strong earthquake ground motion for magnitudes 4 < M < 7 and for frequencies 0.1

103 

digitization scheme. The computer programs (Trifunac and Lee, 1973) were first written for the 

IBM 7094 and IBM 360 computers. These programs involved the following steps: 

(i) Volume I Processing (scaled data; Hudson et al, 1969): 

The half-second timing marks are first checked for "evenness" of spacing, and then smoothed by 

a 1/4, 1/2, 1/4 running average filter. The ^-coordinates of each trace are next scaled to units of 

time in seconds. Each fixed trace (baseline) is smoothed and subtracted from the corresponding 

acceleration trace, with the ^-coordinates subsequently scaled to units g/10 (g = 9.81 m/s 2 ). 

(ii) Volume II Processing (corrected data; Hudson et al., 1971): 

The scaled uncorrected Volume I acceleration data is next corrected for instrument response 

(Trifunac, 1972) and baseline adjustment (Trifunac, 1970; 1971). The data is first low-pass 

filtered with an Ormsby filter having a cutoff frequency f c = 25 Hz and a roll-off termination 

frequency / = 27 Hz. Instrument correction is next performed using the instrument constants. 

These constants are the natural frequency and ratio of critical damping of the instrument, 

considered as a single-degree-of-freedom system (Trifunac and Hudson, 1970; Todorovska, 

1998). These are determined from calibration tests for each accelerograph transducer. The data 

is then baseline corrected by a high-pass Ormsby filter. The cutoff and roll-off frequencies of 

the filter are usually determined from the signal-to-noise ratio of each component (Trifunac and 

Lee, 1978). The acceleration data is then integrated twice to get velocity and displacement. To 

avoid long period errors resulting from uncertainties in estimating the initial values of velocity 

and displacement, the computed velocities and displacements are high-pass filtered at each stage 

of integration, using the Ormsby filter with the same cutoff and roll-off frequencies as for the 

corrected accelerogram (Hudson et al., 1971). 

(in) 

Volume III Processing (response spectra; Hudson et aL, 1972a): 

Using an approach based on an exact analytical solution of the Duhamel integral for successive 

linear segments of excitation, the Fourier and Response Spectra for up to 91 periods and 5 

damping ratios are calculated. The times of maximum response for all periods and damping 

ratios are also recorded (Lee and Trifunac, 1979; 1986). 

Following the development of the automatic digitization system (Trifunac and Lee, 1979), the 

above computer programs, originally developed in 1969/70 (Trifunac and Lee 1973), were 

modified in 1978/79 (Trifunac and Lee 1979) to run on a Data General mini computer. The new 

approach, which took advantage of image processing techniques, increased the speed of the 

overall data processing by one order of magnitude (Fig. 1). 

(iv) Volume IVProcessing (Fourier amplitude spectra; Hudson et al., 1972b) 

Using the Fast Fourier Transform (FFT) algorithms (Cooley and Tukey, 1965) for applications 

that require equally spaced data on Fourier amplitude spectra, Volume IV data were computed 

and presented in the series of "Blue Book Data Reports" since 1972. This data was used in 

numerous empirical studies of Fourier spectrum amplitudes. It offered an opportunity to 

routinely identify significant frequencies in the records using Fisher (1929) test of significance in 

harmonic analyses. At present, Volume IV data processing is not performed routinely because of

104 

the high speed with which FFT can be calculated with modern computers, as required for each 

particular application. 

3.2 The Current System 

In general, the principles and the requirements governing the routine data processing of strongmotion 

records have changed little, if any, since the early 1970s. However, since then, we have 

witnessed a remarkable progress in digital signal processing techniques, in their accuracy, 

efficiency, speed of execution and in the major hardware cost reduction. These improvements 

have been incorporated into the routine data processing of strong-motion accelerograms (Lee and 

Trifunac, 1984,1990). 

4. HIGHER-ORDER CORRECTIONS 

So far in this paper, the standard data processing algorithms were described, which are suitable 

for routine, large scale processing of analog strong motion records. For specialized studies 

requiring higher accuracy further corrections may be required. In the following we describe two 

such corrections. 

4.1 Misalignment and Cross-axis Sensitivity 

In routine data processing of three-component acceleration records it is assumed that the three 

sensitive axes (longitudinal, transverse and vertical) are mutually perpendicular. Careful tilt table 

measurements show however that this is not so and that each sensitivity vector can be misaligned 

by small (few degree) angles (Todorovska 1998; Todorovska et al, 1995; 1998; Wong and 

Trifunac, 1977). For accelerometers, which are sensitive to static tilt, the misalignment angles 

can be measured by simple tilt tests followed by data processing (Todorovska et al., 1995; 1998). 

These angles then can be used to perform exact corrections of cross-axis sensitivity and 

misalignment, resulting in corrected accelerations along three mutually orthogonal coordinate 

axes. 

Measurement of the misalignment angles and corrections for misalignment and cross-axis 

sensitivity have been performed for all instruments of the Los Angeles strong motion network 

(Todorovska et al., 1998) and for the stations of the Los Angeles Department of Water and 

Power (Todorovska et al., 1999) for strong motion data recorded during and following 

Northridge, California earthquake of 17 January, 1994. 

4.2 Corrections for Tilt and Angular Accelerations 

Let Xj, X 2 and X 3 be the mutually perpendicular coordinate axes of displacement in longitudinal 

(L), Transverse (T) and Vertical (V) directions, and let 0 y , 0 2 and fa represent the rotations about 

X l9 X 2 and X 3 directions. For small transducer deflections y t = r, a, where a { is the angle of 

deflection of the z-th pendulum from its equilibrium position, and r t is its corresponding lever 

arm, the equations of motion of the three penduli of an accelerometer (e.g. SMA-1) are 

(Todorovska, 1998): 

&i + %2 a i 

(2a)

105 

(2b) 

where a;, and £ are respectively the natural frequency and fraction of critical damping of the i-th 

transducer. The second and third terms on the right hand side of equation (2a) and (2b) represent 

contributions from tilting ((p, and 0 3 ) and angular acceleration (

106 

8. Gold, B. and C.M. Rader (1975). "Digital Processes of Signals", McGraw Hill, New York, New York. 

9. Gupta, I.D. Rambabu, V., Joshi, R.G., Trifunac, M.D., Todorovska, M.I and Lee, V.W. (1993), "Strong 

Earthquake Ground Motion Data in EQINFOS for India": Part IA, Dept. Civil Eng. Report No. CE 93-03, Univ. 

of Southern Calif., Los Angeles, California. 

10. Hudson, D.E. (1970). "Ground motion Measurements in Earthquake Engineering", in Earthquake Engineering 

edited by R.L. Wiegel, Prentice Hall Inc., Englewood Cliffs, New Jersey. 

11. Hudson, D.E. (1976). "Strong Motion Earthquake Accelerograms-Index Volume", Earthquake Eng. Research 

Laboratory, Report EERL 76-02, Calif. Inst. of Tech., Pasadena, California. 

12. Hudson, D.E. (1979). "Reading and Interpreting Strong Motion Accelerograms", Earthquake Eng. Res. Institute, 

Monograph Series, 262 Telegraph Ave., Berkeley, California. 

13. Hudson, D.E. (1984). "Golden Anniversary Workshop on Strong Motion Seismology March 30-31", 1983, 

Dept. of Civil Eng., Univ. of Southern California. 

14. Hudson, D.E., Brady, A.G. and Trifunac, M.D. (1969). "Strong-Motion Earthquake Accelerograms, Digitized 

and Plotted Data", Vol. 1, Earthquake Engineering Research Laboratory, Report EERL 70-20, California 

Institute of Technology, Pasadena, California. 

15. Hudson, D.E., Trifunac, M.D., Brady, A.G. and Vijayaragharan, A. (1971). "Strong-Motion Earthquake 

Accelerograms, II, Corrected Accelerograms and Integrated Velocity and Displacement Curves", Earthquake 

Engineering Research Laboratory, Report EERL 71-51, California Institute of Technology, Pasadena, 

California. 

16. Hudson, D.E., Trifunac, M.D. and Brady, A.G. (1972a). "Strong-Motion Accelerograms, III, Response Spectra", 

Earthquake Engineering Research Laboratory, Report EERL 72-80, California Institute of Technology, 

Pasadena, California. 

17. Hudson D.E., Trifunac, M.D., Udwadia, F.E., Vijayaraghavan, A. and Brady A.G. (1972b). "Strong Motion 

Earthquake Accelerograms, IV, Fourier Spectra", Earthquake Engineering Research Laboratory, Report EERL 

71-100, California Institute of Technology, Pasadena, California. 

18. Lee, V.W. (1984). "A New Fast Algorithm for the Calculation of Response of a Single-Degree-of-Freedom 

System to Arbitrary Load in Time", Int. J, Soil Dynamics and Earthquake Eng., 3(4), pp. 191-199. 

19. Lee, V.W. (1990). "Efficient Algorithm for Computing Displacement, Velocity, and Acceleration Responses of 

an Oscillator for arbitrary Ground Motion", Int. J. Soil Dynamics and Earthquake Eng., 9(6), pp. 288-300. 

20. Lee, V.W. and Trifunac, M.D. (1979). 'Time of Maximum Response of Single-Degree-of-Freedom Oscillator 

for Earthquake Excitation", Department of Civil Engineering, Report No. 79-14, University of Southern Calif., 

Los Angeles, California. 

21. Lee, V.W. and Trifunac, M.D. (1982). "EQUINFOS (The Strong-Motion Earthquake Data Information 

System)", Department of Civil Engineering, Report No. 82-01, University of Southern Calif., Los Angeles, 

California. 

22. Lee, V.W and M.D. Tnfunac (1984). "Current Developments in Data Processing of Strong Motion 

Accelerograms", Report No. CE 84-01, Dept. of Civil Eng., Univ. Southern California, Los Angeles 

California. 

23. Lee, V.W. and Trifunac, M.D. (1986). "A Note on Time of Maximum Response of Single Degree of Freedom 

Oscillator to Earthquake Excitation", Int. J. Soil Dynamics and Earthquake Eng., 5(2), pp. 119-129. 

24. Lee, V.W. and Trifunac, M.D. (1987). "Strong Motion Ground Motion Data in EQUINFOS: Part I, Department 

of Civil Engineering", Report No. 87-01, University of Southern California, Los Angeles, California. 

25. Lee, V.W. and Trifunac, M.D. (1989). "A Note on Filtering Strong Motion Accelerograms to Produce 

Response Spectra of Specified Shape and Amplitude", European Earthquake Eng., Vol. Ill, No. 2, pp. 38-45. 

26. Lee, V.W. and Trifunac, M.D. (1990). "Automatic Digitization and Professing of Accelerograms Using PC", 

Dept. of Civil Eng., Rep. No. CE 90-03, Univ. Southern California, Los Angeles, California.

107 

27. Lee, V.W. and Y.Y. Wang (1983). "On the Instrument Correction of the RDZ-I Strong- Motion Pendulum 

Galvanometer in China", Earthquake Eng. and Eng. Vibration, Vol. 3, Part 4. Dec. 

28. Lee, V.W., Trifunac, M.D. and Amini, A. (1982). "Noise in Earthquake Accelerograms", ASCE Eng Mech 

D/v., 108, pp. 1121-1129. 

29 Nenov, D., Georgiev, G., Paskaleva, I., Lee, V.W. and Trifunac, MD. (1990). "Strong Earthquake Ground 

Motion Data in EQINFOS: Accelerograms Recorded in Bulgaria Between 1981 and 1987", Bulg. Academy of 

Sciences, CLSMEE, Sofia 1990, also Rep. No. 90-02, Dept. of Civil Eng., Univ. Southern California, Los 

Angeles, California. 

30. Nigam, N.C. and Jennings, P.C., (1969). "Calculation of Response Spectra from Strong-Motion Earthquake 

Records", Bull. Seism. Soc, Am., 59, pp. 909-922. 

31. Novikova, E.I. and M.D, Trifunac (1991). "Instrument Correction for a Coupled Transducer-Galvanometer 

System", Report No. CE 91-02 Dept. of Civil Eng., Univ. Southern California, Los Angeles California. 

32. Novikova, E.I. and M.D. Trifunac (1992). "Digital Instrument Response Correction for the Force-Balance 

Accelerometer", Earthquake Spectra, 8(3), pp. 429-442. 

33. Todorovska, M.I. (1998). "Cross-axis Sensitivity of Acceierographs with Pendulum Like Transducers: 

Mathematical Model and the Inverse problem", Earthquake Eng. and Structural Dynamics, 27(10), pp. 1031- 

1051. 

34. Todorovska, M.L, Trifunac, M.D., Novikova, E.I. and Ivanovic, S.S. (1995). "Correction for Misalignment and 

Cross Axis Sensitivity of Strong Earthquake Motion Recorded by SMA-I Acceierographs", Dept. of Civil Eng. 

Report No. 95-06, Univ. Southern California, Los Angeles, California. 

35. Todorovska, M.L, Trifunac, M.D., Novikova, E.I. and Ivanovic, S.S. (1998). "Advanced Acceleerograph 

Calibration of the Los Angeles Strong Motion Array", Earthquake Eng. and Strut. Dynamics, 27(10), pp. 1053- 

1068. 

36. Trifunac, M.D., (1970). "Low frequency Digitization Errors and a New Method for Zero Baseline Correction of 

Strong Motion Accelerograms, Earthquake Engineering Research Laboratory", Report EERL 70-07, California 


37. Trifunac, M.D. (1971). "Zero Baseline Correction of Strong-Motion Accelerograms", Bull. Seism. Soc. Am., 

62, pp. 401-409. 

38. Trifunac, M.D. (1972). "A Note on Correction of Strong-Motion Accelerograms for Instrument Response", Bull 

Seism. Soc. Am., 62, pp. 401-409. 

39. Trifunac, M.D. (1977). "Uniformly Processed Strong Earthquake Ground Accelerations in the Western United 

States of America for the Period from 1933 to 1971: Pseudo Relative Velocity Spectra and Processing Noise", 

Department of Civil Engineering, Report No. CE 77-04, University of Southern Calif., Los Angeles, California. 

40. Trifunac, M.D. and Brune, J.N. (1970). "Complexity of Energy Release During Imperial Valley", California, 

Earthquake of 1940, Bull. Seism. Soc. Am., 60, pp. 137-160. 

41. Trifunac, M.D. and Hudson, D.E. (1970). "Laboratory Evaluation and Instrument Corrections of Strong Motion 

accelerographs", Earthquake Engineering Research Laboratory, Report EERL 70-04, California Institute of 

Technology, Pasadena, California.. 

42. Trifunac, M.D. and Hudson, D.E. (1971). "Analysis of the Pacoima Dam Accelerogram-San Fernando, 

California, Earthquake of 1971", Bull Seism. Soc. Amer., 61(5), pp. 1393-1411. 

43. Trifunac, M.D and Lee, V.W. (1973). "Routine Computer Processing of Strong Motion Accelerograms", 

Earthquake Eng. Res. Lab., Report EERI 73-03, Calif. Inst. of Tech., Pasadena, California. 

44. Tnfunac, M.D. and Lee, V.W. (1974). "A Note of the Accuracy of Computed Ground Displacements from 

Strong-Motion Accelerogram", Bull Seism. Soc. Amer., 64, pp. 1209-1219. 

45. Trifunac, M.D., and V.W. Lee (1978). "Uniformly Processed Strong Earthquake Ground Accelerations in the 

Western United States of America for the Period from 1933 to 1971: Corrected Acceleration, Velocity and

108 

Displacement Curves", Dept. of Civil Engr., Report No. CE 78-01, Univ. of Southern California, Los Angeles, 

California. 

46. Trifunac, M.D. and V.W. Lee (1979). "Automatic Digitization and Processing of Strong Motion 

Accelerograms", II and I. Dept. of Civil Eng., Report No. 79-15, Univ. of Southern California, Los Angeles, 

California. 

47. Trifunac, M.D. and Todorovska, ML (200la). "Evolution of Accelerographs, Data Processing, Strong Motion 

Arrays and Amplitude and Spatial Resolution in Recording Strong Earthquake Motion", Soil Dynamics and 

Earthquake Eng., 21(6), pp. 537-555. 

48 Trifunac, M.D. and Todorovska, M.I. (200Ib). "A Note on Useabie Dynamic Range in Accelerographs 

Recording Translation", Soil Dynamics and Earthquake Eng., 21(4), pp 275-286. 

49 Trifunac, M.D , Udwadia, F.E. and Brady, A.G. (1971). "High Frequency Errors and Instrument Corrections of 

Strong-Motion Accelerograms", Earthquake Engineering Research Laboratory, Report EERL 71-05, California 


50. Trifunac, M.D, Udwadia, FE. and Brady, A.G. (1973a). "Analysis of Errors in Digitized Strong-Motion 

Accelerograms", Bull. Seism. Soc. Amer., 63, pp. 157-187. 

51. Tnfunac, M.D, Udwadia, RE. and Brady, A.G. (1973b). "Recent Developments in Data Processing and 

Accuracy Evaluations of Strong Motion Acceleration Measurements", Proc. 5 th World Conf. Earthquake 

Engineering, Rome, Italy, Vol. I, pp. 1214-1233. 

52 Trifunac, M.D, Todorovska, M.I. and Lee, V.W. (1999a). "The Rinaldi Strong Motion Accelerogram of the 

Northridge California Earthquake of 7 January, 1994", Earthquake Spectra, 14(1), pp. 225-239. 

53. Tnfunac, M.D., Todorovska, M.I. and Lee, V.W. (1999b). "Common Problems in Automatic Digitization of 

Strong Motion Accelerograms", Soil Dynamics and Earthquake Eng., 18(7), pp. 519-530. 

54. Wong, H.L. and Trifunac, M.D. (1977). "Effects of Cross-Axis Sensitivity and Misalignment on Response of 

Mechanical-Optical Acceierographs", Bull Seism. Soc. Amer., 67, pp. 929-956.

109 

FIGURE CAPTIONS 

Fig. 1 Trends in capabilities and cost of accelerogram digitization and data processing 

systems. 

Fig. 2 Hardware components of the automatic digitization and data processing system. 

Fig. 3 Fourier spectra (dashed lines) and Response Spectra (solid lines) for damping £ 0 = 0, 0.02, 

0.05, 0.10 and 0.20 (top to bottom), for records 45 seconds long (a-left) and 90 seconds long (bright). 

Fig. 4 Comparison of different noise spectra associated with different recording instruments and 

different methods of digitization (modified from Trifunac and Todorovska, 200Ib). 

Fig. 5 Comparison of Fourier amplitude spectra of translation, X t with spectra of contribution 

from fyg +0/ A , analog digitization noise, digital digitization noise (PDR), and microtremor and 

microseism noise (redrawn from Trifunac and Todorovska, 200Ib).

I 000,000 

100,000 

Number of uniformly processed records 

in USC strong motion database 

Cost of 

digitization system $ 

10,000 

Mechanical Spectrum Analyzer 

Biot (1941,1942) 

Scanner 

resolution 

Capacity of 

disk storage - Mb 

1,000 

100 

10 

Time required to compute one set 

of standard response spectrum 

curves for five damping values - mm 

Optronics 

drum scanner 

Typical record 

digitization 

ime - mm 

Mechanical 

analyzers 

Analog 

computers 

Nova (1977) 286 386 486 'Pentium 

Mini 

Personal computers 

computers, computer^ 

1940 1950 1960 1970 1980 Year 1990 2000 

Figure 1

Figure 2 

m

Response and Fourier Spectra of Digitization Noise 

"1' i ii n 1 1—i'"i" i'i M| 1 1—rnrrrrrj— 

PSV 

CO 

10. 

10* 

Record length ~ 45, 

Record length = 90 s 

Period - s 

10 

t I t I 1 

I 

1 

Period -s 

10 

Figure 3

113 

TT] r—i j 11iii| 1 i n inij r i i niiij 1 i i | n^ 

10 3 

Extrapolated Fourier amplitude spectra of acceleration 

R=10 km, H= 0 krn SMA-3/SMP-1 

Photoscan/ pStrong *& em 

- noise in 

digitization/ \ motion digitization 

- Japanese noise 

noise 

10 2 PHR! 

~ accelerogram 

£ M-125 of fixed 

- straight linec/) 

10' 

010° 

J, 

\JL 

10- 1 

io- 2 

10" 3 

1C" 3 10' 2 10" 1 10° 10 1 

Frequency - Hz 

Figure 4

114 

10 3 

10 2 r 

| I lll| I | I | I liq 4 I I 1 I ill) I I I 1 I )ll| 

Extrapolated Fourier amplitude spectra of acceleration 

R=10 km, H= 0 km, s=2, s L =0 

Typicai digitization noise 

in analog records digitized 

manually or automatically 

SMA-1,50dB 

DR, 11 bits, 60 dB 

Noise in 

PDR-1+FBA-13 

10- 3 10 1 10 2 

Frequency - Hz 

Figure 5




EVALUATION OF DYNAMIC SOIL PROPERTIES AND 

LIQUEFACTION POTENTIAL BY SEISMIC PIEZOCONE TESTS 

T. Liao and P.W. Mayne 

Georgia Institute of Technology, School of Civil & Environmental Engineering, 

Atlanta, GA 30332-0355 USA 

ABSTRACT 

Seismic piezocone tests (SCPTu) provide an expedient and efficient means to obtain site-specific 

information on the soil stratigraphy, dynamic soil properties, and ground liquefaction susceptibility, 

since five separate readings are obtained hi the same sounding: cone tip stress (q-r), sleeve friction (f s ), 

penetration porewater pressure (ut>), dissipation-time recordings (tso), and shear wave velocity (V s ). The 

shear wave data are used to determine the small-strain stiffness that is required in ground amplification 

analyses (e.g., SHAKE, DESRA, DEEPSOIL). Layering and soil types are evaluated from the 

penetration readings to detect soils that are prone to liquefaction (i.e., sands, silty sands). The 

liquefaction potential is assessed using two independent readings from the sounding that are 

normalized for the effects of effective confining stress levels (q-n and V s i). Illustrative results from 

seismic piezocone tests at a paleoliquefaction site in the New Madrid Seismic Zone in midwestern 

USA are presented. 

INTRODUCTION 

Dynamic soil properties and liquefaction potential are among the major concerns in evaluating the level 

of ground shaking and disruption that might occur during earthquakes. Key dynamic soil properties 

include the small-strain shear modulus (Gmax), Poisson's ratio (v), damping ratio (D s ) (Woods, 1978; 

Campanella, 1994). Parameters related to the liquefaction potential of soils are addressed separately 

because of their unique destructive nature. Dynamic soil properties and liquefaction potential can be 

evaluated by laboratory techniques, such as strain-rate-test, resonant column, ultrasonic pulse, cyclic 

(triaxial, simple shear, torsional shear) test, bender-element, and shake table. Although the lab tests are 

versatile and can provide the desirable dynamic soil properties, undisturbed samples of sandy soils are 

either very expensive or impossible to obtain. Furthermore, after the soil samples are taken from the 

field, their in-situ state cannot be fully recovered in the lab, leading to errors hi test results. For these 

reasons, field testing is becoming more popular for site characterization in seismic areas and the 

evaluation of dynamic soil properties and liquefaction potential. 

The most widely used field technique is the standard penetration test (SPT), but it is not realistic for the 

single N-value from this test to fulfill the task to provide the many parameters needed for engineering

116 

analysis. Other techniques include non-intrusive surface geophysics (e.g. seismic refraction/reflection, 

steady-state surface wave, spectral analysis of surface wave) and borehole geophysics (e.g. cross-hole, 

down-hole, up-hole, supension PS logging). These tests provide the compression wave velocity (V p ) 

and/or shear wave velocity (V s ) that are needed to derive the dynamic moduli. Data collected by the 

non-intrusive surface geophysics methods are difficult to interpret without experienced users. The cost 

of the borehole geophysics methods is very high, because the holes are drilled, cased, and grouted, then 

cured for several days before the tests are performed (Figure 1). 

The seismic piezocone test (SCPTu) is a hybrid field method that combines the virtues of the cone 

penetration test (CPT) with downhole geophysics (Campanella, et al., 1986). Five separate readings are 

obtained in the same sounding: cone tip stress (qi), sleeve friction (f s ), penetration porewater pressure 

(Ub), dissipation-time recordings (tso), and shear wave velocity (V s ), providing site-specific information 

on the geostratigraphy, dynamic soil properties, and ground liquefaction susceptibility (Figure 2). 

Compared with the SPT and the geophysics methods mentioned before, the SCPTu is more costeffective, 

its procedure is simpler and more standardized, and its continuous record of soil properties 

with depth enables a better look at soil variability. 

bownhole Testing 

Horizontal Plank 

Seismic Piezocone 

Tests (SCPTu) 

; Shear Wave Velocity K-^2!S 

[ V,. = AR/A-t 

Figure 1. Conventional Downhole 

Geophysics Test Using Cased Borehole 

Figure 2. Soil Property Evaluation from 

Seismic Piezocone Test Results 

SEISMIC PIEZOCONE TEST 

The SCPTu system used by the authors includes 5-, 10-, and 15-tonne Hogentogler penetrometers, field 

computer, and anchored hydraulic rig. Each penetrometer consists of a 60° angled-apex cone at the 

front end, two load cells, a transducer, an inclinometer, and one geophone. The front load cell directly 

measures the force over the tip area to give the cone tip resistance, q c . A second load cell records the 

axial force over the cylindrical sleeve area to provide the sleeve friction, f s . A saturated porous element 

is used to measure the porewater pressure. The pressures m and U2 are measured either at the midface 

of the tip or at the cone shoulder, respectively. An inclinometer is used to monitor the vertical angle, i, 

as a warning signal that the rods might buckle or shear during hard pushing. The geophone is 

incorporated at the back end of the penetrometer to provide a direct measure of the arrival times from a 

downhole shear wave.

117 

The test is conducted by advancing an instrumented electronic probe with a hydraulic pushing system 

at constant rate of 20 mm/s. The penetration data are logged continuously by computer giving 

unparalleled details on the soil layering profiles, while downhole geophysics are taken at 1-m intervals 

during the successive addition of rods. The 6-tonne truck utilizes twin-screw earth anchors installed to 

depths of 2 m for increased reaction. With lightweight rigs, soundings of 30 m can be achieved in 3 

hours time. Using larger 25-tonne vehicles, penetration depths of up to 90 m are possible. 

The tip resistance (q c ) is a point stress related to bearing capacity and soil strength. The measured q c 

must be corrected for porewater pressure effects on unequal tip areas (Lunne, et. al., 1997), especially 

in clays and silts, and this corrected value is termed q T . The uo position element is required for this 

correctio. The tip resistance (q-r), sleeve friction (f s ), and porewater pressures (ui or 112) are used to 

characterize the subsurface layering, soil behavioral type, and strength properties. Porewater dissipation 

(tso) can be conducted to measure permeability (k) and coefficient of consolidation (c h ). 

Shear waves are generated by striking a horizontal steel plank at the ground surface. A downhole 

geophone is oriented parallel to the plank to detect the horizontally-polarized shear waves. From the 

measured wave train at each depth, a pseudo-interval shear wave velocity (V s ) is determined. The 

velocity is the difference in travel distance between any two successive events divided by the 

difference in travel times (Campanella et al., 1986). The seismic data are useful for subsequent site 

amplification studies, as well as the evaluation of soil resistance to liquefaction susceptibility. 

DYNAMIC SOIL PROPERTIES 

During the cyclic loading by earthquakes, soils exhibit nonlinear and hysteretic stress-strain behavior. 

The secant shear modulus G decreases with the shear strain level j c , but with a less rapid decay than 

that in the condition of mono tonic loading (Figure 3). At very low strain levels (less than 10") the 

small-strain shear modulus (G ma x - GO) is operational and the soil behaves as an elastic medium. As the 

strain goes beyond 10" 5 , the soil response becomes nonlinear, and at large strains (larger than 10" 3 ), the 

soils experience permanent plastic deformation and eventually reach an unstable condition at peak 

strength. Since ground motions during earthquakes may involve both small and large strains, the whole 

curve of shear modulus reduction is desirable for engineering computation and analysis. 

The magnitude of the snear strain for plane waves can be given by the expression j c = u/V s , where u 

is the peak particle velocity (Stokoe, et al., 1978). It is found that the shear strain caused by shear 

waves generated during the SCPTu travels through the soil mass below the elastic strain level. The 

Gmax can therefore be obtained from the following equation: 

Where p T is the total mass density p r = j r /g and g = 9.8m/V . Mayne (2001) compiled data of unit 

weight y sat and V s from saturated geomaterials (Figure 4) to define the following trend: 

(2) 

where y sat is in kN/m 3 , V s in m/s, and z in meters.

It has been found that G max applies to both the initial static monotonic loading and dynamic loading of 

geomaterials (Jardine, et aL, 1991). Moreover, since very small strains have not yet generated any 

excess porewater pressures, G max can be applied to both drained and undrained soil behavior. Once 

G max is known, the shear modulus reduction curve can be derived from the plastic index (PI) of the 

soils (Vucetic & Dobry, 1991). Given the shear modulus G, the corresponding Young's modulus E, 

Bulk modulus B ! , and constrained modulus M' can be derived from the Poisson's ratio v. The range of 

the value of Poisson's ratio v is from 0.1 to 0.2 at these strains. 

118 

0 0001 0 001 

Shear Strain 

Figure 3. Secant Shear Modulus Reduction 

for Monotonic & Cyclic Loading 

100 1000 

Sicar Wsve Velocity, V s (ins) 

Figure 4. Correlation for Unit Weight from 

Depth and V s (Mayne, 2001) 

The amplitude information of the collected shear waves can be used to analyze the material damping in 

terms of damping ratio (D s ) at small-strains. Stewart & Campanella (1993) suggested that the spectral 

ratio slope (SRS) method is the most reliable and consistent approach to obtain the damping ratio D s of 

a soil layer by the SCPTu, but it still requires verification on different soil types to fully validate the 

technique (Campanella, 1994). D s increases with the strain level, and the relationship between D s and 

shear strain y c has also been correlated with the PI (Vucetic & Dobry, 1991). 

LIQUEFACTION EVALUATION 

Recent earthquakes in Gujarat, India (2001), Izmit, Turkey (1999), Chi-Chi, Taiwan (1999) and Kobe, 

Japan (1995) have emphasized the importance of liquefaction hi geotechnical engineering because of 

its significant destructive nature. Liquefaction predominantly occurs in sands to silty sands that are 

difficult or impossible to sample. This warrants the use of field testing, especially the SCPTu, for 

liquefaction analysis, since the SCPTu provides not only the tip resistance qr and shear wave velocity 

V s , which can quantity the susceptibility to liquefaction (Robertson & Wride, 1998; Andrus & Stokoe, 

2000), but also obtains the sleeve friction f s and porewater pressure u 2 that are used for characterizing 

the soil types, layering, and geotechnical engineering parameters. Liquefaction notably occurs at depths 

less than 20 meters, and research found that the presence and thickness of the overlying clay cap 

significantly affects the liquefaction potential of underlying sand layers (Ishihara, 1985; Youd & 

Gams, 1995). Evaluating the liquefaction susceptibility of a certain soil layer and its potential for 

ground damage cannot therefore be confined to testing sandy soil samples only, but requires the 

characterization of its overlying capping clay layer as well

119 

From the collected SCPTu data, the layering and soil types can be evaluated to detect soils that are 

prone to liquefaction (i.e., sands, silty sands) using various soil behavioral classification charts, such as 

proposed by Robertson (1990). Soil layers susceptible to liquefaction are further analyzed for their 

liquefaction potential corresponding to certain magnitude earthquakes. 

In liquefaction analyses, the seismic loading expressed in terms of the cyclic stress ratio (CSR) can be 

given as (Seed & Idriss, 1971): 

(3) 

where a max is the peak ground acceleration (PGA), g is the acceleration of gravity, a vQ and

120 

PALEOLIQUEFACTION STUDIES IN MED-AMERICA 

The New Madrid Seismic Zone (NMSZ) is considered to be a highly-seismic region hi the United 

States. During 1811 and 1812, more than 200 separate earthquake events occurred hi the region, with 

the three largest ones estimated to have moment magnitudes of 7.9, 7.6, and 8.0 (Johnston & Schweig, 

1996). Due to rapid development in the late twentieth century, the NMSZ has become a highly 

populated area and hosts large cities, including Memphis, Tennessee and St. Louis, Missouri. The 

inevitable recurrence of the large seismic events would cause widespread loss of life and damage to 

structures, because of the ground shaking and soil liquefaction. In order to map the seismic ground 

hazards and dynamic soil properties in the NMSZ, series of SCPTu tests were conducted in this region. 

A representative SCPTu sounding performed at the Walker paleoliquefaction site near Marked Tree, 

Arkansas is shown in Figure 6. The small-strain shear modulus and the soil stratigraphy for this 

sounding are also shown. The site mainly consists of sandy soils, with a capping layer of clay between 

the depths of 2 to 5 meters. This site was selected for study because two large sand dikes were found on 

the property. Recent research hi Japan and western USA, where earthquakes occur more frequently, 

notes the important finding that sites that experience liquefaction will likely liquefy again during 

earthquake events of similar magnitude (Youd, 1988; Yasuda & Tohno, 1988). Therefore, the Walker 

site has the potential to liquefy again during large earthquakes hi the future. 

Tip Resistance Sleeve Friction Pore Pressure 

qT(MPa) f, (kPa) u2, uB(kPa) 

0 20 40 0 200 400 -200 0 200 400 

SW Velocity Shear Modulus 

V,(m/s) 

GmM(MPa) 

0 200 400 0 

1 °0 200 

Figure 6. Seismic Piezocone Results at Walker Paleoliquefaction Site in Eastern Arkansas 

The procedure of liquefaction evaluation by SCPTu is depicted in Figure 7, indicating the use of q t , f s , 

and Ub for soil delineation, V s for obtaining G ma x hi site amplification analyses (a raax ), and normalized 

q t i and V sl for evaluating liquefaction potential. The results of liquefaction analysis from the SCPTu 

sounding at the Walker paleoliquefaction site corresponding to a magnitude M w =8.0 earthquake are 

given in Figure 8. At a factor of safety of Fs = 1.0 for liquefaction, the critical values of tip resistance 

and shear wave velocity are superimposed with the measured data. This critical value, which changes 

with depth, is shown in the same graph as the measured value in the figure. Liquefaction is likely 

whenever the measured value is less than the critical value. The gaps in the critical value plots 

represent soil layers that are not susceptible to liquefaction due to their classification (e.g., clayey soils 

for which liquefaction analyses are not relevant).

121 

Piesocone Tests to Evaluate Seismic Ground Hazards 

^J-aywlyVWf 

vj 

J%UT. 

I** 2, y,2 

Uy#« s, v,s 

Cayw-4; \^ 

Soil Type 

-Nonliquefactron 

Susceptible 

j ' 

'"'-' ''' 

Site-Ampiificafi 

(SHAKE Profile 

*qT^ 

sc- Based qel 

Figure 7. Procedures for Liquefaction 

Evaluation by SCPTu 

Figure 8. Liquefaction Analyses for Walker 

Site SCPTu Sounding 

CONCLUSIONS 

The seismic piezocone (SCPTu) provides five soil measurements in a single sounding, including: tip 

resistance, sleeve friction, porewater pressure, dissipation, and shear wave velocity. It is an efficient 

way to assess soil stratigraphy, dynamic soil properties, and liquefaction potential in seismically active 

areas. The shear wave velocity (V s ) is used to obtain the small-strain shear modulus (G^ that is 

required in site-specific ground amplification studies. Both deterministic and probabilistic approaches 

are available to define the degree of susceptibility to liquefaction. The soil resistance to liquefaction 

can be ascertained from either the normalized tip stress qri and/or the normalized shear wave velocity 

V s i. Such redundancy provides a higher level of confidence in the conclusions. 


Financial and technical support has been provided by the Mid-America Earthquake Center (MAE) 

under grant EEC-9701785 by the National Science Foundation (NSF). 

REFERENCES 

Andrus, R.D. and Stokoe, K.H. (2000). Liquefaction resistance of soils based on shear wave velocity. 

Journal of Geotechnical &. Geoenvironmental Engineering 126:11, 1015-1026. 

Campanella, R.G. (1994). Field methods for dynamic geotechnical testing. Dynamic Geotechnical 

Testing II, (STP1213), American Society for Testing and Materials, Philadelphia, 3-23. 

Campanella, R.G., Robertson, P.K. and Gillespie, D. (1986). Seismic cone penetration tests. Use ofln- 

Situ Tests in Geotechnical Engineering (GSP 6). ASCE, Reston, VA, 116-130. 

Ishihara, K. (1985). Stability of natural deposits during earthquakes. Proceedings, 11 th International 

Conference on Soil Mechanics & Foundation Engineering, Vol. 1, Balkema, Rotterdam, 321-376,

Jardine, R.J., Potts, D.M., StJohn, H.D., and Hight, D.W. (1991). Some practical applications of a 

nonlinear ground model. Proceedings, 10th European Conference on Soil Mechanics and Foundation 

Engineering (1), Firenze, 223-228. 

Johnston, A.C. and Schweig, E.S. (1996). The enigma of the New Madrid earthquakes of 1811 - 1812. 

Annual Review of Earth and Planetary Sciences 24, 339-384. 

Juang, C.H., Jiang, T. and Andrus, R.D. (2002). Assessing Probability-based Methods for Liquefaction 

Potential Evaluation. Vol. Journal of Geotechnical Engineering 128:7, 580-589. 

Lunne, T., Robertson, P.K., and Powell, JJ.M. (1997). Cone Penetration Testing in Geotechnical 

Practice. Blackie Academic, Routledge Publishers, New York, 312. 

Mayne (2001). Stress-strain-strength-flow parameters from enhanced in-situ tests. Proceedings, Intl. 

Conf. on In-Situ Measurement of Soil Properties & Case Histories, Bali, Indonesia, 27-48. 

Robertson, P.K. (1990). Soil classification using the cone penetration test. Canadian Geotechnical 

Journal 27:1, 151-158. 

Robertson, P.K. and Wride, C.E. (1998). Evaluating cyclic liquefaction potential using the cone 

penetration test. Canadian Geotechnical Journal 35:3, 442-459. 

Seed, H.B., and Idriss, I.M. (1971). Simplified procedure for evaluating soil liquefaction potential, 

Journal of Soil Mechanics and Foundation Division 97:9, 1249-1273. 

Stewart, W.P. and Campanella, R.G. (1993). Practical aspects of in situ measurements of material 

damping with the seismic cone penetration test. Canadian Geotechnical Journal 30:2, 211-219. 

Stokoe, K.H., Anderson, D.G., Hoar, RJ. and Isenhower, W.M. (1978). In-situ and lab shear velocity 

and modulus. Earthquake Engineering and Soil Dynamics, Vol. Ill, ASCE, New York, 1498-1502. 

Vucetic, M. and Dobry, R. (1991). Effect of soil plasticity on cyclic response. Journal of Geotechnical 

Engineering 117:1, 89-107. 

Woods, R.D. (1978). Measurement of Dynamic Soil Properties, State of the Art Report. Earthquake 

Engineering and Soil Dynamics, Vol. I, ASCE, Reston, VA, 91-178. 

Yasuda, S. and Tohno, I. (1988). Sites of re-liquefaction caused by the 1983 Nihonkai-Chuba 

earthquake. Soils and Foundations 28:2, 61-72. 

Youd, T.L. (1988). Recurrence of liquefaction at the same site. Proceedings of the Eighth World 

Conference on Earthquake Engineering. San Francisco, Vol. Ill: 231-238. 

Youd, T.L. and Gams, C.T. (1995). Liquefaction-induced ground-surface disruption. Journal of 

Geotechnical Engineering, 121:11, 805-809. 

Youd, T. L., et al. (2001). Liquefaction resistance of soils: Summary report from the 1996 NCEER and 

1998 NCEER/NSF workshops on evaluation of liquefaction resistance of soils. Journal of 

Geotechnical and Geoenvironmental Engineering, 127:10, 817-833. 

122




ATTENUATION FUNCTION OF GROUND MOTIONS FOR 

GUANGDONG REGION OF SOUTHERN CHINA 

WONG Yuk Lung 1 , ZHENG Sihu 2 , LIUJie 2 , KANG Ying 3 , TAM Cheuk Ming 4 , 

LEUNG Yin Kong 4 , ZHAO Xinquan 5 

1 Department of Civil and Structural Engineering, The Hong Kong Polytechnic University, Hong Kong 

2 Center for Analysis and Prediction, China Seismological Bureau, Beijing, 100036 

3 Seismological Bureau of Guangdong Province, Guangzhou, 510070 

4 Hong Kong Observatory, Hong Kong 

5 Institute of Geological and Nuclear Science, Gracefield, Lower Hutt, New Zealand 

ABSTRACT 

We made use of the first ever set of horizontal-component digital seismic data of 44 earthquake 

events (M L = 2.5 to 5.1) from September 1999 to October 2000, recorded by 14 stations of the 

Guangdong Telemetered Network to determine the attenuation function of S-wave phases of ground 

motions for the Guangdong Region of southern China. 

It was found that the geometrical spreading function of the region could be best represented by a trilinear 

piecewise continuous function. At a distance less than 45 km from a seismic source, the 

geometrical spreading coefficient was 0.97. Between 45 and 100 km, the coefficient was nearly zero. 

Beyond 100 km, the coefficient was 0.5. The frequency-dependent Q was estimated to be Q = 

481.5f°' 31 . 

We then evaluated the source spectra for each of the 44 events from the records of the YHK Station of 

the Hong Kong Observatory's seismic monitoring network, and the proposed attenuation model. It 

was found that the so-obtained source spectra derived from YHK Station were comparable to those 

derived from the stations of Guangdong Network. Hence, the proposed attenuation function was 

likely to be applicable for the Hong Kong region as well. 

INTRODUCTION 

In seismology, the analysis of attenuation characteristic of ground motions provides a better 

understanding on wave transmission inside the earth, and the seismic source parameters. In 

engineering applications, the attenuation relationships of S-wave phases of ground motions are of 

most interest as their amplitudes are approximately five times larger than the associated P-wave 

phases, causing most of the earthquake damages. Therefore, a quantification of the S-wave 

attenuation is necessary for seismic hazard analysis of a region. 

Traditionally, researchers tended to study the attenuation and hazard issues for seismically active 

regions. After the Newcastle Earthquake in 1989 which occurred in a region of low seisrnicity and 

caused significant damages valued at about 700 million Australian dollars (Brunsdon, 1990), 

earthquake hazard analysis of a region of low seismic risk but high consequence of economic losses

124 

has received more attention. In recent years, as there has been a tremendous economic growth in 

Guangdong, a province of low to moderate seismicity, the need to proper assess the attenuation of 

ground motions and seismic risk of this region is apparent. 

It has been recognized that the attenuation of wave transmission in a region can be reliably estimated 

from calculations based on extensive ground motion data recorded in the region. As the quantity of 

useable strong ground motion data in China is small, even for high seismic regions, the attenuation 

characteristics have been formulated using the intensity records in China. However, there is a concern 

on the accuracy of such approach. 

In 1999,the Guangdong Digital Seismograph Network was established in the Guangdong Province of 

China, and recording of digital ground motion data from 14 telemetered stations of the Network 

commenced in August of the same year In collaboration with Guangdong Seismological Bureau, the 

first ever-complete set of digital ground motion data of the region was used to evaluate S-wave 

attenuation Results of the study, including the geometrical spreading coefficients, the anelastic 

attenuation and site responses of the Guangdong region, and the applicability of the proposed 

attenuation function for the Hong Kong region are reported in this paper. 

SEISMOGRAM DATA 

The dataset consists of 249 selected horizontal-component digital seismograms from 44 earthquake 

events (M L = 2.5 to 5.1) between September 1999 to October 2000, recorded by the Guangdong 

Telemetered Network. Figure 1 shows the locations of the stations and the events. The distances 

between the earthquake hypocenters and the observation stations vary from 10 km to 500 km. 

Figure 1. Distribution of stations and seismic events (open triangles indicate location of 

stations and solid circles indicate locations of seismic events) 

FOURIER SPECTRUM OF GROUND MOTION 

The observed shear wave spectral amplitude of event f at station j can be written as follows 

where A |0 is the source term of event i, G is geometric attenuation function, R t is hypocentral 

distance, 5 y Is site response, and Q is frequency-dependent Q, and j3 is shear wave velocity. 

(i)

125 

Taking logarithm on two sides, we get the following equation 

where c(/) is the coefficient of anelastic attenuation, which is 

/) (2) 

GEOMETRICAL SPREADING FUNCTION G(R U ) 

For the geometncal attenuation function, early studies suggested that at near source distances, the 

geometric spreading coefficient was 1, and at regional distances where the dominant phase was Lg 

wave, the geometric spreading coefficient was 0.5. However, recent studies (Burger et. al, 1987, Ou 

and Herrmann, 1990) indicated that the shape of geometric attenuation function was complex even for 

a simple layered crustal model. Within a medium source distance, the amplitudes might increase or 

remain constant. In this study, we tried linear, bilinear and trilinear pieciewise continuous functions 

to fit the ground motion data, and found that the trilinear model (Equation 4) yielded the best results 

or the least residual error (as defined below). Here we just outlined the procedure for determining the 

geometrical spreading coefficients of the trilinear model which was based on the method proposed by 

Atkinson and Mereu (1992). 

~ b ' - Rf" - R~ bl R }

126 

and log A tQ (/) is the mean source amplitude for event £, which is defined: 

where n, is the number of stations recording the event i, n 0 is the total number of events used in the 

analysis. 

The residual Res is a function of (b { , b 2 , b-. and c), the frequency /, and sources distances (R\, R 2 ). 

In order to reduce the number of unknowns in the inversion process, we took b 3 equal to 0.5, judging 

from the previous studies on Lg waves. 

For a specific frequency / , the site response of all stations was initially taken to be zero, and some 

trial values of /, and R 2 were selected. We used Genetic Algorithm (GA) to search for the 

combination of attenuation parameters (6,, b 2 , c) that gave minimum residual error. The increment 

of the site response correction Aflog 5 ; (/)] for station; was calculated from Equation 9. 

(8) 

where m } is event numbers recorded at station;. 

(9) 

The above-mentioned process was repeated, using these new site terms. Iteration continued until no 

further reduction in the residual error was possible. We then changed the values of R { and R 2 , and 

repeated the whole process again and again till a pair of ( R { and R 2 ) yielded the least residual error, 

so that the best set of attenuation parameters and site response were finally obtained. 

Coefficients of Geometrical Spreading Function 

Based on the aforesaid process, we derived the best pair of R { =45km and R = 2 lOOfcm. The 

associated geometrical attenuation coefficients b { , b 2 and site terms S y (/) for each sampling 

frequency were also determined. By definition, the geometric attenuation coefficients b } and b 2 

should be independent of frequency. However, as the inversion process was conducted for each 

sampling frequency, we obtained one set of (b { , &, , c ) for each sampling frequency. Figure 2 shows 

the values of b } and b 2 . We took the average value of each parameter, and obtained: b { = 0.97 ±0.1 1, 

b 2 = 0.0097 ±0.15, and b 3 = 0.5 as our final solutions. 

In order to study the attenuation due to the path effects only, we defined the so-called normalized 

amplitude LogA'//) (Equation 10), by subtracting the source term and the site term of Equation 2. 

Figure 3 compares the observed decay of normalized amplitude with that obtained from the proposed 

attenuation function for/= 8.91 Hz, as a typical example. It is evident that there is a significant 

flattening in the decay of spectral amplitudes in the transition zone of the source distances ranging 

from 45 km to 100km.

127 

I 

'++ 

-j- 

4. 

+4•N-. 

k 

. 

j_ 

.."*" + 

. + 

4 

~ib° 10' 

Frequency (Hz) 

.1 n- 

"16° 10' 


Figure 2. Values of b { and b-> vs frequency / (thick line indicates average value) 

1 10-4 

Q) 

TJ 

10* 

o 

2 

10 2 10 3 

Hypocentrai Distance (km) 

Figure 3. Observed decay of normalized amplitude (for/ = 8.91/fz). 

ANELASTIC ATTENUATION 

Figure 4 plots the coefficients of anelastic attenuation, c(f), as well as 2(/) • For the frequency 

below 2 Hz, the calculated Q values are quite scattered. Therefore, we only regressed the results 

above 2 Hz, and obtained the frequency-dependent Q as Q = 481.5 • /°' 31 .

128 

O 10 s 



Figure 4 Coefficient of anelastic attenuation c and Q avs frequency 

(straight line is Q = 481.5 -/ a31 ) 

Figure 5 compares our Q model for S waves in the Guangdong region with the Q models for S or Lg 

waves in other regions [Mark "1" Q = 130/ 10 for Apennines of Italy (Malagnini et al., 2000a); Mark 

"2" Q = 150/ 5 for West America (Chin and Aki, 1991); Mark "3" Q = 56/ 01 for Oaxaca of Mexico 

(Castro and Munguia, 1993); Mark "4" Q = 4QO/ 42 

for Central Europe (Malagnini et al., 2000b); 

Mark "5" Q = 670/ 33 for Eastern Canada (Atkinson and Mereu, 1992); Mark "6" Q = 508/ 48 for 

Indian shield region (Singh et al., 1999); and Mark "7" Q = 680/ 36 s for Eastern North America 

(Atkinson and Boore, 1995)]. We found that the Q model for shear waves in the Guangdong region 

is similar to that in Central Europe. 

10 2 10° 


Figure 5 Comparison of 0 models of 5 waves of different regions (P - Guangdong region, 1 • 

Apennines, 2 - west America, 3 - Oaxaca of Mexico, 4 - Central Europe, 5 - Eastern Canada, 6 - 

Indian shield region, 7 - Eastern North America)

129 

CORRELATION WITH GROUND MOTION DATA RECORDED BY THE HONG KONG 

OBSERVATORY 

For geological considerations, Hong Kong is part of the Guangdong region. In the derivation of the 

above-mentioned model and parameters, we have not used the ground motion data recorded in Hong 

Kong. In fact, the Hong Kong Observatory operates a network of eight stations to record-ground 

motions in digital format since 1997. 

In this section, we tried to establish the source spectra of the same 44 events from the digital signals 

recorded by YHK station of the Hong Kong Network, assuming that there was no site amplification 

effect of YHK station as it was sited on hard bedrock, and using our proposed geometric spreading 

function and Q function derived from the Guangdong Digital Seismograph Network data. 

It was found that source spectra of the selected events derived from the 14 Guangdong stations and 

from YHK station were comparable, and the discrepancies were well within an acceptable range. 

Figure 6 is an example of comparison of the source spectra of the Heyuan Earthquake (M s = 3.1, 

occurred on 4 November 1999) calculated from 5 Guangdong stations and the YHK station. The 

thick line is derived form YHK station. It also confirms insignificant site amplification effect at YHK 

station. 

Figure 6 Comparison of source displacement spectra recorded at different stations for Heyuan 

Earthquake of 4 November 1999 (M s = 3.1) (thick line is from records of YHK Station after corrected 

for attenuation) 

f/Hz 

SUMMARY OF FINDINGS 

Based on the 249 Fourier spectra of 44 earthquake events recorded by the 14 stations of the 

Guangdong Telemetered Network, the attenuation model and source parameters were estimated using 

the Genetic Algorithms inversion. 

The results can be summarized as following: 

* The geometrical spreading function shows three distinct sections for the Guangdong region. At 

a source distance less than 45 km, corresponding to geometrical spreading of the direct wave, 

the geometrical spreading coefficient is 0.97. At a source distance between 45 and 100 km, the

130 

geometrical coefficient is nearly zero. At a source distance beyond 100 km, corresponding to 

the Lg phase, the geometrical spreading coefficient is 0.5. 

The frequency-dependent Q is estimated as Q — 481.5 • /° 31 . 

Preliminary study of the digital seismograph data of the same 44 events recorded by Hong Kong 

Observatory seismic monitoring network-indicates that the proposed attenuation model is likely 

to be applicable for Hong Kong Region 

Acknowledgments 

The work reported in this paper is a part of the ASD Project on ground motion and response spectra 

for seismic design in Hong Kong funded by The Hong Kong Polytechnic University. 

References 

Atkinson, G. M. and D. Boore (1995). New ground motion relations for eastern North America, Bull 

Seism. Soc. Am., 85, 17-30. 

Atkinson, G. M. and R. F. Mereu (1992). The shape of ground motion attenuation curves in 

Southeastern Canada, Bull. Seism. Soc. Am. 82, 2014-2031. 

Brunsdon, D.R. (1990). The December 28, 1989 Newcastle , Australia Earthquake. Bulletin of the 

New Zealand National Society for Earthquake Engineering, Vol. 23, No. 2, 102-120. 

Burger, R., P. Somerville, J. Barker, R. Herrmann, and D. Heimberger (1987). The effect of crustal 

structure on strong ground motion attenuation relations in eastern North America, Bull. Seism. 

Soc. Am 77, 420-430. 

Castro, R R. and L. Munguia (1993). Attenuation of P and S waves in oaxaca, Mexico subduction 

zone, Phvs Earth Planet. Interiors, 76, 179-187. 

Chun, K., G. West, R Kokoski, and C. Samson (1987). A novel technique for measuring Lg 

attenuation: result from eastern Canada between 1 to 10 Hz, Bull Seism. Soc. Am. 77, 389-419. 

Hasegawa, H. (1983). Lg spectra of local earthquakes recorded by the Eastern Canada Telemetered 

Network and spectral scaling, Bull. Seism. Soc. Am. 73, 1041-1061. 

Malagnmi, L., R. B. Herrmann, and M. D. Bona (2000a). Ground-motion scaling in the Apennines 

(Italy), Bull Seism. Soc. Am., 90, 1062-1081. 

Malagnini, L., R. B. Hermann, and K. Koch (2000b). Regional ground-motion scaling in central 

Europe, Bull Seism. Soc. Am., 90, 1052-1061. 

Ou, G. and R. Herrmann (1990). A statistical model for peak ground motion from local to regional 

distances, Bull. Seism. Soc. Am. 80, 1397-1417. 

Shin, T. and R. Herrmann (1987). Lg attenuation and source studies using 1982 Miramichi data, Bull. 

Seism. Soc. Am. 77, 384-397. 

Singh, S. K., M. Ordaz, R. S. Dattarayam, and H. K. Guputa (1997). A Spectral analysis of 21 May 

1997, Jabalpur, India, Earthquake (M W =5.S) and estimation of ground motion from future 

earthquakes m the Indian shield region, Bull. Seism. Soc. Am., 89, 1620-1630.




SLOPE STABILITY AGAINST EARTHQUAKE 

CONSIDERING THREE DIMENSIONAL EFFECT 

I. Yoshida 1 , T. Kaneto 2 and M. Takao 2 

Structural and Geotechnical Engineering Department, Tokyo Electric Power Services Co.,Ltd., 

Tokyo, JAPAN 

2 Nuclear Power Engineering Department, Tokyo Electric Power Company 

Tokyo, JAPAN 

ABSTRACT 

In order to discuss slope stability against seismic force considering three dimensional (referred as 3-d, 

hereafter) effect, an evaluation method based 3-d dynamic FEM is developed. The model for the 3-d 

dynamic FEM used in this study is composed of one 3-d model, four 2-d far field models and four 1-d 

far field models. They are connected by viscous boundaries. In the proposed method, a sliding surface 

is modeled by a set of planes or an oval. The problem to search the sliding surface with minimum 

safety margin is formulated as optimization problem, in which objective function is safety factor, and 

design parameters are parameters describing the sliding surface such as center and radius of the sliding 

arc. The optimization problem is solved by using Genetic Algorithm, which attracts attention and is 

widely used as practical global optimization method in various engineering problems recently. In 3-d 

analysis, directions of excitation and sliding are also important, so that these directions for minimum 

safety factor are also searched simultaneously. Several numerical simulation results show that safety 

factor based on 2-d analysis tends to be conservative, though the response of 3-d model is larger than 

that of 2-d model partially. 

INTRODUCTION 

After Kobe earthquake (1995), many Japanese aseismic design codes have been revised, and 

design earthquake and spectra tends to become large. For the large design earthquake, we need to 

estimate the limit state more carefully than before. If the limit state is examined under lots of 

conservative assumptions, the obtained safety margin might be smaller than the level required in the 

new design code. Consequently a lot of budget for the strengthening is needed irrationally. One of the 

standpoints for more precise estimation is to consider three dimensional (referred as 3-d, hereafter) 

effect of the slope. For the stability analysis of slopes, there are several methods, such as limit 

equilibrium method and finite element method (FEM). However most of the method that have been 

performed so far are based on two dimensional shape of slope and sliding surface. 

In this paper, an evaluation method based on 3-d dynamic FEM is studied. Search of critical sliding 

surface can be formulated as optimization problem hi which the objective function is the safety factor 

of the sliding surface and design parameters are parameters which define the sliding surface. Genetic 

Algorithm (referred as GA here) is used for the optimization. GA is one of the most prospective

132 

optimizing method in many engineering problems, which is learned from the way of evolution of life, 

especially mechanism of genes. 

METHODOLOGY FOR ESTIMATION OF SAFETY FACTOR 

Response Analysis and Safety Factor 

Dynamic response of a slopes is calculated in time domain by 

Newmark p method which is one of direct integral methods, in this 

study. Viscous boundaries can be adopted for the base and the four 

sides of a FEM model As shown in Fig.l, the 3-d model is 

connected with four 2-d free field models by viscous boundaries. 

The each 2-d model is also connected with two 1-d free field 

models by viscous boundaries. The each model has also viscous 

boundary at the base. 

In conventional 2-d analysis, sliding surface is usually modeled 

by a set of lines or an arc. As a natural extension to 3-d analysis, 

a sliding surface is modeled by a set of planes or an oval The 

modeled sliding surface is divided into many triangle panels for 

the numerical calculations. The vertex of the panel is 

determined from the intersection of side of FEM element and 

the sliding surface. Fig.2 shows two triangle panels, cl-c3-c4 

and cl-c2-c3, in an hexahedron element. 

Safety factor of the modeled sliding surface is obtained as 

the ratio of resistance force to sliding force. The resistance 

force working on a triangle panel is calculated by normal stress 

to the panel, area of the panel and material strength (cohesion 

and internal friction). Sliding force on the panel is also 

calculated by shear stress to sliding direction and the area of the 

panel Safety factor S/ is obtained as the ratio of total sum of 

each force as follows. 

Fig.l Free Fields and 

Boundary Conditions 

Fig.2 Triangle Panels Forming 

Sliding Surface 

where, RI : resistance on panel iÂ\: area of panel f, t: time,0 p 0 2 : directions of excitation and sliding 

S l : sliding force on panel z, n : number of the panels, x : a vector of parameter which define a sliding 

surface. Minimum value with respect to these parameters is used as a safety factor of a slope or ground. 

Optimization by Real Coded Genetic Algorithm 

The problem to search the sliding surface with minimum safety margin is formulated as optimization 

problem, in which the objective function is safety factor, and design parameters are parameters 

providing the sliding surface such as center, radius of the surface, and so on. In this optimization 

problems, there might be many local optimum solutions. In that case, local search method such as

133 

1st Generation 

dGenerate individuals randomly ! 

generations 

[ Evaluation of fitness ) 

I Preserve the elite solutions j 

Selection 

O 

D 

parent 

child 

.0 a 

..-Parent-] 

/ 

0 

•Parent .2 

F 

Parent-3 

\. 

0" 

gQ-ârent _2 

a-' 

%r'DD 

0-Sn 

.©"' 

..•Parent- ! 

Parent-3 

Q 

/ ••. 

f*B\ 

/%£ )D 

••' Q 

•'* n 

(£""" 

Parent- 1 

••*"'" 

Type-1 Type-2 Type-3 

Fzg. 4 Outline of Crossover in Real Coded GA 

..-••"" Parent- 

Crossover and Mutation J modified Newton methods (DFP, BFGS, etc.) can not be 

applicable. In this study, we adopt GA (Genetic algorithms) 

Optimum solution 

which attracts attention as practical global optimization 

methods. 

Fig.3 Flowchart of Genetic Algorithm 

Traditionally, binary codes have been used to represent 

the design variables including real number parameters, and 

the crossover operation is performed by the exchange of a 

part of the binary bit. However, some researchers are using 

the GA with real coded crossover operators (Tsutsui et al,1997) and reported the advantage of their 

methods. Among them, the unimodal normal distribution crossover (UNDX) is reported to show good 

performance in optimizing problems (Kita, et al, 1999). 

Basically, the outline of real coded GA which is shown in Fig.3 is same as that of binary coded GA, 

except crossover operator and mutation. There are several models proposed for real coded crossover 

operator with Gaussian noise. The concepts of the crossover operator which are used in this paper are 

illustrated in Fig.4 (Yoshida, 2000). In the case of Type-1, the children are derived by following 

equation. 

g_ A+P 2 Pi-P* here, a * N(0,l) (2) 

V 

2 2 

c, p l and p 2 are vectors in design parameter space, c is a child made from parents p l and p 2 . The 

parents are randomly selected in proportion to their fitness, which is the inverse of safety factor in this 

slope stability study. The children are generated in one-dimensional subspace of the parameter space in 

Type-1. In Type-2, n dimensional perturbation is given to Type-1 children, consequently Type-2 

children are generated in n dimensional space, when there are n design variables. The magnitude of the 

perturbation is defined based on the distance from Parent-3 to the line with Parent-1 and 2. In Type-3, 

the children are generated in 2-dimensional subspace made by three parents as shown in Fig.4. In all 

types, the generated children are in the neighbourhood of the gravity center of their parents. 

Perturbation v is added to each design variable as mutation. In this study, Cauchy distribution is 

used. 

NUMERICAL EXAMPLES 

Models and Characteristics of Dynamic Response 

Two types of slope model are used for the numerical calculations. FEM model of them are shown 

in Fig.5. The height and width of the slope are indicated in Fig.6. The number of node and element of 

Type-1 model is 8848 and 7557, while those of Type-2 are 11336 and 9883. The sides and base of the

134 

Type-2 

Fig.5 Three dimensional Slope Model for the Numerical Calculation 

Z 

A 

C 800 ..£400 . c 800 x, 

r* *n^ "^ *n 

/ V / 

270 

lower layer 

Fig. 6 Dimension of the Slope Model 

' 

540 

810 

unit: m 

-01 00 01 (m/s) 

Fig. 7 Response Acceleration in y-Direction 

(perpendicular to Excitation x-Direction) 

model are connected with free field by viscous boundaries. Several cases with respect to stiffness and 

strength properties are studied. 

As an example of response analysis, the response acceleration in y direction at specific time with 

homogeneous property model (shear velocity 340m/sec., Poisson's ratio 0.48, damping ratio 0.0) is 

shown in Fig.7. Since the direction of excitation is in ^-direction, the response in y-direction 

corresponds to the component perpendicular to the excitation, which means response due to 3-d effect. 

The spectral ratios to the doubled input motion in ^-direction are shown in Fig.8. The response at point 

A becomes large, while that of point B becomes small compared with 1-d response due to the slope 

configuration. The response of Type-1 is larger than that of Type-2 generally. 

Sliding Surface With Minimum Safety Factor 

The sliding surface with minimum safety factor is searched by GA, for both models Type-1 and 

Type-2. It is assumed that the models are divided into two layers, upper and lower layer as shown in 

Fig.6. The properties of both layers are summarized in Table 1. The time history and Fourier spectrum 

of the input motion, which is the record of Kera county earthquake at Taft, are shown in Fig.9. The 

sliding surface is defined by crossing surface of the oval and the 3-d slopes. The parameters of the oval 

are shown in Fig. 10. The coordinate of center (x, v, z), radius (a, b, c), angle around z axis, of the oval 

are determined such that the safety factor becomes minimum by GA. The directions of sliding 9 l and 

excitation #, for minimum safety factor are also searched in addition to the above parameters at the 

same time. Population size and generation size for GA are is 150 and 40. 

The obtained sliding surfaces with minimum safety factor of Type-1 are shown in Fig. 11 and 12. 

Fig. 11 shows the critical surface when the search is performed under the limitation that the sliding

135 

10° 10' 1 


10° 10 1 Fig.9 Acceleration Time History of 


/ n^wr Motion 

c 

Fig. 8 Spectral Ratio to Doubled Input Motion 

Fig. 10 Parameters describing the shape 

of sliding Surface 

surface is generated only inside the model, which means size of sliding block is limited by some 

reasons in an actual slope. While Fig. 12 shows the critical surface when no limitation is considered. 

The safety factor with the limitation is 2.38 and the safety factor without the limitation is 2.20. For the 

comparison, 2-d analysis is also performed, and the obtained safety factor is 2.14, In the case without 

the limitation, the shape of sliding surface is close to the shape obtained by 2-d analysis and the 

directions of excitation and sliding are close to the direction of the steepest of the slope. This result 

suggests us that the critical state without the limitation is close to the state of two dimensional analysis. 

On the other hand, Fig. 11 suggests that the shape of sliding surface, directions of excitation and sliding 

becomes complicated, if there is some limitations with respect to the area of sliding surface. 

The same analysis for Type-2 model with the limitation is performed. The critical surface is shown 

in Fig. 13. The direction of sliding and excitation are close to the steepest direction of the slope 

although the limitation is considered. It can be interpreted that the response around the 3-d comer of 

the slope of Type-2 is smaller than that of 2-d slope so that the critical slope tends to be formed to 

avoid 3-d corner of the slope. The obtained safety factor for Type-2 slope is 2.60, which is larger than 

that of Type-1 model. 

Table 1 Properties of Model 

Upper 

Lower 

Unit Weight 

(kN/m 3 ) 

17.8 

16.8 

Poisson's 

Ratio 

0.48 

0.45 

Shear 

Velocity 

(m/s) 

338.0 

499.0 

Cohesion 

(kPa) 

98.1 

1500.0 

Internal Friction 

(degree) 

38.6 

0.0 

Tensile Strength 

(kPa) 

0.0 

222.0

136 

plane 

plane 

sliding 

x-z section 

y-z section 

excitation 

excitation 

Fig. 11 Critical Sliding Surface with Limitation 

of the Area for Type-1 Slope Model 

CONCUSION 

sliding 

Fig. 12 Critical Sliding Surface without 

Limitation of the Area for Type-1 Slope Model 

In this paper, an evaluation method of safety 

factor of sliding surface based on 3-d dynamic 

FEM is studied and numerical examples are 

shown. Within the simulations performed in this 

study, the safety factors obtained by 3-d model is 

larger than those by 2-d model Though it is 

difficult to conclude that the estimation by 2-d 

model is always conservative, our results suggest 

that actual 3-d slope has more safety margin than 

that estimated by 2-d model in many cases. 

y-z section 

plane 

excitation 

sliding 

References 

Fig. 13 Critical Sliding Surface with Limitation 

of the Area for Type-2 Slope Model 

KITA H., ONO L and KOBAYASHI S. (1999). 

Theoretical analysis of the unimodal normal 

distribution crossover for real-coded genetic algorithms (in Japanese), Transactions of the society of 

instrument and control engineers, vol.35-11, pp. 1333-1339 

TSUTSUI S. and GHOSH A., CORNE D. and FUJIMOTO Y. (1997). A real coded genetic algorithm 

with an explore and an exploiter populations", Proceedings of the 7th International Conference on 

Genetic Algorithms, pp.238-345 

YOSHIDA I. (2000), Comparison of Real Coded GA(Genetic Algorithm) and Binary Coded GA 7 

Proceedings of International Conference on Monte Carlo Simulation, pp. 81-87

SEISMIC RISK AND 

DISASTER MANAGEMENT




< on 

THE MID-AMERICA EARTHQUAKE CENTER RESEARCH 

PROGRAM TOWARDS DEVELOPMENT OF CONSEQUENCE- 

BASED SEISMIC RISK MITIGATION 

D.Abrams 1 , A.Elnashai 2 and J.Beavers 3 

1 MAE Center, University of Illinois at Urbana-Champaign 

Urbana, IL, USA 


Urbana, IL, USA 


Urbana, EL, USA 

ABSTRACT 

A new framework for managing seismic risk, termed Consequence-Based Engineering (CBE), has been 

adopted and articulated by the Mid-America Earthquake (MAE) Center. This entails undertaking rapid 

assessment, comparing expected and acceptable consequences, re-assessing hazard, vulnerability and 

inventory, investigating various possible vulnerability reduction measures and decision making regarding 

benefit and cost of the various options. This process is not limited to structural systems, but rather to all 

societal endeavors, including economic and social ramifications. Moreover, the framework, or paradigm, 

of CBE is not limited to seismic nsk, but pertains to other forms of exposure and indeed to multiple 

hazard scenarios. To elaborate this framework and to provide a comprehensive set of application 

guidelines, the MAE Center has formulated a core research program comprising a number of fully 

coordinated projects aimed at covering all aspects of seismic risk mitigation measures viewed from a 

consequence-based perspective. The research program is described in this paper. The individual projects 

are described and their role towards development of CBE is discussed. The intricate process of meshing 

the various research projects, items of input and output and their inter-relationship is described. Finally, 

the anticipated outcome of the MAE research program, namely (i) a complete and state-of-the-art seismic 

risk mitigation environment, (ii) technical developments amenable to exploitation by industry and other 

stakeholders and (iii) the advanced IT visualization system through which the technical developments are 

projected to aid decision- and policy-makers, are presented.

140 

CONSEQUENCE-BASED ENGINEERING - DEFINITION AND APPLICATION 

Consequence-Based Engineering (CBE) is a new paradigm for seismic risk reduction across regions. It 

quantifies the risk to societal systems and subsystems by working with policy-makers, decision-makers 

and stakeholders to ultimately develop risk reduction strategies and implement mitigation actions. 

Therefore, it is a framework for risk assessment in all its aspects. With a coordinated approach to 

engineering on the basis of anticipated consequences across regions rather than individual structures, a 

user of CBE can assess probable seismic hazards, synthesize likely damage across regions of specific 

stakeholder interests and minimize consequences of such an event through selected interventions. 

Because social-economic impact is considered across a population of construction, the benefits of seismic 

risk-reduction measures can be better assessed through this new systems approach. The paradigm provides 

opportunities to perform the following tasks in vulnerable communities: 

• assessment of risks and understanding of the most critical components of the region 

• provision of intervention scenarios and their cost-benefit equation 

• setting priorities in intervention and disaster management policies 

• re-assessment of system vulnerability after implementation of mitigation measures 

• determination of research priorities to minimize uncertainty in the above applications 

The new paradigm is intended for use by practicing engineers responding to the needs of their clients who 

have a significant vested interest in mitigating seismic risk and are thus termed "stakeholders." 

Stakeholders are groups or individuals making decisions to invest in a particular intervention to mitigate 

possible earthquake losses. Examples of stakeholders relying on the results of a CBE analysis include 

insurance executives, city managers, state highway officials and owners of large building stocks. Using 

new technologies that support CBE, practitioners will be able to demonstrate to their stakeholder clients 

via high-end data-mining and visualization tools, what types of consequences are likely for their 

respective systems and how various intervention measures, such as retrofit or relocation of system 

ingredients, can reduce such consequences. 

As developed by the MAE Center the steps of a CBE analysis are given in the flowchart of Figure 1. 

Initially, a rapid assessment is executed to define the relevant system, approximate the probable hazard, 

project what consequences are likely and delineate what types of consequences might be acceptable. Next, 

a four-step decision tree is used to determine if: (a) estimated consequences are acceptable, (b) acceptable 

consequences should be redefined, (c) modeling parameters should be refined and (d) further system 

interventions should be considered. If anticipated consequences exceed tolerable ones and no further 

refinement of acceptability is feasible, then parameters defining the hazard and built environment can be 

refined to reduce anticipated losses (assuming that the preliminary analysis were conservative) and/or 

system intervention may be prescribed to minimize anticipated losses. An interactive damage synthesis 

module developed with advanced data-mining and visualization tools is used to determine and view 

consequences for various problem definitions and mitigation scenarios. Accessing this module iteratively, 

consequences may be visualized for a number of different system intervention strategies with various 

input parameters describing the hazard or the built environment.

141 

1. System Definition 

define system of interest 

define hazard 

define characteristics 

Are Consequences 

Acceptable'' 

Done 

confident that 

consequences will be 

acceptable 

2. Rapid Estimate of 

Consequences 

quick assessment of 

likely consequences 

Should Acceptable 

Consequences be 

Redefined'' 

• 4. Decision Making 

3. Define Acceptable 

Consequences 

define stakeholder needs 

Rapid Assessment' 

Should 

Parameters 

be Refined 

Should System 

Interventions; be 

Considered" 7 

Refine Hazard Estimate 

further refine hazard for more 

pfecise loss assessment 

Refine Inventory Estimate 

further refine inventory of built 

environment for more precise loss 

Assessment 

10. Prescribe System 

Interventions to Minimize 

Consequences 

rehabilitate or demolish 

vulnerable structures, construct 

new structures, re-route network 

flows, re-manage land use, etc 

Consequence Minimization ' 

7. Refine Fragility Relations 

further refine vulnerability of built 

environment with refined hazard 

and more precise response 

analyses 

8. Re-Assess Social Impact 

assess social and economic 

consequences of event in terms of 

refined hazard and inventory 

estimates and fragility relations 

Damage Synthesis 

9. Re-Visualize Consequences 

examine effects of system 

alterations on reducing 

consequences 

FIGURE 1: OPERATIONAL STEPS OF CBE APPLICATION

142 

RESEARCH AND DEVELOPMENT REQUIREMENTS OF CBE APPLICATION 

CBE may be used with existing technologies. Indeed, it is in current use, though unintentionally and 

perhaps in a less structured manner than that depicted in Figure 1. However, its true appeal is in the fact 

that its requirements set research and development priorities that, in turn, bolster its application and 

increases confidence in decisions based on the outcome of its application. The MAE Center Core 

Research Program (CRP) has been formulated to respond to the challenges of developing the next 

generation of risk assessment environments, based on the consequence-based engineering approach. The 

hierarchical structure of the CRP is shown in 

Figure 2. 

Damage 

Synthesis 

CBE 

Framework 

MAE Center Core 

Research Program 

Consequence 

Minimization 

Hazard 

Definition 

FIGURE 2: MAE CENTER CORE RESEARCH PROGRAM 

The structure of the CRP closely follows requirements set by CBE as well as research management, interdisciplinary 

teaming of researchers and industrial interest in research products. It comprises three thrust 

areas: Hazard Definition (HD), Damage Synthesis (DS) and Consequence Minimization (CM). The CBE 

Framework Development (FD) is an overarching activity that guides the projects, develops further the 

CBE approach and manages uncertainty throughout the research programs. Therefore, the CBE FD is a 

capstone for the three thrust areas. Each thrust area, as well as the capstone FD, comprise a number of 

projects that are designed to fulfill a specific requirement in the CBE paradigm development and 

application. Broadly, the DS thrust provides damage estimates, both rapid and refined, before an event. 

The CM thrust area provides measures of minimizing the damage estimated in DS as well as means of 

selecting from a number of possible interventions. The HD thrust provides seismic input (hazard) for all 

projects in all required forms. 

The projects under each thrust area and the capstone alongside their inter-relationship are shown in Figure 

3. In each thrust area there is a gateway project that acts as the outlet from the thrust area to the rest of the 

research program. The gateway projects of the three thrusts are: Damage Visualization Module (DS-1), 

Decision Support Tools (CM-1) and Synthetic Earthquake Hazard (HD-1).

143 

f«l-.J 

WJjfc >. « 

Relations 

-Ji 4 * 

EQ Path 

Modeling 

* " a """f'* 

Modeling 

Hazards Definition Thrust Area 

JT* " Defortr 

key [ |j systems integration [ henabling technologies \...\..~ If fundamental knowledge 

FIGURE 3: PROJECTS OF THE MAE CENTER PROGRAM GROUPED BY THRUST AREA 

MAE CENTER RESEARCH PROJECTS IN A CBE CONTEXT 

Hereafter, the individual projects are briefly described, and their contributions to the development of the 

CBE paradigm are indicated in Table 1.

144 

Development of Consequence-Based Engineering Framework (FD) 

As discussed above, three projects guide the MAE Center research program by developing and 

articulating the application steps and operational rule of CBE; by steering the research effort towards 

minimizing uncertainty and by providing insight into opportunities for and impediments to the adoption of 

a new paradigm in the earthquake engineering community. The projects in this area provide leadership 

and steering power to the remainder of MAE Center activities (not least the core research program) and 

complete the process of elaboration and formalization of the CBE paradigm, with relevant application 

rules and verification examples. The three projects are: 

* FD-1 Consequence-Based Engineering Framework Development: This systems-level project 

develops and details the CBE paradigm as a framework for earthquake risk mitigation, and indeed 

other forms of risk, to vulnerable regions and systems. This project builds on and formalizes the 

operational rules for application of CBE through interaction with the various projects included 

under the core and stakeholder research programs and provides real-life examples of application, 

from conception to acceptance, documented in publications transferable to stakeholder groups. 

* FD-2 Systematic Treatment of Uncertainty: Develops procedures, models and control 

measures for managing uncertainty throughout the steps comprising the CBE paradigm application 

and represented in the core and stakeholder projects. 

* FD-3 CBE Implementation Opportunities and Challenges: Investigates the opportunities 

offered by, and the hurdles facing, the adoption of the paradigm by all different sectors of industry 

and society at large. This study will enable the MAE Center and its partners to launch measures 

through outreach to overcome the identified challenges and capitalize on the opportunities offered. 

Core Thrust Area on Damage Synthesis (DS) 

The ultimate objective of this core thrust area is development of innovative methods and refinement of 

existing ones to assess and visualize seismic risk across regions. The thrust area is focused on project DS- 

1 (Damage Visualization Module) which will integrate outputs from all DS projects, as well as fuse 

outcomes from corresponding "driving" projects CM-1 and HD-1 of the other two thrust areas. Towards 

this end, the following DS research projects are currently underway: 

• DS-1 Damage Visualization Module: This is a gateway project that integrates results of the DS, 

CM and HD thrust areas from the basic ingredients of hazard and vulnerability (before and after 

intervention) to project risk; hence, losses across societal systems. The project combines 

engineering seismologists, structural and geotechnical earthquake engineers and computer 

scientists and may be viewed as the deliverable from the entire MAE Center research program. 

» DS-2 Advanced Identification Technologies: Advanced techniques of data definition, selection 

and collection are studied here; such as remote sensing and image processing methods, both to 

collect inventory and to monitor losses after earthquakes. 

• DS-3 Response Analysis Tools: In this project, advanced concepts and techniques of analysis 

are developed and tested, with the aim of providing guidance for the derivation of fragility 

(vulnerability) functions and to increase the accuracy of damage estimates. 

• DS-4 Vulnerability Functions: Develops innovative approaches to investigate the vulnerability 

of structural systems and to derive fragility curves, with the objective of providing the most 

accurate and least uncertain set of fragility functions possible.

145 

DS-5 Response Simulation across Regions: The response of large populations of the built 

environment is characterized in this project for use in the visualization tool. 

DS-6 Network Economic Loss: Deals with refined estimates of economic losses due to network 

damage, taking into account time sequencing and a flexible spatial distribution of network 

ingredients. 

DS-7a Network Vulnerability Modeling: Aims to develop models for estimating the economic 

and social losses due to damage to large and complex transportation networks including the effect 

of disruption, down- and recovery-time and impact. 

DS-7b Damage-Functionality Relations: Derives relationships between reduction or loss of 

functionality on the one hand and level of damage on the other. 

DS-8 Social-Economic Impact Assessment: Studies the effect of an earthquake on social and 

economic impact measures in a region. 

DS-9 Risk Assessment Modeling: Develops new risk assessment models through establishing 

the relative merits of existing risk assessment approaches and applications results, leading to 

improved risk and loss modeling procedures. 

Core Thrust Area on Consequence Minimization (CM) 

The set of projects included in the area of CM use the outcome from the DS damage assessment; 

investigate and compare various measures of affecting the consequences through system intervention; and 

employ decision making and land-use management, alongside societal impact studies to zoom on to the 

most cost-effective consequence minimization measure to be applied to the system. The CM process also 

includes consulting the risk visualization facilities in DS to assess the effect of the consequence reduction 

measures proposed. The following CM projects are currently underway in the MAE Center: 

• CM-1 Probabilistic Decision Support for Regional Risk Assessment: This gateway project 

develops criteria for setting priorities for intervention and mitigation, based on optimization 

algorithms that are applied cyclically to the system under consideration, taking into account 

possible changes in its characteristics. 

• CM-2 Acceptable Consequences: Evaluates the level of acceptable consequences from a societal 

system viewpoint through developing reaction profiles and refining them based on field data, 

leading to the development of a dynamic societal response model. 

• CM-3 Network Retrofit and Rerouting Strategies: Develops procedures for balancing costbenefits 

of retrofit and/or re-routing, to aid in decision making as to the most effective intervention 

approach. 

• CM-4 Structural Retrofit Strategies: Addresses the level and type of structural intervention for 

populations of structure, to achieve a specified level of consequence minimization, by relating the 

intervention technique to changes in structural characteristics and deriving intervention-sensitive 

fragility relationships. 

• CM-5 Multi-Hazard Application of Regional Damage Synthesis: A test bed project that 

confirms and calibrates the damage synthesis module using a region, whose infrastructure system 

and hazards can be easily defined. 

• CM-6 Land Use Management: Minimizes consequences through the effective re-use of urban 

spaces, including relocation of vulnerable and hazardous buildings, widening of transportation 

arteries for emergency response and similar measures.

'• 

146 

Core Thrust Area on Hazard Definition (HD) 

The following HD projects are currently being executed, all of which are aimed at refining hazard 

characterization towards more accurate applications of the CBE procedure, through improving models and 

methods as well as providing new calibration data 

• BDD-1 Synthetic Earthquake Hazard: Peak ground parameters, synthesized earthquake timehistones 

and consequently spectra, are investigated in this gateway project for the purposes of 

representing source, path and site characteristics in areas where no actual earthquake records exist 

• HD-2 Intraplate Source Modeling: Robust understanding of intraplate earthquake mechanisms 

is the purpose of this project, with emphasis on the Central United States 

• HDD-3 Intraplate Ground Motion Data: Monitors and uses strong-motion data and fault creepshpoftheNMSZ 

• HD-4 Gujarat-NMSZ Correlations: Develops the relationship between the deep intraplate 

earthquake of Gujarat and earthquakes on similar fault systems, with special reference to the 

NMSZ, making use of the aftershock data collected by the MAE Center after the damaging 

earthquake in India, in January of 2001 

• HD-5 Seismic Path Modeling: Studies the attenuation of strong-motion in various media in the 

horizontal and vertical directions 

• HD-6 Site Modeling Site representation, both in terms of sub-soil condition and topographical 

effects, are developed in this project using measurements of earthquake motion on different soil 

conditions and geological features 

• HD-7 Ground Deformation Modeling: Denves factors affecting liquefaction and other forms of 

large ground deformations that may pose a hazard to the built environment 

TABLE 1 

CONTRIBUTION OF RESEARCH PROJECTS TO CBE DEVELOPMENT 

MAE Research Thrust Area 

Research Project 

CBE Framework Development 

FD-1 CBE Framework Development 

FD-2 Systematic Treatment of Uncertainty 

FD-3 CBE Implementation Opportunities 

Damase Synthesis 

[BS-I 

S.DS-2 

-BS-3 

IRS-4 

UDS-S 

•n-6 - 

•BS-7a 

!D&4* 

iDS-8 

'BS-9 

Damage Vjsoaliza&on Module 

• Advanced Identifi.catiQn' Technologies 

Response Analysis Toofe * 

" YHfcerabOlty Faactjons * 

Response Simulation across Regions "* ; 

êtwoaÊcoaamK-Los^ ^%i* 

* 

Mfttwrfffc ITrclttftraTvrFtfy Mnd&Fmg T : 

I teia^RiBc^^ « 

't Scoai-Bc^BOinic Impact Assessmflfft 

Risk-MsessmeBtModefim^, , 

Consequence Minimization 

CBE Paradigm Step 

1 2 3 4 5 6 1 8 9 10 

X 

X 

X 

X 

x 

X 

X 

j»r 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

x 

X 

X 

X 

X 

' 

X 

X 

X 

X 

"$, 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X. 

* X 

"X 

X, 

X 

X 

X 

X 

X 

X 

X 

X 

X 

1 J6 

X 

X 

X 

* X 

X 

X 

X 

X 

• X 

X ' 

X ; 

> X \ 

X i 

• X ' 

' X < 

* >'

147 

CM 1 

CM 2 

CM 3 

CM-4 

CM-5 

CM-6 

Hazard Definition 

,HD-1 

HD-2 

HD-3 

HD-4 

HD-5 

HD-6 

HD-7 

Probabilistic Decision Support 

Acceptable Consequences 

Network Retrofit/Rerouting Strategies 

Structure Retrofit Strategies 

Multi-Hazard Application of DS 

Land-Use Management 

Synthetic Earthquake Hazard -* 

Intraplate Source Modeling 

Intraplate Ground Motion Data 

Gujarai-NMSZ Correlations "- »- * 

Seismic Path Modeling 

Site- Modeling 

Ground Deformation Modeling 

X 

X 

X 

J, 

X 

X 

X 

X 

*. 

X 

_ 

* 

- 

— 

- 

X 

X 

X 

-, 

"*"" 

X 

X 

X 

X 

X.* ~ -„ 

X 

x: ** 

"X- 

£-* 

X 

X 

- 

X 

X 

X 

AC 

"* j 

*" 

i~ a. "* 

*. 

- 

X 

X 

X 

X 

X 

X 

_ 

~ 

, „ 

X 

X 

X 

X 

^X -4 

•"• 

* & 

*_ 

w 

* 

-, 

X 

X 

X 

X 

X 

X 

t,j> 

w~ 

^ 

* « 

•" - 

« 

CBE AND STAKEHOLDERS 

The structure of the MAE research program is based on the fusion of three core research thrust areas, with 

four stakeholder research thrust areas representing insurers, owners of large building stocks, construction 

associations and transportation officials To maximize leveraging potential, core thrust area research, 

supported with NSF and institutional matching funds, is implicitly intended to serve research interests of 

not one but multiple, stakeholder groups, as denoted graphically m Figure 4 In turn, research of 

individual stakeholder groups extends core research towards applications pertinent to their respective 

goals Thus, neither core nor stakeholder research thrusts by themselves can provide the appropriate 

balance between systems, enabling technologies and basic knowledge However, combinations of 

carefully configured core research with individual stakeholder research result in balanced research 

agendas having the most impact on earthquake risk reduction Further details on the Stakeholder Research 

Program of the MAE Center are given on the web site http.//mae ce umc edi* 

Transportation 

Officials 

Buildin 

Ow 

Construction 

Industry 

Insurance 

Industry 

FIGURE 4 RELATIONSHIP BETWEEN CORE AND STAKEHOLDER RESEARCH PROGRAMS 

The strategy for engaging stakeholders in learning about and indeed influencing core research projects, is 

critical to the success of infusing CBE m the earthquake engineering community, since affecting losses 

through the application of mitigation measures can only take place through stakeholders The four 

stakeholder groups targeted by the MAE Center, following a thorough review of national seismic nsk

148 

reduction priorities, have specific requirements that influence the strategic research planning for the CRP. 

Transportation officials are interested in assessing and minimizing the consequences of reduction or loss 

of traffic flow on their networks. The Construction Industry is interested in the market share appropriated 

by different construction form and material. Building owners are primarily interested in the probable 

damage to their properties from an occupant safety and continued operation viewpoints. Finally, the 

insurance industry is interested in quantifying the levels of relative risk on regional and national scales to 

refine setting premiums and deductibles as well as spread their exposure over areas of different levels of 

risk. The common thread amongst the stakeholder groups is that loss assessment and visualization, before 

and after the application of a number of risk reduction measures, is essential for their continued operation 

and effective future planning. This requirement is the motivation behind the development of the DS-1 

visualization module, which in turn drives the technical developments within the CRP. 

Another group with an interest in MAE Center research is engineering practitioners who will be early 

adopters of the CBE paradigm. Whereas consulting firms do not generally fund research, advice of 

practitioners with respect to the practicality of implementing a new engineering approach is essential for 

its ultimate success. Consultants with experience in providing seismic risk assessments for stakeholder 

clients are engaged with the MAE Center in refining further the research programs, as well as in 

developing specific projects, such as DS-9 on risk assessment modeling. 

CONCLUDING REMARKS 

The research projects currently underway at the MAE Center have been presented, starting from the 

definition of the guiding paradigm of CBE and its operational steps (Figure 1), to the contents of the 

individual projects. The relationships (input-output) between projects were presented in Figure 3, whilst 

the contribution of each project to the CBE application steps is shown in Table 1. The relationship 

between core and stakeholder research was briefly described and diagrammatically represented in Figure 4. 

By virtue of the above presentation, it is emphasized that the framework of CBE provides an effective tool 

for setting research priorities towards global seismic risk assessment and mitigation. The MAE Center 

research program presented above, in a CBE context, is comprehensive and fully coordinated. The 

ensuing research projects also provide a core of developments and application examples that are of use to 

a wide range of industrial projects and stakeholder interests. 

ACKNOWLEDGEMENT 

Development of the research plan described in this paper is funded primarily by the Earthquake 

Engineering Research Centers Program of the National Science Foundation under Award Number EEC- 

9701785. The authors have benefited substantially from discussions with the MAE Center Executive 

Advisory Board, Stakeholder Advisory Board and others.




APPLICATION OF EARLY DAMAGED AREA ESTIMATION 

SYSTEM USING DMSP/OLS TO RECENT DESTRUCTIVE 

EARTHQUAKES 

Kouichi HASEGAWA 1 , Mitsuhiro HIGASHIDA 1 , Masayuki KOHIYAMA 2 , 

Norio Maki 1 , Haruo Hayashi 1 , Herbert W.KROEHL 3 , 

Christopher D.ELVIDGE 3 and V. Ruth HOBSON 3 

1 Earthquake Disaster Mitigation Research Center (EDM), NIED 

2465-1 Mikiyama, Miki, Hyogo, Japan 

2 Institute of Industrial Science, University of Tokyo 

4-6-1 Komaba, Meguro, Tokyo, Japan 

National Oceanic and Atmospheric Administration, National Geophysical Data Center 

E/GC 325 Broadway, Boulder, Colorado USA 

ABSTRACT 

The authors are developing the Early Damaged Area Estimation System (EDES), which provides the 

information as to the estimated impacted areas within the first 24 hours after any significant 

earthquake based on nighttime lights observed by the Defense Meteorological Satellite Program 

(DMSP) Operational Linescan System (OLS). The estimation method is based on the detection of 

significant reductions or loss of lights in nighttime images following the event, for it can be expected 

that city lights will observably decrease after a large earthquake. 

The EDES targets Asia-Pacific region, which doesn't have tools to estimate earthquake disaster area, 

to support their activities after big earthquakes by providing the useful spatial information. In addition, 

humanitarian relief aids by various organizations all over the world should be distributed properly and 

fairly to those who need such aids in the impacted area, and it is desirable for international disaster 

response and relief communities to identify the exact location of impacted areas as soon as possible as 

for their quick mobilization. The EDES will also be able to support these communities with the 

information. We expect this system will contribute to global disaster management activities. 


It is necessary to grasp the spatial distribution of the damaged areas in early time in a large earthquake 

disaster in order to allocate limited human and physical resources efficiently. Shortly after a disaster, 

interruptions of communication and confusions of information would interfere with precise 

assessment of the disaster situation, and relief actions would come to confusion. Therefore, the early 

information of damaged area distribution is indispensable for effective disaster responses.

150 

In Japan, on reflection of the Hanshin-Awaji Earthquake Disaster 1995, national and local 

governments have developed damage estimation systems using GIS, which are based on social 

inventory database and observed seismic information. These systems provide real-time disaster 

information and enable efficient disaster responses. But only a few countries in the world have such 

damage estimation systems as developed in Japan. International rescue and research team should be 

deployed properly to the impacted areas in order to make their assistance effective. In addition, 

humanitarian aids by various organizations should be distributed properly and fairly to those who 

need such aids. Thus, it is desirable for international disaster response and relief committees to 

identify the location of impacted areas as soon as possible as the task for their mobilization. 

For the purpose of supporting relief and recovery activities by the governments or other organizations, 

the damaged areas were estimated using nighttime images observed by the Defense Meteorological 

Satellite Program's Operational Linescan System (DMSP/OLS). As examples two large earthquakes, 

the Turkey Kocaeli Earthquake (Mw7.4) on August 17 1999 and the India Gujarat Earthquake 

(Mw7.7) on January 26 2001 were picked up. 

2. DAMAED AREA ESTIMATION BASED ON 2 NIGHT LIGHT IMAGES BEFORE AND 

AFTER EARTHQUAKE (BASIC METHOD) 

It can be expected that city lights will observably decrease after a large earthquake due to various 

reasons such as electricity failure, building collapses, evacuation to shelters or the suspension of 

commercial activities. Therefore, the significant reduction in nighttime lights can be an indication of 

possible impacted areas due to earthquake disasters. The satellite images observed by the DMSP/OLS 

are suitable for the early identification of the damaged areas for following reasons: 

a) Due to the sensitive scanner, nighttime images are available. 

b) The nighttime images are observed twice a day by two DMSP satellites. 

These mean that we could detect significant reduction in nighttime lights at any day on a daily basis. 

The DMSP/OLS imagery has spatial resolution of 2.7 km, and the resolution is not as high as that of 

the Landsat/TM or the SPOT. But the recurrent periods of the satellites with high-resolution sensors 

are more than two weeks and the chances to observe the image immediately after a disaster are very 

low. Consequently, the short recurrent period is necessary for emergency use of the observed imagery. 

The steps of damaged area estimation are shown in Figure 1. Each pixel in the visible-near infrared 

(VNIR) images has digital number (DN) ranging from 0 to 63. We calculated the differences of DNs 

on a pixel basis between before and after the earthquake in the sampled area. The cloud influences 

were checked using the thermal infrared (TIR) images and the stable light images. Base on the 

histogram of the differences, the areas that show the reduction in nighttime lights withp > 0.995 were 

determined as significant reduction due to the earthquake disaster (Figures 2). The result maps were 

disseminated to the world through the Web page of EDM. The series of the estimation method is 

automatically executed in the server machine of EDM.

151 

Number of Pixels 

(Frequency) 

Threshold (p>99 5%) 

Estimated 

damaged areas 

Normal distribution 

based on sample 

ixels 

Difference of digital numbers (change of brightness) 

Figure 1: 

estimation 

Flowchart of damaged area 

Figure 2: 

estimation 

Criteria of damaged area 

3. NEW ESTIMATION METHOD BASED ON TIME-SERIES IMAGES 

Some problems shown in the followings remain in the estimation results based on two images before 

and after earthquake. 

a) Cloud influence remains even if Stable Light Image is applied. 

b) Reflection of water area leads to mis-judgment 

c) Small cities are neglected due to the variance of the night light 

These problems would be resolved, if pixel-based average and standard deviation of digital number 

were analyzed for several images before an earthquake and significant test were applied to post-event 

image. This procedure is called "Time-Series Images Method". It takes about 10 hours to complete 

analysis using more than 30 images. It is important to notice that the gain of the sensor estimated 

originally affects the results in the period that moon light is bright. So it is better to use images in 

about 10 days before and after new moon. As this method is operated manually now, automated 

system for this method should be built. 

4. APPLICATION OF BASIC METHOD TO THE TURKEY MARMARA EARTHQUAKE 

DISASTER 

We estimated the possible impacted areas of the 1999 Marmara Earthquake Disaster that occurred in 

northwest Turkey on August 17. The DMSP/OLS images in this region were provided from National 

Oceanic and Atmospheric Administration's National Geophysical Data Center (NOAA/NGDC). The 

nighttime images before and after the earthquake are shown in Figures 3 and 4 respectively. 

Fortunately, the two images have little cloud influence considering the TIR images (Figures 5 and 6). 

The histogram of digital number differences between two VNIR images is shown in Figure 7. 

The estimated impacted areas spread widely in Yalova, Kocaeli and other provinces (Figure 8). These 

estimation results showed with a high degree of correspondence with the real damages.

152 

Before EQ 

After EQ 

Figures 3 and 4: VNIR images before and after the earthquake in Turkey 

Before EQ 

After EQ 

Figures 5 and 6: TIR images before and after the earthquake in Turkey 

-50 -40 -30 -20 -10 0 10 20 30 40 50 

Figure 7: 

Histogram of digital number differences between two images (log-scaled)

153 

Red: possible impacted areas Gray: saturated data 

Figure 8: Estimated damaged area in the Marmara Earthquake Disaster 

5. APPLICATION OF BOTH METHODS TO THE GUJARAT, INDIA EARTHQUAKE 

DISASTER 

We estimated the possible impacted areas of Gujarat, India Earthquake Disaster on January 26, 2001 

using basic and new method of EDES. The VNIR and TIR image on January 24 shown in Figure 9 

and 10 was selected as an image before the earthquake. Figure 11 and 12 shows VNIR and TIR 

images after the earthquake. These images were used for basic method. The distribution of the 

average and standard deviation of digital number with each pixel for new method are shown in Figure 

13 and 14. The estimated damaged areas using basic method and new one are shown in Figure 15 and 

16. The estimated impacted areas spread in Gandhi Dham, Bhuj and Morbi. Some low reliability 

estimation areas caused by cloud, water or other reasons remain in basic method estimation image 

(Fig. 15), but there exist no such areas in new method estimation image (Fig. 16). Comparing these two 

estimation results, it is revealed that new method estimation is better than basic one.

154 

Before EQ 

After EQ 

Figures 9 and 10: VNIR images before and after the earthquake in India for basic method 

Before EQ 

After EQ 

Figures 11 and 12: TIR images before and after the earthquake in India for basic method 

Average 

Standard deviation 

Figures 13 and 14: Distribution of average and standard deviation acquired from time-series 

images for new method

155 

Low reliability 

-(water-reflection 

5 n) - - -„ - -'*>/ -- /- - < 

"V Epicenter (Mw7.7) / l-S 

A - a... .:._.„,] " '.UdaipuK-- _ -, 

Figure 15: 

• Estimated damaged areas 

• (P>99%) 

Estimated damaged areas 

(P>95%) 

" Saturated data 

_, 

|g Pre-event clouds 

Post-event clouds 

Keshod 

Mangn 

200 km Veraval 

Red, Yellow: possible impacted areas Gray: saturated data 

Estimated damaged area using Basic Method in the Gujarat Earthquake Disaster 


• (P>99%) 


(P>95%) 

a Saturated data 

Red, Yellow: possible impacted areas Gray: saturated data 

Figure 16: Estimated damaged area using New Method in the Gujarat Earthquake Disaster

156 

6. CONCLUSION 

We applied the proposed method of the early damaged area estimation system to two earthquake 

disasters. It is revealed that the basic method estimation is considerably accurate at least in case that 

the cloud influence is little, whereas the estimation by new method using time-series images is hard to 

suffer the several kind of influences and result in more reliable. 

The basic method system has been automated, but the new method one is manually operated. We're 

going to build automation system for new method within a couple of years. 


The DMSP-OLS images used in this study are provided by NOAA/NGDC. We wish to thank 

NOAA/NGDC and Dr. Haruhiro Fujita for the cooperation. 

REFERENCES 

Hashitera, S., Maki, N., and Hayashi, H. (1999), "The Potential of Using Satellite Images to 

Determine an Index of Recovery from Natural Disaster: A Case Study of the Great Hanshin-Awaji 

Earthquake Disaster," Proceedings of the 6th Japan/United States Workshop on Urban Earthquake 

Hazard Reduction, pp. 492-495.




157 

CCD CAMERA SYSTEM APPLICATION 

FOR DISASTER MANAGEMENT 

Mitsuhiro HIGASHIDA 1 , Norio MAKI 1 and Haruo HAYASHI 2 

'Earthquake Disaster Mitigation Research Center, 

National Research Institute for Earth Science and Disaster Prevention 

Hyogo, JAPAN 

2 Disaster Reduction System Research Center, 

Disaster Prevention Research Institute, Kyoto University 

Kyoto, JAPAN 

ABSTRACT 

The goal of this study is to establish a numerical disaster process simulation model, which 

makes it possible to estimate when the recovery from disaster will complete, what factor 

accelerates physical recovery. This paper discusses about the following topics, as the first step 

of this study; 1) the method to get the time series of physical recovery degree from a series of 

observed images, 2) the relationship between the physical recovery degrees and behavior of 

impacted people. 


Though seven years have passed, the reconstruction urban planning after the 1995 Kobe 

earthquake is still on the way. It does not means that the quickness is the first priority in the 

reconstruction urban planning. The upgrading not only in the physical level but also in the 

regional economy and the community activities level are goals of the reconstruction urban 

planning. However, seven years is very long time in our life. 

Present reconstruction urban planning shows only the drawing of complete image. Thinking 

that it takes long time to complete, not only designing the complete image but also designing 

the process is necessary. Urban planning should become to have the sense of time. It says that 

planner should explain what the area would become in each time phase of recovery or 

reconstruction process. To do so, the simulation on disaster process of a planned area is 

necessary. However, the urban reconstruction study from disaster has spotlighted to the 

counter-measures or reconstruction urban planning itself until now. Even the evaluation studies 

on these counter-measures are limited. So there does not exit the data sets to trace the disaster or 

disaster process, which is indispensable data to make simulation. 

We continue to study on "disaster process" or "recovery process" after natural disaster. The 

knowledge on disaster process is constructed from the systematic knowledge on 1) variety of 

problems, 2) scale of problems, 3) chronological data on each problem and 4) the 

countermeasures to problems. The concept of disaster process will be shown in Fig.l. The data 

on how the impacted people behaved in each phase of disaster process has been collected by the

method of questionnaire sheet survey and interviews until now. (Hayashi et. al. 1997, Kimura et. 

al. 1999, 2000, Tanaka et. al. 2000) 

However, these data should be understood with the relation between physical environment 

change data such as physical damage, situation of shelters, dismantle of damaged buildings, a 

land use change and reconstruction of buildings. It is necessary to clarify the relationship 

between behavior of people and physical recovery to understand a holistic disaster process. 

The goal of this study is to establish the numerical disaster process simulation model, which 

makes it possible to estimate when recovery from disaster will complete and how we progress 

recovery in the most effective manors. And we will distinguish what factor accelerates physical 

recovery by more quantitative manner by analyzing the relationship between activities for 

recovery and physical recovery. To do so, long term continuous monitoring is necessary to 

record physical recovery. It is essential data in a natural science field. For example, continuous 

ground motion monitoring is very common method in seismology. 

How can we monitor physical recovery An aerial photograph is the best way to monitor 

physical environment changes of broad area and it is easy to convert into GIS data. However, it 

is difficult to collect continuous data of physical recovery using aerial photograph. The 

continuous monitoring using CCD camera system is used for monitoring natural hazards such 

as volcanic eruption and flood. And we have established a continuous disaster process 

monitoring system using CCD camera system. This system can monitor not only the physical 

recovery but also many data on disaster process, such as damages, situation in shelters, and 

activities at the impacted area. The implementations of this system into a real disaster response 

will be discussed in anther paper. 

This paper discusses about following topics as the first step of this study; 1) the method to 

get the time series of physical recovery degree from a series of observed images, 2) the 

relationship between the physical recovery degrees and behavior of impacted. 

158 

Fig.l Disaster Process 

2. OUTLINE OF THE MONITORING SYSTEM 

2.1 Outline of the monitoring system 

We have started continuous monitoring on disaster process at impacted areas for a disaster 

such as Kobe-Japan, Chungliao-Taiwan and Miyake Island-Japan and collected the digital 

images of the disaster process of impacted areas twice a day using CCD camera system. 

Observation data of each day is now accumulated to data server in Japan through the Internet. 

And these data is distributed through the Internet at the following address: 

http://inpaku.dpri.kyoto-u.ac.jp/jp/join/live/index.html. The outline of the system will be shown 

in Fig.2.

Modem or Router 

Dial-up 

ISP 

The Internet 

WWW Server(X) 

FTP Server 

159 

CCD Camera 

System 

Pnvate network 

FTP 

Internet Client 

Fig.2 Outline of the monitoring system using CCD camera 

2.2 Continuous monitoring at Nagata, Kobe 

Misuga-west of Nagata, Kobe was severely impacted by the 1995 Kobe earthquake of Jan. 17, 

1995 and the fire induced by this earthquake. The damage statistics in this area is as follows; 

major damage and burnout - 242 buildings, moderate damage - 34 buildings among 334 

buildings in this area. So severe damage (Major and Moderate) ratio amounts to 83%. The data 

of this area is distributed from two sets of CCD camera which were set on the roof of Mikura 5, 

where the "Machi-communication" that is community based organization working for 

community development for this area locates. This building was reconstructed after the disaster. 

The data collection was started from March 19, 2001. The location and the images of CCD 

camera system will be shown in Fig.3. 

,^P *el ,^ 

++ 

r^ 

8^4SPy35 

v* &£l§^4E2i[ 

."* Ml 

Hnli 

I | %& 

Ui .1* •^^ 

CCD1 

"^^"^T Vifclil^ /T^TM 

""V^ % 

4^'* 

Fig.3 Monitoring at Nagata, Japan 

2.3 Continuous monitoring at Chungliao, Taiwan 

Chungliao, Taiwan was severely damaged by 9.21 Taiwan earthquake, 1999. The building 

damage at Chungliao is 2, 542 major damage buildings, 1,424 moderate damage buildings, and 

human death amount to 179 peoples. The cameras were set at Yunping village at Chungliao. 

The community based reconstruction program, which include creating small business and 

rediscovering the regional heritage such as the canal that has not used for many years as well as 

the reconstruction of buildings, was progressed in this village supervised by Prof. Yu of 

Chung-Yuan University in Taiwan. The temporary town including the village office, library,

kindergarten, library, and dwellings were established at the other area where these official 

buildings located before the disaster, because all these official buildings were collapsed for the 

earthquake. The reconstruction of original area where these facilities were located before the 

disaster is still on the way. Continuous monitoring on both temporary town and reconstruction 

area is conducted using five CCD camera sets. Observations were started from April 19, 2001. 

The location and the images of CCD camera systems will be shown in Fig.4. 

160 

CCD1 CCD2 CCD3 CCD4 

Fig. 4 Monitoring at Chungliao, Taiwan 

CCD5 

3. IMAGE PROCESSING 

3.1 Processing procedure 

These images are useful to understand present status and sequential images are useful to 

understand a physical disaster process by intuition. However, amount of data is huge and the 

development of automated data processing method is necessary to get the numerical data on the 

physical recovery. And to make a comparative study, the standardized procedure is 

indispensable. The method to get the time series of the physical recovery from a series of 

observed images will be discussed in this chapter. We will define the amount of buildings as the 

physical recovery degree from disaster. 

The observed data contains the noises such as 1) moving objects such as cars, people, 2) the 

shadow of buildings from the viewpoint, and weather such as rain. So removing these noises is 

the first step for image processing. In this paper, the basic procedure to get physical recovery 

degree using observed data is discussed. On the other hand, moving objects means physical 

recovery. This is a very important element to understand recovery process. 

3.2 Removing the noises 

3.2.1 Removing moving objects 

The method to get averaged image of observed images was used to remove moving objects

from observed images Fig 5 shows the results getting averaged image of three images. It shows 

that the cars observed in the image of March 25 and people observed m the image of April 6 are 

removed in the averaged image. 

3.2.2 Removing shadows in observed Images 

There exist shadows of many objects such as trees, cars, and buildings in observed images. 

These have possibility recognized as the physical recovery during image processing The 

observed images have been composed from three colors, red, blue, and green. Each color has its 

own charactenstics against shadows. The method to remove shadows using these characteristics 

was developed. The procedure of this process is as follows. 

a Observed images are recomposed to law data of red, blue, and green (Fig 5) 

b The value of each color is adjusted into the value that does not have effects from the 

sunlight 

R'=(R-G/R+G)xlOO (1) 

B'=(B-G/B+G)xlOO (2) 

R' value has characteristic that effects from shadows is weak And R' value is used for die 

following analysis. 

c Comparing two sets of adjusted image according to following formulation. 

AR'=|RO'-R1'| (3) 

AR': Comparison result, RO'* Base Averaged Image , Rl': Reference Averaged image 

d. The compared images (acquired from process of c) are classified into two values, black and 

white according to a threshold. Threshold value 5 is used in this analysis. Acquired image 

will be shown in Fig.6. 

3/25 10 00 4/6 10 00 4/20 10 00 

Averaged image of 3/25 4/6 and 20 Averaged image of 5/181 Sand 20 

Red Value Green Value Blue Value 

Fig.5 Removing Moving objects 

Fig.6 Shadows removed image 

3.3 Acquiring amount of physical recovery 

There still exist many noises without physical recovery (ref. Fig6). These noises are 

generated from the effect of sunlight against existing buildings and the edge of roads. These 

noises are removed by the method of smoothing. The result is shown in Fig.7. And acquired 

results will convert to the GIS data. About the method for conversion will be discussed another 

paper.

162 

Fig.7 Amount of physical recovery 

4. DISASTER PROCESS AT MISUGA-WEST AT NAGATA, KOBE 

4.1 Physical Recovery Degree 

Fulfillment ratio to FLR or BLR designated by City Planning Act can be thought as the 

recovery degree or index at the impacted area. Misuga-Nishi is the dwelling area and 

designated FLR is not usually fulfilled. So BLR is used to describe the physical recovery 

degree. 

4.2 Physical disaster process 

Disaster process monitoring at Nagata starts from March 19, 2001 and monitoring period is 

limited. The physical disaster process from the beginning was supplemented using the field 

survey data conducted by "City Planning Institute of Japan (CPU) and Architectural Institute 

of Japan (ALT)" and "Machi-communication". Fig.8 shows the damage in this area (CPU and 

AU) and Fig.9 shows the situation at relief phase ("Machi-communication") and Fig. 10 show 

the present situation. The circled buildings are the buildings picked up from the image 

processing. 

Using the time series data of CCD camera and supplement data, the relation between time 

and recovery degree is acquired as Fig. 11. 

Ma, or damage 

Mods-ate damage 4$& ^ %.W* 

Slight Damage * , "«** 1 Temporary buildings 

No damage \J^~ I^J Survived buildings 

Burned out 

No data 

B d n BRI ° 

^" Based on Mach ' Communications 

;ao 

Fig. 8 Damage (survey by CPU and AU) Fig.9 Relief Phase (Survey by Machi-com)

163 

Picked up building from 

image processing 

JBSj 

Reconstructed 

Based on Machi Communications 

Fig.10 Present Status (Survey by Machi-com) 

40.0% 

Dismantle pf voluntary temporary 

Reconstructioii 

20.0% 

0.0% 

500 1000 

(days) 

1500 2000 2500 

Fig.ll decision-making and physical recovery 

4.3 Decision making and physical recovery progress 

Table. 1 shows decision making on reconstruction urban planning at Misuga-west The 

relation between decision-making and physical recovery will be shown in Fig.ll. 

5. COMMENTS 

This paper shows the basic procedure on disaster process monitoring and very first results on 

disaster process monitoring using CCD camera. There are still many points to need to be 

improved on image processing. Future subjects are 1) improvement of reliability to convert 

observed images to observed values and 2) improvement of the analysis method of a physical 

recovery degree. However, the possibility to monitor physical disaster process was clarified. 

It is usually said that it takes at least ten years to complete recovery from the disaster.

Chronological GIS based datasets both on human behavior and physical recovery continue to be 

collected at Nishinomiya, eastern part of Kobe, which contains the data on damage, relief, 

recovery, and reconstruction. We plan to be established same dataset in Chungliao, Taiwan by 

the cooperation with researchers at Taiwan. Using these datasets, the numerical disaster process 

simulation model will be established in near future. 

164 

Date 

17-Jan-95 

jTjMariS 

_2C-Apr-95. 

1 -Mav-95 

J-Sepc95 

J3-Sepr95 

1 4-Jan-97 

Ji8rFeb-91 

LQHy1air97 

^7_-N°vr97 

_lL-Jarr98. 

Jan-99 

Apr-99 

Dec-99 

Mar-00 

^T- 

TABLE.l DECISION MAKING PROCESS 

Top,c 

0 1995 Kobe earthquake 

60 First Decision _qn reconstruction urban.pjanntng „__ _ __ 

_ -96 Machizukun organi28tionjAjQS_established __ _ „ _ .„ 

104 Inquiry sheet survey on reconstruction urbanjjlanning 

224 |nquiry_sheet survey __qn__reconstruction urban planning _____ 

„ _235 .Urban planning madejpy Machizukun organisation were prppqsed to_Kobe city government _ 

717 Practical plan '-was decided 

-J^JJSepJ^rdjdejqisionjsn „. 

_713^Qpmmittee_ numbers on_thejand readjustment was elected _ _ _ 

_ »LQ30,Distnct P!anning_was established 

___ 

1075 J-and .readjustment starts _ _ „.,_„ 

1424 Construction of Mikura 5 starts 

1514 Construction of Public apartment in this area starts 

1 754 Complete Mikura 5 

1J344 Complete the public apartment _ 

Acknowledgement 

About setting CCD camera, "Machi-Communications" at Kobe and Prof. Yu of Chung-Yuan 

University of Taiwan, Prof. Lian-Chen Chen of National Taiwan University, and Chungliao 

government office at Taiwan, were extremely helpful. I would like to show the special thanks to 

them. 

Reference 

Building Research Institute, Final report on damage from the 1995 Kobe earthquake, 1996, (in 

Japanese). 

Hayashi, H., Shigekawa, K, Producing Disaster Ethnography for the Development of Disaster 

Ethnology, Papers of the Annual Conference of the Institute of Social Safety Science, No.7, 

ISSS, Japan, 1997, pp.376-379, (in Japanese). 

Higashida, M., Maki, N., Hayashi, H., Development of Recovery Process Observation System 

using CCD camera after a disaster, Journal of Social Safety Science, No.3, ISSS, Japan, 2001, 

pp.95-100, (in Japanese). 

Kimura, R., Hayashi, H., Tatsuki, S., Urata, Y, Clarifying the human behavior of the disaster 

victims after the Great Hanshin-Awaji earthquake, Journal of Social Safety Science, No.l, ISSS, 

Japan, 1999, pp.93-102, (in Japanese). 

Kimura, R., Hayashi, H., Tatsuki, S., Determinants and Timing of housing Reconstruction 

Decision by the Victims of he 1995 Hanshin Awaji Earthquake Disaster, Journal of Social 

Safety Science, No.2, ISSS, Japan, 2000, pp. 15-24, (in Japanese). 

MacM-Communication 

Tanaka, S., Hayashi, H., Shigekawa, K., Kameda, H., Development of Standardized Procedure 

for. Disaster Ethnography Interview, Case Comparison, Coding, and Disaster Process 

identification-, Journal of Social Safety Science, No.2, ISSS, Japan, 2000, pp.267-276, (in 

Japanese).




A SIMPLIFIED METHOD TO EVALUATE LIQUEFACTION 

OF SUBSOIL OF BUILDING 

Jing Liping and Wu Zhaoying 

Institute of Engineering Mechanics 

China seismological Bureau, Harbin 

Abstract 

In this paper, based on the simplified method proposed by Men et al and Martin's model of the 

pore water pressure, a new simplified method is presented to evaluate the pore water pressure in 

the subsoil of buildings during earthquake. According to the cone model theory the subsoil of 

buildings may be approximately divided into three zones, in which the seismic shear stress and 

additional dynamic stress can be evaluated with the use of Seed's simplified method and the 

cone model. Then the pore water pressures in these three zones may be approximately evaluated 

with the use of Martin's model The method could be directly applied to analyze the initial 

liquefaction potential of subsoil of building with pore pressure ratio. 

Key words subsoil of building, cone model, pore water pressure, seismic liquefaction 

Introduction 

As is well known up to the present soil liquefaction evaluation methods available have been set 

up largely for the free field without building, These methods may be divided into four groups i.e. 

(1) Seed's simplified method, (2) seismic response analysis method including both total and 

effective stress approach, (3) empirical formulae, (4) probabilistic and statistical method. Since 

the 1964 Niigata Earthquake and Alaska Earthquake brought about an extensive damage of soil 

liquefaction, many investigators have devoted much attention to study seismic liquefaction 

problems and remarkable achievements have been gotten as shown in the methods of four groups 

cited above. When buildings are existed the pattern of liquefaction would be different from that 

in the free ground. However seismic liquefaction evaluations of subsoil of buildings have not yet 

been resolved in such a way that general and practicable methods with sufficient accuracy and 

plausibility can be established for the use of engineering design and analysis. This fact may be 

due to that (1) building subsoil are much more complicated to deal with in contrast with free 

field of ground since the existence of building makes a non-uniform stress field and a 

complicated wave propagation pattern due to soil-structure interaction and (2) a variety of 

influencing factors relevant to building such as size, shape, rigidity, base area, buried depth of 

foundation, dead and seismic loading, etc. have to be considered There have been only limited 

results of study on the problem and model tests 1 " 4 , but these studies bring large time and 

manpower consumption and could not meet the needs of evaluating liquefaction of subsoil of 

building in engineering 

Men et al (1997) 5 have presented a new simplified method to evaluate seismic liquefaction of 

subsoil of buildings. They have devised the method in such a way that it is simple to handle and 

has clear physical insight and sufficient engineering accuracy by taking advantage of Seed's

166 

simplified method for free ground site and of the cone model concept well developed by Meek 

and Wolf in dynamics for scattered field. Their conclusion showed the reasonableness and 

tendency in good accord with results of small model tests and of finite element analyses. 

In this paper based on Men's simplified method and considering the nonlinear constitutive 

relationship of soil and Martin's model of pore water pressure, another method is suggested for 

evaluating pore water pressure of subsoil of building. The method is more suitable for evaluating 

seismic liquefaction potential of subsoil of buildings in engineering 

Men's simplified method for evaluation of seismic liquefaction of subsoil of building 

Men et al adopted following assumptions in their method proposed. 

1. The building is rigid as a whole. 

2. The building is situated on the ground surface and represented only by a footing 

3. The soils may be regarded as elastic under dead and seismic loading condition. 

4. The rocking component of building motion is neglected in evaluation of liquefaction 

potential because only shear stress is important to induce residual pore water pressure. 

5. A plane shearing body wave SH is propagated vertically upward from the source of 

earthquake. 

Based on the above assumptions and the substructure method to deal with the structure reaction 

to seismic input, Men's simplified method can be summed up as consisted of the following 

procedures. 

For given displacement, the wave field in subsoil can be written as following 

u~u f +u^ (1) 

where u^ is the free field ground motion and « 4 is the scattered field motion induced by 

building, and u is the total motion of subsoil of building. If the input motion from the 

earthquake source is u t = u ia f (t-z/c) then the free field motion of homogeneous elastic media 

will be 

(2) 

where / (t) is an input function, c s is a shear wave velocity, z is a vertical ordinate downward. 

For given stress, the dynamic stress in subsoil of building can be written as following 

where rôâre the total dynamic shear stress and total normal stress in subsoil of building, 

respectively, T / ,a / are the dynamic stresses in free ground, T 4 ,cr, are the additional dynamic 

stresses due to existing building. 

In the simplified method proposed by Men et al, the dynamic stress in free ground can be

167 

evaluated by Seed's approximate method. The additional ground motion or the additional 

dynamic stress induced by the presence of building can be evaluated approximately with the 

cone model well developed by Meek and Wolf in dynamics in recent years. 

When the cone model is used, the shape of foundation should be equivalent to a fixed shape, 

such as a circular footing, square footing or a strip footing. According to the cone model theory. 

The two cones, one shear and another compression can confine the influence of structure. 

Because the apex height of the shear cone and the apex height of the compression cone are 

different, the total solution are divided into three zones as named as I, II and III, as shown in 

Figure 1. 

There are the additional vertical dynamic, static stresses and the additional dynamic shear 

stress in zone I. There is only the additional dynamic shear stress in zone II. The stress 

distribution is same as of the free ground in zone III. The procedure of evaluating seismic 

liquefaction of subsoil of building using the simplified method proposed by Men et al as follows. 

1. According to the cone model theory, the subsoil of building are divided into the three zones as 

shown in Figure 1. 

Determine the free field dynamic shear stress by using 

Seed's simplified method. 

2. Determine the additional shear stress and normal 

stress of each zone by using the cone model. 

3. Determine the total dynamic stress. 

4. Use triaxial dynamic compression apparatus or 

dynamic simple shear test devise to obtain the soil 

liquefaction resistance r^for relevant point to be 

considered. 

5.Make a comparison between r d and 1 R for the every 

point to see if that soil point may liquefy, as do in 

Seed's simplified method. 

Figure l.the cone models 

The simplified method to evaluate the pore water pressure in subsoil 

It is well known that volume of dry sand is changed under shear, i.e. the so-called dilatancy 

under cyclic shear loading. For saturated sand this compression is gradually compensated by the 

elastic rebound deformation of sand skeleton enabling thus the residual pore water pressure to 

generated and the effective stress to decrease until ultimately the liquefaction occurs for certain 

loading level with sufficient number of cycles 

We neglect the inertial effect and the compressibility of pore water and sand grains here and then 

following Martin et al (1975) 6 we can write out the pore pressure increment as 

c s 2 

A£... =c,(y-c,£,,,) + ——— ( 5 ) 

( 6 ) 

where e dv is the volume compression due to cyclic shear, i\£ dv ,i\G w the are increment of

168 

volume and residual pore pressure within one load cycle, respectively. t r is the tangent rebound 

modulus of sand skeleton, y is the shear stain, c { ~~ c 4 are the test constants. 

Note that t r is relates to the effective confined pressure and can be expressed by an empirical 

formula, for example of the form (Martin et al 1975) 

E= (7) 

Where a v0 is the vertical effective normal stress at the beginning of unloading, a v is the 

residual normal stress after unloading, & 2 , m,n are the test constants, respectively. 

Using Massing' s model, the non-linear relationship of the shear stress and strain can be 

expressed as follows 

Therefore 

T==(e7,)'"-- (8) 

a + by 

1— 

where a,b are test constants, respectively. 

Substituting formula (9) into formula (5), the pore water pressure can be approximately 

calculated with formula (4) in terms of r after Nth cyclic loading which the dynamic shear 

stress T can be calculated with the simplified method proposed by Men et al 

Evaluating seismic liquefaction potential of subsoil of building 

After obtaining pore water pressure, based on the initial liquefaction criterion, the estimation of 

liquefaction can be made with the pore pressure ratio. The pore pressure ratio is defined as a 

ratio of the pore pressure to the vertical confined stress. According to the cone model, the ratio of 

pore pressure in the three zones of subsoil is as follows, respectively 

Zone I R(Z,N)= - - - (10) 

zone nan 

where a£ is the pore water pressure after Nth cyclic shear loading.

169 

R(z,N)is reached to 1 or more, the initial liquefaction occurs. 

Typical example 

To show the basic ideal and principle of the method proposed in this paper, as a complete 

example, we give a concrete evaluation of liquefaction of subsoil under a strip footing with the 

width of 2b Q , which rest on the ground surface of an isotropic homogeneous semi-space. 

According to the cone model theory, Men et al gave the stress distribution of the each zone in 

subsoil, as follows 

Zone I r f = 

§ 

°>-^7 

( 12 ) 

w a mf 

a. =±- z+zA 2g 

Zone II ( 13 ) 

Zone III 

If r — W*VJ 

( 14 ) 

where i d , a are the total dynamic shear stress and total compression stress, respectively. r s , 

is the additional dynamic shear stress due to existing building, w is the vertical static load 

undertaken by the footing, and A is the area of footing, zj, ZQ are tne a P ex ne ig nt of shear 

cone and compression cone ^respectively, p is the volume weight of soil, p, is the buoyancy 

volume weight of soil, a mdx is the peak acceleration of ground, g is the acceleration of gravity. 

Using the above formulas, the equivalent cyclic shear stress amplitude and the vertical 

compression stress of each point with arbitrary depth in each zone can be calculated respectively. 

Then, the equivalent cyclic shear strain can be determined in terms of nonlinear constitutive 

relationship (9). Finally, the pore water pressure produced under a definite cyclic shear loading 

value in different zones of subsoil can be evaluated.

170 

For an example, taking Possion ratio of soil /^=0.33, p = 20&V/m 3 , pj=10W/m 3 , 

w = 100&V , b Q = 1m ,

171 

subsoil and combined with the nonlinear constitutive relationship of soils and Martin's model of pore water 

pressure, a simplified method is suggested for evaluating pore water pressure of subsoil of building This 

method is simple, practical, and convenient for the prediction of seismic hazards due to liquefaction potential 

of subsoil of ordinary building. 

REFERENCES 

1. Y. Yoshiaki et al., (1977). Settlement of building on saturated sand during earthquake, Soil 

Foundation. 17 1. 

2. Huishan Liu et al.,(1984). Liquefaction behavior of saturated sand layer under footing, 

Aseismic Behavior of Base Soils and Industrial Construction, Seismology Press (in Chinese),. 

3. Finn and M. Yogendrakumar, (1987). Analysis of pore water pressure in seismic centrifuge 

tests, in Soil dyn. liquefaction, Elsevier, Amsterdam. 

4. R. Popescu and J. H. Prevost, (1993). Centrifuge validation of a numerical model for dynamic 

soil liquefaction, SDEE, pp. 73-90. 

5 Fu-Lu MEN and lie Cui, (1997). Influence of building existence on seismic liquefaction of 

subsoils, EESD, Vol 26, 691-699. 

6. Martm,G.R.et al, (1975). Fundamentals of liquefaction under cyclic loading. J.Geot. Eng. 

Dw.,ASCE, 101(5).




A GENERAL APPROACH TO SEISMIC PERFORMANCE 

ASSESSMENT 

H. Krawinkler 1 

'Department of Civil and Environmental Engineering, Stanford University 

Stanford, CA 94305-4020, USA 

ABSTRACT 

Performance-based earthquake engineering implies design, evaluation, and construction of engineered 

structures whose seismic performance meets the diverse economic and safety needs of owners and 

society. As the First major step towards an integrated design/assessment approach, the Pacific 

Earthquake Engineering Research (PEER) Center has focused on the development of procedures, 

knowledge, and tools for a comprehensive seismic performance assessment of buildings and bridges. 

This paper focuses on the general performance assessment methodology developed by PEER 

researchers for buildings. The approach is aimed at improving decision-making about seismic risk by 

making the choice of performance goals, and the tradeoffs they entail, apparent and transparent. In the 

approach, decision variables are identified whose quantification, together with an assessment of 

important uncertainties, will make it feasible to characterize and manage economic and societal risks 

above and beyond potential loss of life and injuries. 

INTRODUCTION 

The following paragraph was written at a time when the value, practicality, and direction of 

performance-based earthquake engineering (PBEE) was a matter of much debate in professional and 

academic organizations (Krawinkler, 1997): 

"Performance-based seismic engineering is a noble concept. Its implementation, however, has a long 

way to go. There are legal and professional barriers, but there are also many questions whether it will 

be able to deliver its promises. It appears to promise engineered structures whose performance can be 

quantified and conforms to the owner's desires. If rigorously held to this promise, performance-based 

engineering will be a losing cause. We all know that we cannot predict all important seismic demands 

and capacities with confidence, even in a probabilistic format. There are, nevertheless, compelling 

reasons to advocate performance-based engineering as a critical area for research and implementation. 

The objective of seismic engineering should be to design and build better and more economical 

facilities. Both terms, "better" and "more economical", are relative to the status quo. In the writer's 

opinion, significant improvements beyond the status quo will not be achieved without a new and 

idealistic target to shoot for. We need to set this target high and strive to come close to its 

accomplishment. We may never fully reach it, but we will make significant progress if we have a well

174 

defined target. Performance-based seismic engineering is the best target available, and we need to 

focus on it." 

Now, five years later, the idealistic target has moved closer to reality. We still have a long way to go, 

but we hope to have laid the groundwork, established a framework, and provided basic tools that make 

PBEE realizable and an attractive alternative to conventional analysis and design. In particular, 

significant progress has been made in establishing performance metrics and developing procedures for 

a rigorous performance assessment that combine seismological, engineering, financial (e.g., $ losses) 

and societal (e.g., life safety) considerations. This paper tries to summarize the approach developed by 

PEER researchers for this purpose. 

One concern needs to be expressed, and one notion needs to be dispelled, up front. The concern is that 

the proposed approach relies heavily on the availability of seismological, engineering, and financial 

data (e.g., construction costs after an earthquake). Much of the needed data is of questionable 

reliability at this time. This is not a drawback of the approach, but points out an urgent need for data 

acquisition and modeling. The notion to be dispelled is that the future will consist of complex 

probabilistic approaches in routine structural design. This should be an option at the discretion of the 

engineer or owner, but should not be necessary for routine designs. The main objectives of the 

development is to 

• Facilitate decision making on cost-effective risk management of the built environment in areas 

of high seismicity 

• Facilitate the implementation of performance-based design and evaluation by the engineering 

profession, and 

• Provide a foundation on which code writing bodies can base the development of performancebased 

provisions. 

The latter should result in relatively simple but more transparent and risk consistent provisions than are 

contained in present codes and standards. 

COMPONENTS OF PERFORMANCE ASSESSMENT APPROACH 

This paper is concerned with performance assessment of buildings. The assumption is that all relevant 

building systems, i.e., the soil/foundation/structure system as well as the nonstructural and content 

systems, are given. The components of the performance assessment approach are summarized here (in 

part from Cornell and Krawinkler, 2000) and illustrated in Fig. 1, and are elaborated on in later 

sections. 

By definition, PBEE is based on achieving desired performance targets. The latter are of concern to 

society as a whole or to specific groups or individual owners. Performance targets are of the type 

expressed in the first column of Fig. 1. The metrics of specific interest in the PEER research are life 

safety, dollar losses, and downtime (or loss of function), and it is postulated that a performance target 

can be expressed in terms of a quantifiable entity and, for instance, its annual probability of 

exceedance. The quantifiable entities on which performance assessment is based, are referred to as 

Decision Variables, DVs. For instance, X s (y), the mean annual frequency 1 (MAP) of the loss 

exceeding y dollars, could be the basis for a performance target. Alternatively, a performance target 

could be that the expected annual loss should be less than x dollars. In the assessment methodology 

1 The MAP is approximately equal to the annual probability for the small values of interest here.

175 

the key issue is to identify and quantify, with due consideration to all important uncertainties, decision 

variables of primary interest to the decision makers. Examples of DVs of interest are number of 

casualties, $ losses, and length of downtime, see second column of Fig. 1. 

Perform. 

Targets 

•Collapse & 

Life safety 

P f

176 

(probability of being in or exceeding a specific damage state, given a value of EDP). The DMs include, 

for example, descriptions of necessary repairs to structural or nonstructural components. If the 

fragility functions for all relevant damage states of all relevant components are known, the DVs of 

interest can be evaluated either directly or by means of cost functions that relate the damage states to 

repair/replacement costs. The result of this last operation is G(DV\DM), the conditional probability 

that DV exceeds a specified value, given a particular value of DM. 

These steps, which form the basis of performance assessment, can be expressed in the following 

equation for a desired realization of the DV, such as the MAP of the DV, A(D V), m accordance with the 

total probability theorem: 

This equation, which often is referred to as the framework equation for performance assessment, 

suggests a generic structure for coordinating, combining and assessing the many considerations 

implicit in performance-based seismic assessment. Inspection of Eq. (1) reveals that it "de-constructs" 

the assessment problem into the four basic elements of hazard analysis, demand prediction, modeling 

of damage states, and failure or loss estimation, by introduction of the three "intermediate variables", 

DM, EDP, and IM. Then it re-couples the elements via integration over all levels of the selected 

intermediate variables. This integration implies that in principle one must assess the conditional 

probabilities G(EDM\IM), G(DM\EDP) and G(DV\DM) parametrically over a suitable range of DM, 

EDP, and IM levels. 

In the form written, the assumption is that appropriate intermittent variables (DMs and EDPs) are 

chosen such the conditioning information need not be "carried forward" (e.g., given EDP, the DMs 

(and DVs) are conditionally independent of /M, otherwise IM should appear after the EDP in the first 

factor.) So, for example, the EDPs should be selected so that the DMs (and DVs) do not also vary with 

intensity, once the EDP is specified. Similarly one should chose the intensity measures (IM) so that, 

once it is given, the dynamic response (EDP) is not also further influenced by, say, magnitude or 

distance (which have already been integrated into the determination of A(7MJ). This condition can 

make selection of records a challenge. 

Equation 1 may take on various forms, depending on the purpose and the decision variable of interest. 

For instance, Miranda and Aslani (2002) use the following form to compute the expected annual loss 

for component;, £"[£,], if the damage to the component can be expressed by m damage states: 

] = £ J £[L y | DM = dm,] P(DM = dm, \ EDP J = edp) P(EDP J > edp \ IM = im) dv(1M \E DPd m 

dIM 

(D 

The expected loss for the building is then computed as the sum of the expected damages for the 

individual components. Clearly, expected losses do not provide a comprehensive probabilistic 

performance assessment, and the assumption that component losses can be computed independently 

and then summed over all components is a simplification, but Miranda's approach is the first 

comprehensive implementation of the framework equation and constitutes a significant step forward in 

performance assessment.

177 

SPECIFIC ISSUES AND CHALLENGES IN PERFORMANCE ASSESSMENT 

Each step summarized in the previous section has many issues that need much discussion and further 

research. The following sections address a few of these issues and point out some of the challenges 

that have to be confronted in order to permit a consistent implementation of the framework equation. 

Because of space limitations, only a few specific issues concerned with IMs and EDPs are addressed. 

Intensity Measures, IM 

Intensity measures are quantities that describe the magnitude (M) and distance (R) dependence (other 

parameters such as fault mechanism also could be considered) of ground motions characteristics that 

significantly affect the upstream variables of the performance assessment approach. In the context of 

Eq. (1), this implies evaluation of the MAP of IMs through seismic hazard analysis. Of course, 

simplicity favors scalar measures, and in particular, measures for which hazard analysis results are 

available. Choosing, for example, PGA for the IM may be initially appealing, but it implies that the 

distribution G(EDP\IM) may have a very broad variability. This in turn means that it will require a 

large sample of records and nonlinear analyses to estimate G(EDP\IM) with sufficient confidence. A 

well selected spectral ordinate (e.g., S a at the fundamental period T ; ) is an improvement over PGA and 

has been the popular choice in the recent past. It is a matter of debate whether S a (Tj) is indeed the 

"best" choice ("best" implies a balance between simplicity and accuracy). 

S a (Tj") does not account for the frequency content at T y T 17 which dominates higher mode effects (T < 

Tj) and period elongation effects for inelastic systems (T > T t ). Again, this may lead to G(EDP\IM) 

distributions with a very broad variability. This is illustrated in Fig. 2, which shows normalized 

incremental dynamic analyses (IDAs), together with statistical values, for a 6-story frame structure 

subjected to 40 ground motions that are scaled so that the S a at the fundamental period (T, = 0.6 sec.) is 

the same. Scaling records to a common S a at T, (which is implied by using S a (T|) as the IM) results in 

a median spectrum that resembles a typical spectrum for ordinary ground motions, but results in large 

variability in spectral ordinates at periods even very close to T,, see Fig. 3. 

NORMALIZED MAXIMUM STORY DRIFT 

N-6, T,=0 6, £=0 05, Peak-oriented model 9=0.010, BH, K,,S,, LMSR-N 

ELASTIC STRENGTH DEMAND SPECTRA 

Scaled Records (T=OJ s), LMSR, § = 0.05 

1 2 3 4 5 

Norm. Max. Story Drift Over Height, 9ull)l)

178 

start to dominate the hazard at long return period hazards. Several efforts are in progress to find 

improved IMs, ranging from the use of combinations of spectral values at modal periods (Luco and 

Cornell, 2001) to the use of vectors that incorporate near-fault parameters such as a an equivalent pulse 

period. 

Engineering Demand Parameters, EDPs 

Engineering demand parameters are the product of response prediction, most appropriately from 

inelastic dynamic analysis that considers the soil-foundation-structure system resting on top of bedrock. 

The challenge of formulating complete system models, which should incorporate modeling 

uncertainties and should account for all important uncertainties inherent in geotechnical and structural 

material, component, and system properties, will remain a challenge for many years to come. 

Relevant EDPs depend on the performance target and the type of system of interest. They include 

story drifts, component inelastic deformations, floor accelerations and velocities, but also cumulative 

damage terms such as hysteretic energy dissipation. Once identified, they can be computed from 

different procedures such as the by now widely employed incremental dynamic analysis (IDA) 

procedure. In this procedure, the soil-foundation-structure system is subjected to a ground motion 

whose intensity is incremented after each inelastic dynamic analysis. The result is a curve that shows 

the EDP plotted against the M used to control the increment of the ground motion. IDAs can be 

earned out for a sufficiently large number of ground motions to perform statistical evaluation of the 

results. This implies that for a given value of IM, the median value and a measure of dispersion (e.g. 

84 th percentile) of the response EDP values are evaluated, with results as shown in the right part of Fig. 

4. This provides the information for dG(EDP\IM) of Eq. (1). 

For use in Eq. (1), dG(EDP\IM) must be evaluated for the full range of 1M that contributes 

significantly to the final value of DV. The IDA, however, has a limited range of applicability because 

the ground motion frequency characteristics change with magnitude, particularly for long return period 

events that may be dominated by near-fault ground motions. Thus, caution must be exercised in 

defining the range of applicability of an IDA and the associated values of dG(EDP\IM). 

Presuming that the IDA curves and their statistics are valid for the full range of interest, the 

information shown in Fig. 4 can be used to develop a hazard curve for the EDP, using the equation 

*£OP 00 = P[^DP > y \ IM = x] | dX m (x) | (3) 

This hazard curve can be obtained from numerical integration of results of the type shown in Fig. 4, or 

through a formulation described in Luco and Cornell (1998), which results in the following expression 

for the mean annual frequency of EDP exceeding a value y: 

X EDP (y) = P(EDP > y} = k t ylar k exp-^£DP|(M (4) 

This equation holds if the IM hazard can be described by the widely used equation 

l ai (x) = P[lM> X ] = k e x' k (5) 

and the following relationship can be fit locally (around the return period of primary interest) to the 

median EDP - IM data:

179 

= a(lM) b 

(6) 

A typical result of an EDP hazard curve is shown in Figure 5. 

' * Incremental Dynamic Analysis 

AVERAGE DRIFT HAZARD CURVE-T^l.8 sec. 

N=9, ysfl 10, §=0 05, Peak oriented model, 6*0 060, BH. K,, Si, LMSR 

1 • 

J# ft fl1 • 

V 

\ | — Hazard Curve Obtained from Numerical Integration 

1 N xJ _i_ ' 

f 

EDP (e.g., max. interstory dnft) 

Figure 4. Incremental Analyses and Their Use in 

Probabilistic Seismic Demand Analysis 

A nnrvi • 

0 0.005 0.01 0.015 0.02 O.I 

Average of Maximum Story Drifts, 9 

Figure 5. Example of a Hazard Curve for 

an Engineering Demand Parameter 

A special problem arises when the limit state of collapse is of interest. Collapse is caused by 

detenoration in strength, and for this reason response prediction has to resort to the use of deteriorating 

systems. Models for such systems are available, see Fig 6 (Ibarra et al., 2002). The effect of 

detenoration on the seismic response is illustrated in Fig. 7, which shows IDA curves for a 9-story 

frame without and with deterioration. The implication of the IDA turning horizontal is that an 

infinitesimal increase in ground motion intensity causes a very large increase in the maximum story 

drift. This implies dynamic instability (global collapse), which cannot be evaluated by means of 

nondeterioratmg systems. 

Pinching Hysteretic Model, Hahl-Column 1,P-A=0, 

a=Q.10,ac,p=-024,ic=fl.5,Yit=100,Y,=50,Y«=30,YB=40,5c=2.35y 

MAX. STORY DUCTILITY vs. NORM. STRENGTH 

N=», T,=0.9, §=O.OS, K,, S,, BH, 9=0.015, Peak-Oriented Model. LP89svl 

s 10 is 

Maximum Story Ductility Over the Height, fi^,. 

Figure 6. Comparison Between Exp. Results and 

Analytical Predictions for a Deteriorating System 

Figure 7. IDA for a MDOF Structure Without 

and With Deteriorating Elements 

The use of deteriorating systems permits a direct evaluation of the probability of collapse in 

accordance with Eq. (1), without the help of an intermittent DM, Collapse statistics can be performed 

on the last point of the ID As, see Fig. 8, which provides statistical information in terms of a suitable 

intensity measure. In Fig. 8, S a (Tj) is used as the IM (j is the base shear coefficient V/W). 

Alternatively, collapse fragility curves of the type shown in Fig. 9 can be developed, which permits an 

evaluation of the probability of collapse for different deterioration properties (Ibarra et al., 2002).

180 

MAX STORY DUCTILITY vs NORM. STRENGTH 

Ns9,T>0.9,5=0 05, KI, S,, BH, 9aO 015, Peak-Oriented Model, 

tt,=0 05,5^5,*4, ae= 0 10. y.aS Yc»8 . rk=8 , Y.=» , A.=0, LMSR 

(Sj/Tl vs PROBABILITY OF COLLAPSE, 1=0.5 s 

Peak Oriented Model, LMSR-N, £=5%, P-A='0.1N' 

o,=0.05, oc^Var, 5,y6y=Var, Ys.c^=Var 

I «1 

/ — Ywjyr'SO, 5^=4, Oe.-0.30 

YIAM* 100, 5^=4, a^-O.lG 

/ __ Y.tM-Inf, 5^=6, a^-0.10 

10 20 30 

Maximum Story Ductility Over the Height, li5JB«, 

Figure 8. ID As for a Deteriorating System and 

Collapse Statistics 

Figure 9. Collapse Fragility Curves for Systems 

with Various Deterioration Properties; T = 0.5 sec. 

For performance targets associated with monetary losses and downtime, the need exists (at least at this 

time) to use DMs as intermittent variables m order to close the loop from EDPs to DVs. The paper by 

Miranda and Aslani (2002) discusses the use of DMs in the context of loss estimation. This paper, and 

many others that cannot be quoted because of space constraints, provide an overview of the status quo 

of the research effort on performance-based earthquake engineering carried out by the PEER Center. 

ACKNOWLEDGEMENTS 

The work summarized in this paper is part of a multi-year research effort on seismic performance 

assessment. Many individuals have contributed to this research, which is a multi-institutional and 

multi-disciplinary effort directed and sponsored by the Pacific Earthquake Engineering Research 

(PEER) Center. Funding for the Center is provided by the National Science Foundation, by various 

agencies of the State of California, and by industry. This support is gratefully acknowledged. 

REFERENCES 

Cornell, A. and Krawinkler, H. (2000). Progress and Challenges in Seismic Performance Assessment. 

PEER News, April 2000. 

Ibarra, L, Medina, R., and Krawinkler, H. (2002). Collapse assessment of deteriorating SDOF 

systems. Proceedings of the 12 th European Conference on Earthquake Engineering, London. 

Krawinkler, H. (1997). Research issues in performance based seismic engineering. Seismic Design 

Methodologies for the Next Generation of Codes, Fajfar, P., and Krawinkler, H., Editors, Balkema, 

Rotterdam, 47-58. 

Krawinkler, H., Medina, R., and Alavi, B. (2003). Seismic drift and ductility demands and their 

dependence on ground motions. Accepted for publication in the Journal of Engineering Structures. 

Luco, N. and Cornell, C.A. (2002). Structure-specific scalar intensity measures for near-source and 

ordinary earthquake ground motions. Earthquake Engineering & Structural Dynamics. 

Miranda, E. And Aslani, H. (2002). Probabilistic estimation of building response for loss estimation. 

Proceedings, ICANCEER 2002.




THE REAL-TIME EARTHQUAKE INFORMATION SYSTEM 

IN THE NORTHERN KYUSHU, JAPAN WITH A SMALL 

SCALE GEOINFORMATION DATABASE FOR SEISMIC 

INTENSITY ESTIMATION 

Hidemori Narahashi 

Assoc. Professor 

Department of Architecture 

Kyushu Sangyo University, Japan 

ABSTRACT 

The system to estimate seismic intensity distribution quickly after earthquakes has been 

developed since 1999, regarding to a local densely populated area of 20x20 square miles in 

the northern Kyushu, Japan. It is called KSU Real Time Earthquake Information System. The 

system consists of two subsystems, those are the seismic observation and data acquisition 

subsystem and seismic intensity estimation subsystem. In this paper the performance and 

goals of the system are outlined and matters to be improved are discussed. 

The two properly oriented observation stations and the eight K-NET stations are deployed in 

the area to observe three components ground acceleration for twenty-four hours. Soon after 

two or more stations will have been triggered with an event the center collects data from the 

other stations via public telephone line. It takes about forty minutes to receive a data set from 

all of the stations. Since April 2002 additional seismic intensity data at the 108 stations of 

Fukuoka Prefecture are transferred very promptly via ISDN telephone line to the data 

acquisition subsystem. 

After the data acquisition subsystem receives a data set from the ten ground motion 

observation stations it is about fifteen minutes until the latter subsystem estimates earthquake 

ground motion for more than ten thousands of blocks of 250x250 square meters in the area. 

The topography of each block is assigned on the GIS database of this subsystem in order to 

estimate JMA seismic intensity based on the data from triggered observation stations. 

Eventually the system gives 16 times as dense seismic intensity estimation results than the 

DIS system of the National Land Agency, Japan. 

The system is characterized as an earthquake information system for the local governments of 

the area. The procedures for interpolating the base rock motion from not too many ground 

surface data, as well as the procedure for evaluating ground motion amplification effect with 

the surface topography, should be improved for practical use. In addition, the third subsystem 

should be prepared to open the simulated results for local governments and the community of 

The area.

Proceedings of the International Conference on \ 33 



DEVELOPMENT OF VIRTUAL EMERGENCY 

RESPONSE NETWORK AND APPLICATION 

Sunao Nishimura 

AJBS Consulting, EQE Japan Division 

Tokyo, Japan 

INTRODUCTION 

In a catastrophe earthquake, individuals, companies and organizations are required to respond to 

protect themselves and minimize following risks, because full public assistance is not expected due to 

simultaneous accidents in everywhere in impacted areas, and communication problems. 

Various types of emergency information management platforms and systems have been developed with 

enhanced information technology these days. Most of those products, however, works independently 

apart from practical use, and enough attention is not paid to the integration of different functions with 

proper assumptions on what is likely to occur and how personnel need to respond to varying situations. 

We have developed an integrated emergency information management program for a complex area, the 

Hammi Island Triton Square (HITS), in Tokyo, to protect facilities, employees, visitors and tenants' 

businesses. The program includes a networked information system called Virtual Emergency 

Response Network (VERN) and emergency response planning for optimized response organization. 

The system attempts to encompass wide-ranging emergency information from earthquake intensity and 

prompt damage assessment on/around the site, through response and recovery information 

indispensable for proper decision-making procedure during a disaster. This paper describes concept 

of the network and actual application to the HITS complex. 

VIRTUAL EMERGENCY RESPONSE NETWORK: VERN 

In order to make a community more self-sufficient during an earthquake disaster by expediting the 

collection and processing of emergency information and providing it real-time and under various 

formats to emergency managers, tenants, shoppers, visitors and employees, we developed a concept of 

VERN which is defined as a virtual network where information gathers and responders discuss 

situations to initiate prompt response operations. On this network, authorized responders 

communicate and make decisions based on information that is placed by responders or that is 

estimated and recommended by the system. Figure 1 shows overall structure of information which is 

treated on the network, and associated response expected to be determined based on the information.

184 

Estimated Damage 

Information 

Decision Support 

Information 

EQ Info^ 

Provider** 

Shaking Intensity 

Distribution 

(PGV PGA) 

Initial 

Response 

Figure 1. VERN Emergency Information Structure 

VERN consists of four elements to cover four different types of information management: 1) 

Earthquake Monitoring System, 2) Early Post Earthquake Damage Estimation Tool, 3) Emergency 

Information System, and 4) Decision Tree. Functions of these components are described as follows: 

Earthquake Monitoring System: 

Monitors earthquake occurrence information released by observing agents such as Japan 

Meteorological Agency (JMA), and calculates ground motion distribution. The information includes 

epicentral information and shaking intensity recorded at local sensors. The VERN system directly 

connected to JMA catch the information within 10-30 minutes and suggests initial response level 

associated with severity of earthquake at site. 

Early Post Earthquake Damage Assessment Tool: EPEDAT. 

Triggered by input from the Earthquake Monitoring System, initiates damage estimation calculation 

automatically, and place results into database. This system is known as EPEDAT which is originally 

developed for the state of California by EQE. 

Emergency Information System: 

All information related with status (actual damage), and emergency staffing and operations is entered 

through forms, and reports are generated by functions of sections of emergency organization. 

Emergency response staffs are able to obtain necessary information as well as to input.

185 

Decision Making Support Tool' 

Upon reception of emergency information created and stored by above system, the Decision Making 

Support Tool suggests appropnate response to emergency response organization. Judgment criteria 

are determined based on specific situations that govern following actions. 

These components are then integrated on a unified platform so that consistent information traffic, 

which is expected to help responders to manage various situations efficiently, is generated and 

controlled. It is very important to retain functional interface between "information" and actual 

"response". Figure 2 shows general flowchart of disaster management during/after an earthquake. 

Efficient response activities should be based on appropnate information, and VERN is a virtual space 

that creates communication platform. 

INTERACTION 

Figure 2. General Response Operation Flowchart 

APPLICATION TO HITS RESPONSE ORGANIZTION 

VERN concept has been applied to a complex called Harumi Island Tnton Square (HITS). The HITS 

is a complex facility built under the urban redevelopment plan by Tokyo Metropolitan Government. 

The facility contains high-rise residential apartments, commercial shopping area, concert hall, and 

office buildings including 3 33-40 stones high-rise buildings. The total area of the site is 

approximately 84,000m 2 . It is estimated that daytime population of all business tenants is 18,000 to 

19,000, and a number of visitors and shoppers visit this facility. Earthquake disaster management

186 

program was introduced to business and commercial district in order to secure life safety and minimize 

impacts to building properties and businesses. Residential district is excluded in this program. 

1 Central Emergency Operation Center (CEOC) and 5 sub Emergency Operation Centers (EOC) are 

assigned to major building units, and VERN system was networked for responders to establish local 

communication within the HITS site (see Figure 3). CEOC is in charge of building-to-building 

coordination to support activities, as well as manage common spaces. All information, including 

earthquake shaking intensity and epicenter sent from JMA and on-site building floor response records, 

are stored in a server at CEOC, and shared as appropriate. Authorized responders of HITS 

Emergency Organization initiates response operation, update status information, and implement 

response activities under HITS Emergency Response Plan. Critical decisions in regard with 

operation priority, objectives, and staff allocation are discussed and determined utilizing the network. 

VERN 

Floor 

Response 

Office 

Tower Y 

Central 

Emergency 

Operation 

Center 

Figure 3. VERN Network Applied for HITS 

CONCLUSION 

VERN could show a clue to establish a uniform emergency response mechanism and organization to 

minimize complexity of information, and to maximize self-sufficiency capability. This concept can 

be applied to various size of disaster management community which would have existing systems to 

integrate on a common platform. The key to fully utilize this system is to structure operational 

response scenarios per types and requirement of community and facilities, and decision-making 

support tool, as a cornerstone, is expected to help expedite efficient activities by upgrading the level of 

its density and detail.

Proceedings of the International Conference on -, 07 



STRONG MOTION DATABASE AND ANALYSIS METHOD IN 

MAINLAND CHINA 

Haiying YU 

(Institute of Engineering Mechanics, China Seismological Bureau, 

Phone:4-86-45l-6652604,Fax:-h86-451-6664755, Email:yhy6670420@mail.china.com) 

ABSTRACT 

Strong motion records in China have been analyzed and processed using routine processing 

methods and edited by uniform format. Strong motion database and a set of computer software 

used for routine processing of strong motion records have been established, and 3255 

accelerograms obtained during 1968-2001 have been collected in the database. Based on the 

analysis of low frequency errors in acceelerograms recorded by digital accelerographs, a 

correction method is proposed. The results show that the method can be used for data processing 

of accelerograms of a variety of digital strong motion accelerographs, and it is possible to 

acquire the reliable long-period information of strong ground motion up to about 10 seconds. 

Finally, The accelerograms obtained in SMART-1 array of Taiwan are processed and the 

features of long-period response spectra are analyzed. 

Keywords: strong motion database, low frequency error correction, long period response spectra 

1. STRONG MOTION DATABASE IN MAINLAND CHINA 

The first experimental strong motion observation station in Mainland China was established at 

the Xinfengjiang reservoir dam by IEM in 1962. Up to 2000, 283 strong motion stations or 

arrays with 366 accelerographs have been deployed in the main seismic areas of Mainland China. 

After 1966 Xingtai Earthquake, IEM deployed some mobile strong motion stations. Since then, 

mobile strong motion observations are executed 71 times with 335 accelerographs in Mainland 

China. 3255 valuable accelerograms have been accumulated in China during 1968-2001, in 

which, 206 accelerograms with the peak value larger than 100 gal, 59 accelerograms larger than 

200 gal (Figure 1); the maximum horizontal peak ground acceleration is 541.7 gal, and the

188 

maximum vertical peak ground acceleration is 528.2 gal The major strong motion records are 

listed in Table 1. 

TABLE 1 

THE MAJOR STRONG MOTION ACCELEROGRAMS IN MAINLAND CHINA 

No. 

1 

2 

3 

4 

5 

6 

7 

8 

9 

10 

11 

12 

13 

14 

15 

16 

17 

18 

19 

20 

21 

22 

23 

24 

25 

26 

27 

28 

29 

30 


(Array)Name 

Tangshan 

Tangshan 

Arrays 

Guye 

Zhangbei 

Haicheng 

Tonghai 

Lonsling 

Gengma 

Wuding 

Lijiang 

Yaoan 

Shidian 

YongSeng 

Wuqia 

Wushi 

Wusu 

Atushi 

Jiashi 

Songping 

ZhangYe 

Sunan 

JiayuGuan 

Linze 

Gonghe 

Datong 

Huangbizhua 

ng Reservoir 

Arrays 

Xinfengjiang 

Reservoir 

Arrays 

Guanting 

Reservoir 

Arrays 

Douhe 

Reservoir 

Arrays 

Longyangxia 

Reservoir 

Date of Earthquake 

76.7.28—76.11.15 

82.7—84.12 

95.10.6 

98.1.10 

76.2.8—76.2.28 

70.1.14—70.9.4 

76.6.5—76.6.29 

88.11.6—88.12.6 

95.10.25—95.10.27 

95 1.31—96.2.9 

00.1.15—00.1.31 

014.10—01.5.19 

01.10.27 

85.8.28—85.9.20 

87.1.24—90.10.25 

95.5.2—96.3.22 

95.5.2—96.3.22 

96.3.19—98.8.27 

76.8.16—76.9.2 

87.9.12 

88.11.22—88.12.26 

92.1.12 

87.10.25—88.12.26 

94.2.16—94.9.24 

89.10.18 

68.7.25—74.6.6 

68.10.21—78.9.2 

94.12.23 

79.8.25—80.4.16 

90.4.26—94.10.10 

Magnitude 

3.5—7.8 

2.3—5.7 

5.9 

6.4 

3.0—5.4 

2.6—5.7 

2.3—6.2 

2.6—7.6 

4.1—5.3 

7.0 

2.7—6.5 

1.5—5.9 

4.1—6.0 

3.0—6.3 

1.9—6.4 

3.3—6.9 

3.3—6.9 

3.6—6.9 

4.2—7.2 

4.5 

2.7—5.7 

5.9 

5.0—5.1 

5.2—5.6 

5.7 

2.4—5.2 

2.6—5.3 

3.8 

3.6—4.0 

1.9—6.9 

Numbers of 

Event 

38 

52 

1 

1 

24 

11 

11 

92 

10 

13 

9 

41 

2 

18 

23 

7 

7 

14 

9 

1 

5 

1 

2 

3 

1 

6 

24 

1 

3 

3 

Numbers of 

Accelerograms 

267 

555 

24 

3 

366 

87 

87 

406 

24 

54 

38 

132 

6 

52 

55 

24 

24 

420 

36 

3 

33 

6 

6 

8 

20 

70 

345 

10 

25 

50

189 

31 

Array 

Liujiaxia 

Reservoir 

Arrays 

Total 

95.7.22 

68.7.25—01.10.27 

5.8 

1.5—7.8 

1 

434 

19 

3255 

In Mainland China, Strong motion database have been established by leading of Professor Li-Li 

XIE from 1980' s. In the past many years, more than 3000 strong motion records of Mainland 

China have been collected in the database. The above strong motion records have been processed 

using routine methods by IEM. Strong Motion Data Processing Center (Figure 2) established in 

IEM takes the responsibility for the processing and publication of the strong motion data 

obtained by the various units of China Seismological Bureau. The computer management system 

for strong motion database has been established, and the service for strong motion data on 

Internet will soon be brought about. Besides, through data exchange, more and more external 

strong motion records have been collected. 

Figure 1: Statistics of accelerograms 

Figure2: Strong motion data processing center in IEM 

In near future, the strong motion data of Mainland China will be displayed on web site. Figure 3 

and 4 is strong motion database web site system and home page respectively in IEM. The home 

page includes: professional introduction, about us, login, other useful web site and so on, and this 

web site has five access methods: earthquake list, station list, accelerograms list, basic search 

page and advance search page. Strong motion data is downloaded using these methods. The 

researchers of the world can easily use them.

190 

Figure 3: strong motion database web site system 

Figure 4: strong motion database web site home page 

Up to now, 11 volumes of Report on Strong Earthquake Motion Records and 1 volume of Report 

of Long-period Response Spectrum have been published by EEM. In addition, China institute of 

water resources and hydropower research has edited and published 1 volume of Important Data 

on Strong Motion and Analysis in China Hydraulic Structures, including important strong 

motion records that recorded at Xinfengjiang reservoir dam, Huangbizhuang reservoir dam and 

so on. 

Strong motion records of China have been used in draw up the various design codes of structures, 

seismic ground motion parameter zonation map, probabilistic seismic hazard analysis, 

earthquake safety evaluation, seismic design of structures, studies on characteristics of 

earthquake ground motion and attenuation lows of ground motion. For example, the 

accelerograms recorded at Qianan station during the Tangshan aftershock of August 31, 

1976(m=5.8) and the accelerograms recorded at Tianjing hospital station during the Ninghe 

earthquake of November 15, 1976(m=7.1) are used usually as the input seismic wave in 

vibrating table experiment and seismic calculation of structures. Figure 5 to 8 is typical strong 

motion records of Mainland China. 

AUG.31 1976 11:253 TANGSHAN AFTERSHOCK HEBEI M-5.8 3TAT\ONNO.fM0303101ANANL*NHE9RlDGE 

1NTTYPE. : 

'ALLY SPACED INTERVALS OF 0.01 SEC 

NOV.l 5.1571 21 :S3BTVI TANGSHAN AFTERSHOCK HEBEI M-6 9 • 1INHO 

INSTRUMENT-TYPE RDZ1 IF NO OF POINTS. 1 91 2 EQUALLY SEC 

SPACED INTERVALS OF 0.01 

EFFECTIVE FREQUENCY BAND 30 -3500 HZ 

TH7-m COMP. EW Amox-97 356 (cra/s/s) 

TSJ7-H500MP SN Amax-132.392 (ci 

TS27-146COMP.UO Amax=50.493(cm/5/s) 

TS38-21 - 4 COMP UD Amm-aHO mm- (on/s/s) onss 

Figure5: Accelerogram recorded at Qianan Station 

in Taneshan Aftershock of August 31. 1976 

Figure6: Accelerogram recorded at Tianjing Hospital 

Station in Ninghe Earthquake of November 15, 1976

191 

IV STATION YM 

kCED INTERVALS OF C 

•nro&incQMPsa 

SHOWN EARTHQUAKE 3HMAN OMAG 5 3(Msl STATION TAPING 

U^ -'gg'tf K***'**® "TSmSOF 001 SEC 

SOB-DOB CQMPV Am

192 

2.2 Low frequency errors correction method of accelerograms recorded by digital 

accelerographs 

After comparing the results of different methods, the correction method of low frequency errors 

of accelerograms recorded by digital accelerographs is proposed, as follows: 

1. Adjust zero baseline of original accelerogram by subtracting the average value of preevent 

records, and subtracting, the subtract average value of whole length of accelerogram or use 

in the least squares method if there is no pre-event records can be available. 

2. Filtering the accelerograms by Butterworth digital filter, and the cut-off frequency can be 

determined by Fourier spectrum analyses for seismic records and pre-event noises records. If 

there is no sufficient long pre-event records can be used, we can select one or several cut-off 

frequency according to the experiences to trial, and a suitable cut-off frequency can be 

determined according to if the drift of zero baseline of displacement curve is eliminated or not. 

3. Calculating acceleration response spectrum, velocity response spectrum and displacement 

response spectrum of filtered records. Integrating the velocity and displacement time histories, 

adjusting the zero baseline of displacement time history using least squares method in order to 

eliminate linear trend. 

2.3 Features of Long-Period Spectrum of SMART-1 Array Records 

2.3.1 Record and disposal of digital strong motion Accelerograph 

236 accelerograms recorded by SMART-1 array in 8 earthquakes with magnitudes above 6 are 

selected for analyzing the long period component characteristics of seismic ground motion on the 

same category of site in a small region. Table2 list the data of the 8 earthquake. 

All of the epicentral distances in Table2 are the epicentral distances of central point (COO), the 

maximum distance between various stations is only 4km, so the epicentral distances of each 

station can be considered as same in these earthquakes, except No.7 earthquake. All records are 

processed using the method described above. 

2.3.2 Response spectra characteristics of accelerograms obtained in an earthquake 

Figure 11 and Figure 12 show the dynamic amplification factor /3 curves of accelerograms 

recorded at different stations during two earthquakes (No,4 and No.8) respectively (damping 

ratio is 0.05). In the Figures, the symbols "o" and "*" represent average values of /3 and a 

standard deviation respectively. It can be seen that in short period below 4s, although /j curves 

are similar for most of stations, some f$ curves are very different and the standard deviations are 

large. For example, in case of No.4 earthquake, the corresponding periods for peak values of p 

curves are Is or so for most stations, and are 0.3~0.7s for other stations (Figure6). The 

epicentral distances of all stations are almost same in the two earthquakes. So, It shows that the 

acceleration response spectra, in short period part are quite different even if the accalerograms 

are recorded at same category of sites in an earthquake. However, for the accelerograms obtained

193 

at different stations, the differences of p curves in long period part above 4s is small and the 

attenuate rate is almost same. It shows that the long period component of seismic ground motion 

are almost same at same category of site with same epicentral distance in an earthquake and it is 

almost not influenced by the small differences between the local site conditions of various 

stations and other random factors. 

TABLE 2 

EARTHQUAKE DATA 

No. 

1 

2 

3 

4 

5 

6 

7 

8 

Date of earthquake 

81.01.29 

82.12.27 

83.05.10 

83.06.24 

86.01.16 

86.05.20 

86.07.30 

86.11.14 

Magnitude 

6.9 

6.9 

6.3 

7.2 

6.5 

6.9 

6.2 

7.0 

Epicentral 

distance ( km ) 

31 

118 

35 

116 

22 

36 

6 

79 

Period (Second) 

Figure 11: Dynamic magnification coefficient ft of the 

accelerograms recorded at different stations in No.4 

earthquake (damping ratio is 0,05) 

Figure 12: Dynamic magnification coefficient /3 of the 

accelerograms recorded at different stations in No.8 

earthquake (damping ratio is 0.05) 

2.3.3 Response spectral characteristics of accelerograms recorded at a station 

Figures 13 and 14 show the /3 curves of accelerograms recorded in 8 events at Station COO and 

1-03, respectively (damping ratio is 0.05 ), the magnitudes of these 8 earthquakes are all above 6, 

and the epicentral distances are very different from 6km to 118km. It can be seen that the periods 

corresponding to the peaks of response spectra are quite different, and the/3 values in long 

period part of No.8 earthquake (dashed line in Figures 13 and 14) is much larger than that of any 

other earthquakes.

194 

Panod (Second) 

Figure 13: Dynamic magnification coefficient /3 of Figurel4: Dynamic magnification coefficient 0 of 

accelerograms recorded by Station COO (damping ratio is 

0.05) 

accelerograms recorded by Station 1-03 (damping ratio is 

0.05) 

REFERENCES 

1. Li-Li Xie. (1985). The Main Features Of Data Processing Procedure For Strong Motion Records In China, 

Physics of the Earth And Planetary Interiors, 38, 134-143, Elsevier Science Publishers B.V., Amsterdam- 

Pnnted In the Netherlands. 

2. Li-Li Xie, Shabai Li, Jukang Qian, (1981). Study on the Instument Correction of Accelerograms Recorded 

by Accelerograph Coupled with Galvenometers, Journal of Earthquake Engineering And Engineering 

Vibration, 1: 106-Il6(ln Chinese) 

3. Li-Li Xie, Yongnian Zhou, Haiying Yu, Chengxiang Hu. (1991). Report of Long-period Response 

Spectrum(in Chinese), Seismological Press 

4. Yongnian Zhou, Haiying Yu. (1999). Low Frequency Error Correction of Accelerogram Recorded by 

Digital Accelerograph(in Chinese), Proceedings of workshop on key science problems and design theories 

research of large complicate structures, Dalian University of Technology Press,2000 

5. Haiying Yu, Yongnian Zhou. (1992). Design of Strong Motion Database Management System (in Chinese), 

Procedings of National workshop on Strong Motion Observation Technology (Vol.1) > Xiamen » 

Seismological Press 

6. Haiying Yu, Yongnian Zhou. (2001). Features of Long-Period Spectrum of SMART-1 Array Records(in 

Chinese) , Proceedings of workshop on key science problems and design theories research of large 

complicate structures (2000), Shanghai Tongji University Press 

7. Guangyi Gao, Haiying Yu, Shanyou Li. (2001). The strong-motion observation in the mainland of 

China(in Chinese), World information on Earthquake Engineering, Vol.17, No.4 

8. I wan. W. D., M. A. Moser, and C. Y. Peng. (1985) Some observations on strong-motion earthquake 

measurement using a digital accelerograph, BSSA, Vol. 75 

9. Bolt, B. A., Y. B. Tsai. K. Yen, and M. K. Hsu. (1982). Earthquake strong motions recorded by a large 

near-source array of digital seismographs. Earthq. Eng. Struct. Dyn., 10, 561-573. 

10. Hung-Chie Chiu. (1997) Stable baseline correction of digital strong motion data, BSSA, Vol. 87, No. 4 

ILRDobry, R.D.Borcherdt, C.B.Crouse, I.M.Idriss, W.BJoyner, G.R.Martin, M.S.Power, EJE.Rinne, and 

R.B.Seed. (2000). New site coefficients and site classification system used in recent building seismic code 

provisions, Earthquake Spectra, Vol.16




TOWARDS EARTHQUAKE HAZARD MITIGATION ON 

METROPOLIS - A PLATFORM FOR RISK COMMUNICATION 

Ping Zhu 1 , Masato Abe 2 and Junji Kiyono 3 

Research Institute of Science and Technology for Society, Japan Atomic Energy Research Institute 

18F, Atago Green Hills Mori Tower, 2-5-1 Atago, Minato-ku, Tokyo 105-6218, Japan 

2 Department of Civil Engineering, The University of Tokyo 

7-3-1 Kongo, Bunkyo-Ku, Tokyo 113-8656, Japan 

3 Graduate School of Civil Engineering, Kyoto University 

Yoshida Hon-machi, Sakyo, Kyoto 606-8501, Japan 

ABSTRACT 

Severe earthquakes occurred in regions with high dense population cause not only damage of 

structures but also social problems. Resent years, the demands are increasing for risk evaluation and 

communication of before, during and after earthquakes for disaster precaution, mitigation, rescue and 

recovery works. Upon discussing the challenges from both technology and social issues, this paper 

presents a sketch of a platform based on the Internet for a risk evaluation and disaster mitigation 

system for metropolis under severe earthquakes. 

INTRODUCTION 

Metropolises are vulnerable under attacks of severe earthquakes. Catastrophic economic losses around 

the world are huge and increasing in last decade. The 1995 earthquake in Kobe, Japan (Fig. 1) and the 

1999 earthquake in Chi Chi, Taiwan, for instance, caused losses about $120 and $10 billion 

respectively (Ref. 5). The disaster causes not only damage to buildings, structures but problems in 

society. Earthquake engineering professionals have been facing new challenges for helping solving 

economic and social problems with their knowledge. 

Upcoming earthquakes are still threatening metropolises in Japan (Fig. 2). Resent years, the demands 

are increasing for risk evaluation and communication of before, during and after earthquakes for 

disaster precaution, mitigation, rescue and recovery works.

196 

Upon discussing challenges on conducting seismic risk management and disaster mitigation for 

metropolis, from which a project the authors undertaking, this paper depicts a perspective for solving the 

problems using state-of-the-art computer technologies with a framework of a distributed application 

platform based on the Internet and Virtual Reality (VR) technology and presents some initial 

implementations. 

1 

Fig. 1 Collapse of elevated bridges in 1995 Kobe earthquake 

Hokkaldo-Nansel-OKI 

M7.8 1993 " 

* itokkBldo-Toho- 

P*iJwa:U994 

| KushlM-Oki.MT.8,1992 

Kobe (Hvngo-Ken- 

NambU).M7.2,1395 

Salirtkii-Haruka-Qkl, 

M7.5.1994 

: M7.5-8.mmi 

:B5»i [20 Years] 

•98% [30 Years] 

TwinEqs.MB-, 

M3D±10 

Nankai-Tinankal 

M7.5-Jokai 

I 37% [30 Years)) 

Fig. 2 Seismic action in recent years and near future in Japan

197 

PROBLEMS AND CHALLENGES 

Earthquakes, as a kind of natural catastrophe, are of low frequency but high consequence It is difficult to 

accumulate experiences This way, conventional aseisrmc design takes engineering codes as a main 

target, which means a design is created entirely on criteria that can meet the codes Present code design 

procedures, however, cannot give convincing answers to questions such as what might be the 

prospective damage sustained by a structure during an upcoming earthquake or the costs and benefits of 

earthquake protection 

Though proper handling of seismic behavior of a building or a structure is essential, it is still quite a local 

issue on concerning nsk management for metropolis During the 1995 Kobe earthquake the traffic 

service system in the emergency situation experienced substantial difficulties as considerable damage 

happened in elevated bridges (Fig 1 2) Collapsed buildings can also block roads Fires during the 

earthquake may rum several areas within a city for lack of emergent rescue (Fig 3) 

Fig. 3 Areas damaged by fires during the Kobe earthquake 

All the above-mentioned issues will cause losses of lives directly or indirectly Communication with the 

public becomes quite important for saving lives and keeping order of the society Dunng an earthquake 

city infrastructures such as power supply, electronic communication etc might be out of their functions 

To find efficient ways for communication should be a problem 

On conducting seismic nsk management and disaster mitigation for metropolis as problems concerned 

above, current solutions are insufficient in following aspects 

1) For structure seismic analysis, current solutions are weak in recreating and presenting real 

phenomena, such as damage and collapses of structures 

2) Research achievements and data, such as programs developed and data of structures created or 

obtained from design, construction and maintenance, mainly exist as isolated islands They are 

lack of integration and interaction Moreover, it is difficult to reuse the resources 

3) There is lack of concern in providing and distributing information, which is useful to and can be 

easily accessed by estate owners, administrators and the public Furthermore, less mechanism is 

established to listen and gather messages from the public

198 

It can be seen, to undertake seismic risk management and disaster mitigation for metropolis, a complex 

system is needed. To meet the needs of this task, the system should be built concerning functions as 

following: 

1) Data management, to collect and to keep updating data (geographical, structural data etc.) with 

unified format. 

2) Precise simulation of physical phenomena and consequences, including ground motion, 

structure seismic response, fire etc. in both macro and micro levels. 

3) Evaluation of damage and economic losses based on simulations from scenarios. 

4) Presentation and communication, to show results to different audiences (such as professionals, 

governors, estate owners and the public) with different methods; to collect information from 

the public. 

5) Integration of the system, to bring all the above together for a living system. 

6) Robustness, performance of the system and ways of communication during occurrences of 

disasters 

Since it is obviously difficult to achieve all the targets in one step, a practical way is to conduct and to 

accumulate effects on a frame of an integrated platform. 

A PERSPECTIVE SOLUTION 

Rapid growth computer technologies within the last decade, especially the Internet provide human 

beings new measures to deal with engineering and social problems which were hard to solve in 

traditional ways. A distributed application platform based on the Internet is a reasonable solution for 

hosting such a complex system with versatile resources. Furthermore, the Virtual Reality (VR) 

technology, which becomes more mature, can play an important role on performing 3D detailed 

presentation for various purposes. 

Concerning conducting seismic risk management and disaster mitigation for metropolis, basic targets 

of the distributed application platform are as follow: 

To develop a system for risk communication 

To establish a collaborating platform for researchers, professionals and governors 

To integrate resources and applications 

To communicate with the public reciprocally 

Fig. 4 illustrates concepts of this platform. Each functional parts are liked together without considering 

their physical locations. General audiences have their entry point at a web server with web browsers 

and client-end applets and applications; while professionals can utilize resources along with the 

platform, such as to access databases, to invoke dense computations and to cooperate with others.

199 

Database 

s*n/ar_- ^ 

" 

Application 

server 

PC c/uster for 

Communication with 

" " * other groups 

Fig. 4 A perspective view of the distributed platform 

The above picture shows still simple and basic ideas for conducting the system. To introduce the 

Internet into the system is not only for it can integrate resources and applications for a collaborative 

solution but also for it can be an effective way of reciprocal communication in a society. Traditional 

applications (such as a seismic analysis program) provide solutions, as shown in the upper part of Fig. 

5, by local computation, which means both user and machine are in a same physical location. To shift 

traditional applications into a web-oriented environment, generally speaking, interfaces should be 

separated from the main applications and the interface for output should be able to perform on the 

client end (Fig. 5). This makes it possible for people to access applications from the web. 

Local computing 

User 

- 

interface 

:> 

Application 

, Browser 

Web 

' Application 

User 

VRIVIL 

A simple internet based application 

Fig. 5 A simplified application prototype 

In a complex system, no single application can complete the task for final solutions. Fig. 6 shows a 

distributed environment based on the web. For communications between applications at server's part 

and also between client-end applets/applications and server applications, XML (Extensible Markup 

Language) will play an important role. XML has been claimed the most exciting and significant thing 

to hit the Internet since HTML (Cover R. (2001)). XML is a metamarkup language by providing

200 

arbitrary structures for one to make his/her own markup language For sharing data among applications 

for seismic analysis a special XML, which might be called Structure XML or StructureML, can be 

initiated for purposes of structure analyses, damage evaluation result presentation etc 

Users 

Browser 

Client-end < X/WL > Sen/er -e» A PP //car on 

app ^ -7 

{Extensible 

Language) 

Browser 

Client end < - ^ Sen/er ^ App/, car/on 

app ^ - 

Fig 6 Distributed applications on web 

SIMPLE IMPLEMENTATIONS WITH VR 

Tremendous efforts have been made to achieve good communication between human beings and 

computers since the initial work of computer graphics from later 60 Virtual Reality (VR), which 

becomes more popular recently, is among the most important technologies (Donale H, Baker MP 

(1997), McCarthy M, Descartes A (1998)) 

The term of Vutual Reality originally referred to Immersive Virtual Reality, which means that users 

become fully immersed in an artificial 3D world that is completed generated by computers Nowadays, 

the term VR is also used for applications without fully immersive The boundaries are becoming 

blurred 

A practical way to perform VR is using VRML, which stands for Virtual Reality Modeling Language 

(Ames A L el al (1997)) VRML is an open standard for 3D multimedia and shared virtual worlds on 

the Internet (ISO/TEC 14772-1 1997) (Cnspen B (2000)) It is the de facto standard tor sharing data 

between CAD (3D modeling) and animation programs VRML is a text based scene description 

language It is not a programming language but can have scripts, like JavaScript 

VRML can be used to present structures and to convey 3D scene from the Internet Fig 7 shows a 

suspension bndge authored with VRML The model bridge can be displayed interactively using a web 

browser with a VRML plug-in 

More practice of VR has been made based on a work of a general-purpose 3D dynamic analysis 

program for bridges The program performs seismic analysis of elevated badges based on precise 3D 

modeling (Zhu P et al (2002)) Fig 8 shows a three-span steel bndge analyzed by the program A new 

output interface of this program has been added to convert results of dynamic response of the bndge 

into VRML Animations of vibration can be seen through a VRML viewing tool, such as a web 

browser with VRML plug-in (Fig 9) Users can navigate through the scene to find more details of 

seismic response of the bndge, which are difficult to be acquainted by traditional 2D charts

201 

o 

Fig. 7 VRML authoring - modeling of a suspension bridge 

Fig. 8 Seismic analysis of a steel elevated bridge 

Fig. 9 Seismic response presented using VRML through Internet Explorer

202 

CONCLUSIONS 

Upon discussing demands on conducting seismic nsk management and disaster mitigation for 

metropolis, this paper depicts challenges faced by earthquake engineering professionals for helping 

solving economic and social problems. To meet this multidisciplinary task, the paper presents a 

perspective of solutions with a distributed application platform based on the Internet and VR 

technology. As a first step towards this platform, simple implementations with VR have also been 

demonstrated. Good results are obtained on promoting seismic response presentations using VRML. 

ACKNOWLEDGMENT 

Thanks are given to Ali Alaghehbandian, a master student at Dept. of Civil Engineering, Univ. of 

Tokyo, for his work (Fig. 7) of VRML authoring. 

REFERENCES 

1) Arnes A. L., Nadeau D. R. and Moreland J. L. (1997). VRML 2.0 Sourcebook, Second Edition, John 

Wiley & Sons, Inc., New York, USA. 

2} Cover R. (2001). The XML Cover Pages - Introducing the Extensible Markup Language (XML), 

http://xml.coverpages.org/xnillntro.html 

3) Crispen B. (2000). comp.lang.vrml FAQ, http://hiwaay.nct/~ crispen/vrml/faq.html 

4) Donale H., Baker MR (1997). Computer Graphics C Version (Second Edition), Prentice Hall, Inc. 

New York, US A. 

5) EERJ Endowment Fund White Paper. (2000). "Financial Management of Earthquake Risk." 

Earthquake Engineering Research Institute. 

6)McCarthy M., Descartes A. (1998). Reality Architecture Building 3D worlds with Java and VRML, 

Prentice Hall Europe, Great Britain. 

7) Zhu P., Abe M. and Fujino Y. (2002). "Modeling Three Dimensional Non-linear Seismic 

Performance of Elevated Bridges with Emphasis on Pounding of Girders." Earthquake Engineering 

and Structural Dynamics (under publication).

SMART MATERIALS AND 

SMART STRUCTURES




CRITICAL FACTORS FOR MAGNETORHEOLOGICAL 

FLUIDS IN CIVIL STRUCTURES 

J. David Carlson 

Lord Corporation, Materials Division 

406 Gregson Drive, Cary, North Carolina 27511, USA 

E-mail: jdcarlson@lord com 

ABSTRACT 

Magnetorheological (MR) fluid technology has progressed to the point where it is now 

routinely used on a large, commercial scale to enable semi-active control devices, 

particularly for automotive application. Experience in developing MR fluids for successful 

commercial application has shown that the greatest barriers to success of MR fluid 

technology are not the factors originally presumed to be most important in the laboratory and 

early development stages. While the most common responses to the question of what makes 

a good MR fluid are likely to be "high yield strength" or "non-settling", other factors have 

actually proven to be more critical. The present paper looks at conditions found in MR fluid 

devices operating in real-world applications where shear rates may exceed 10 5 sec" 1 and 

devices are called upon to operate for very long periods of time. The problem of "in-usethickening" 

wherein a MR fluid subjected to long-term use progressively thickens until it 

eventually becomes an unworkable paste is presented. The search for a solution to this 

heretofore unrecognized problem delayed the successful introduction of commercial MR 

fluid dampers for heavy-duty trucks for several years. Today, good MR fluids are able to 

operate for long periods with minimum in-use-thickening and settling.




IMPLEMENTATION OF MODAL CONTROL FOR 

SEISMICALLY EXCITED STRUCTURES USING MR DAMPER 

S. W. Cho, B. W. Kim, H. J. Jung, and I. W. Lee 

Department of Civil and Environmental Engineering, 

Korea Advanced Institute of Science and Technology, Daejeon, Korea 

ABSTRACT 

This paper proposes an implementation of modal control for seismically excited structures using 

magnetorheoiogical (MR) dampers. Modal control reshapes the motion of a structure by merely 

controlling a few selected vibration modes. Hence, a modal control scheme is more convenient to 

design the controller than other control algorithms. Although modal control has been investigated for 

the several decades, its potential for semi-active control, especially for the MR damper, has not been 

exploited. Thus, in order to study the effectiveness for MR damper system, a modal control scheme is 

implemented to seismically excited structures. A Kalman filter is included in a control scheme to 

estimate modal states from measurements by sensors. Moreover, a low-pass filter is applied to 

eliminate the spillover problem. The numerical results indicate that the motion of the structure is 

effectively suppressed by merely controlling a few lowest modes, although resulting responses varied 

greatly depending on the choice of measurements available and weightings. 

INTRODUCTION 

Magnetorheoiogical (MR) dampers are one of semi-active control devices, which use MR fluids to 

provide controllable damping forces. A number of control algorithms have been adopted for semiactive 

systems including the MR damper. Jansen and Dyke (2000) discussed recently proposed semi-

208 

active control algorithms including the decentralized bang-bang controller (MaClamroch and Gavin 

1995), the controller based on Lyapunov stability theory (Brogan 1991; Leitmann 1994), the clippedoptimal 

controller (Sack et al. 1994; Dyke 1996), the modulated homogeneous friction controller 

(Inaudi 1997), and the maximum energy dissipation algorithm. They, also, formulated these algorithms 

for use with MR dampers and evaluated and compared the performance of each algorithm. 

Modal control represents one control class, in which the motion of a structure is reshaped by merely 

controlling some selected vibration modes. Modal control is especially desirable for the vibration 

control of civil engineering structure may involve hundred or even thousand degrees of freedom, its 

vibration is usually dominated by the first few modes. Therefore, the motion of the structure can be 

effectively suppressed by merely controlling these few modes. A modal control scheme, which uses 

modal state estimation, is deskable (Meirovitch 1990). 

The purpose of this study is to implement modal control for seismically excited structures that use MR 

dampers and to compare the performance of the proposed method with that of other control algorithms 

previously studied. A modal control scheme with a Kalman filter and a low-pass filter is applied. A 

Kalman filter is included in a control scheme to estimate modal states from measurements by sensors. 

Three cases of the structural measurement are considered by a Kalman filter to verify the effect of each 

measurement; displacement, velocity, and acceleration, respectively. Moreover, a low-pass filter is 

applied to eliminate the spillover problem. 

MODAL CONTROL SCHEME FOR MR DAMPERS 

In this section, a modal control scheme with a Kalman filter and a low-pass filter is implemented to 

seismically excited structure. After the implementation of modal control scheme, numerical simulation 

is presented in a subsequent section for comparisons between control algorithms. 

Modal Control 

Consider a seismically excited structure controlled with m MR dampers. Assuming that the forces 

provided by the control devices are adequate to keep the response of the primary structure from exiting 

the linear region, the equations of motion can be written 

Mx(t} + C±(t} + Ex(t)=Jtf(t}-Mn g (1) 

where M, C and K are the nxn mass, damping, and stiffness matrices, respectively; :c is the n- 

dimensional vector of the relative displacements of the floors of the structure;/= [/;,/2,...,/ m ] r is the

209 

vector of measured control forces generated by m MR dampers; x g is ground acceleration; F is the 

column vector of ones; and A is the matrix determined by the placement of MR dampers in the 

structure. The displacement can be expressed as the linear combination 

(0=*»,r = J.2,...,n (2) 

where r\ r (r) is the r th modal displacement; fy r is the r th eigenvector; 0 is a eigenvector set; and 77 is a 

modal displacement vector. In modal control, only a limited number of lower modes are controlled. 

Hence, / controlled modes can be selected with I < n and the displacement may be partitioned into 

controlled and uncontrolled parts as jr(r) = x c (r) + X R () , where x c and X R represent the controlled 

and uncontrolled displacement vector, respectively. We refer to the uncontrolled modes as residual. 

The eigenvectors are orthogonal and can be normalized so as to satisfy the orthononnality conditions. 

Inserting Eq. (2) into Eq. (1), multiplying by$ r r and considering orthogonal condition between 

eigenvectors, we obtain 

(3) 

where C are r modal damping ratios, ® r is a natural frequency. Then, Equation (3) can be rewritten in 

state-space form such as 

w c (t) = A c w c (t) + B c f(l) + E c x s , y c (t} = C c w c (t} (4) 

where w c is a 2/-dimensional modal state vector by the controlled modes and 

' ° 

-al -/ '' 

/c I s B - - I E - 

— 

Er ~ 

~ '' 

(5) 

are the 21x21, 2Lxm matrixes and a 2/xl vector, respectively, and Ac is the diagonal matrix listing 

2co r £ r ; ii 2 is the diagonal matrix listing fi),..., a$\ B c '=® T A\ and E c '=

210 

Gaussian) methods are advocated because of their successful application in previous studies (Dyke et 

al. 1996). For the controller design, x g is taken to be a stationary white noise, and an infinite horizon 

performance index is chosen that weights the modal states by controlled modes such as 

J slimier f \w T cQw c + u T Ru)dt] (7) 

*-»- r L J ° J 

where R is a 2x2 identity matrix because the numerical example has two MR dampers, and Q is a 

2Zx2/ diagonal matrix. It should be noted that the size of Q is reduced from 2nx2n to 21x21 because the 

limited lower modes are controlled. Therefore, it can be said that it is more convenient to design the 

smaller weighting matrix of modal control. For example, when the lowest one mode is controlled for 

calculating the modal control action, Q is a 2x2 diagonal matrix. 

Modal State Estimation 

An observer for modal state estimation should be provided, since real sensors may not estimate the fu]l 

modal states directly or the system may be expensive to prepare the sensors for the full states. To 

estimate the modal state vector H>cW from the measured output y (£), we consider a Kalman-Bucy filter 

as an observer(Meirovitch, 1990). Not only, in this paper, the state feedback including velocities or 

displacements is considered, but also the acceleration feedback is implemented for the modal state 

estimation using a Kalman-Buch filter. In any case, we can write a modal observer in the form 

$ c (t) = A c w c (t) + B c f(t) + E c x g +L[y(t)~C c w c (t}-D c f(t)] (8) 

where w c (t) is the estimated controlled modal state and L is the optimally chosen observer gain matrix 

by solving a matrix Riccati equation, which assumes that the noise intensities associated with 

earthquake and sensors are known. C c is changeable according to the signals which are used for the 

feedback and D c is generally zero except the acceleration feedback. For modal state estimation from 

the displacements, C c = [4> C 0 ]. For control with the velocity feedback, C c = [ 0

211 

To examine the effect of the control forces on the uncontrolled modes, residual modes can be written 

^R(^= : A R w R (t) J rB R f(t) + E R x g (12) 

where W R is a residual state vector by uncontrolled modes. Substituting Eq. (10) into Eq. (4) and 

considering Eq. (12), we obtain 

w c (r) = A c w c (t)-B c K c w c (t) + E c x g} w R (t)-A R w R (t)-B R K c w c (f) + E R x (13) 

Moreover, substituting Eqs. (10) and (11) into Eq. (8), we can write the observer equation in the form 

* c (t) = (Ac - B C K C )>v c (r) + LC C (w c (t) - w c (r)) + LC R W R (r) + E c x g (14) 

The error vector is defined such as e c (t) = w c (t) - w c (t). Then the Equations can be written in the 

matrix form 

A C -B C K C 0 

0 

~B C K C 

— BffK r 

(15) 

LC R A C -LC C 

Note that the term -B^Kc in Eq. (15) is responsible for the excitation of the residual modes by the 

control forces and is known as control spilloveralas. If CR is zeros, which means the sensor signal only 

include controlled modes, the term -B&Kc has no effect on the eigenvalues of the closed-loop system. 

Hence, we conclude that control spillover cannot destabilize the system, although it can cause some 

degradation in the system performance. Normally, however, the above system can not satisfy the 

separate principle because the term LC# affects eigenvalues of the controlled system by the observer. 

This effect is known as observation spillover and can produce instability in the residual modes. 

However, a small amount of damping inherent in the structure is often sufficient to overcome the 

observation spillover effect. At any rate, observation spillover can be eliminated if the sensor signals 

are prefiltered so as to screen out the contribution of the uncontrolled modes (Meirovitch, 1990) 

Numerical Example 

To evaluate the proposed modal control scheme for use with the MR damper, a numerical example is 

considered in which a model of a six-story building is controlled with four MR dampers (Fig. 1). This 

numerical example is the same with that of Jansen and Dyke (2000) and is adopted for direct 

comparisons between the proposed modal control scheme and other control algorithms. In simulation, 

the model of the structure is subjected to the NS component of the 1940 El Centro earthquake. Because

212 

the building system considered is a scaled model, the amplitude of the 

earthquake was scaled to ten percent of the full-scale earthquake. The 

various control algorithms were evaluated using a set of evaluation 

criteria based on those used in the second generation linear control 

problem for buildings (Spencer and Sain 1997) such as 

FIGURE 1 

SCHEMATIC DIAGRAM 

(Jansen and Dyke 2000) 

where x t (f) is the relative displacement of the zth floor over the entire 

response, x"^ denotes the uncontrolled maximum displacement. h t is the 

height of each floor (30cm), d t (t) is the interstory drift of the above 

max 

ground floors over the response history, d n denotes the normalized 

peak interstory drift in the uncontrolled response, x m (r) is the absolute 

accelerations of the zth floor, x is the peak uncontrolled floor 

acceleration, and W is the total weight of the structure (1335 N). The corresponding uncontrolled 

responses are as follows: x*** =1.313 cm, JT = 0.00981 cml'xT (t) = 146.95 cm/sec 2 . 

The resulting evaluation criteria are presented in Table 1 for the control algorithms previously studied 

(Jansen and Dyke, 2000). The numbers in parentheses indicate the percent reduction as compared to 

the best passive case. To compare the performance of the semiactive system to that of comparable 

passive systems, two cases are considered in which MR dampers are used in a passive mode by 

maintaining a constant voltage to the devices. The results of passive-off ( 0 V ) and passive-on (5V) 

configurations are included. 

FIGURE 2 VARIATIONS OF EVALUATION CRITERIA WITH WEIGHTING PARAMETERS 

FOR THE ACCELERATION FEEDBACK 

For modal control, three cases of the structural measurements are considered; displacements, velocities 

and accelerations. Using each structural measurement, a Kalman filter estimates the modal states. Fig. 

2 represents the variations of each evaluation criteria for increasing weighting parameters in a 3- 

dimensional plot JT is the summation of evaluation criteria, //, /2 and /. We can find the weighting 

for reduction of overall structural responses from the variations of JT, whereas we can find the 

weighting for reduction of related responses from //, J and /j, Designer can decide which to use 

according to control objectives. By using the controller (H2/LQG) with designed weighting matrices

213 

from Fig. 2, we can get the results in Table 2. Similarly, for the displacement and velocity feedback 

cases, Table 2 summarize the results for each minimum evaluation criteria of the designed weighting 

matrices. 

TABLE 1* 

NORMALIZED CONTROLLED MAXIMUM RESPONSES 

DUE TO THE SCAPED EL CENTRO EARTHQUAKE 

Control strategy 

Passive-off 

Passive-on 

Lyapunov controller A 

Lyapunov controller A 

Decentralized bang-bang 

Maximum energy dissipation 

Clipped-optimal A 

Clipped-optimal B 

Modified homogeneous friction 

Ji 

0.862 

0.506 

0.686(4-35) 

0.326(-35) 

0.449M1) 

0.548(4-8) 

0.631(4-24) 

0.405(-20) 

0.421 (-17) 

TABLE 2 

J2 

0.801 

0.696 

0.788(4-13) 

0.548(~21) 

0.791(4-13) 

0,620(-11) 

0.640(~8) 

0.547(-21) 

0.599(-20) 

h 

0.904 

141 

0 756(-16) 

1.39(4-53) 

1.00(4-11) 

1.06(4-17) 

0.636(~29) 

1.25(4-38) 

1.06(4-17) 

J 4 

0.00292 

0.0178 

0.0178 

0.0178 

0.0178 

0.0121 

0.01095 

0.0178 

00178 

(* Jansen and Dyke 2000) 

NORMALIZED CONTROLLED MAXIMUM RESPONSES OF THE VARIOUS FEEDBACK DUE 

Control strategy 

Modal control A n 

Modal control Aj 2 

Modal control A^ 

Modal control Ar T 

Modal control Dn 

Modal control D n 

Modal control D^ 

Modal control Dn* 

Modal control V;, 

Modal control V J2 

Modal control V n 

Modal control V rT 

TO THE SCALED EL CENTRO EARTHQUAKE 

Weighting parameters 

qmd=400, qmv=sl500 

qmd=l, qmv=oOO 

qmd=2200, qmv=100 

qmd=500. qmv=600 

^md=100, gmv=4900 

qmd=100, qmv=4900 

qmd=200, qmvs=4900 

qmd=3300, qmv=4700 

qmd-700, qmv=800 

qmd=l, qmv=400 

qmd=1300, qmv=100 

qmd=600, qmv=500 

Ji 

0.310(-39) 

0.398(-21) 

0.549(4-8) 

Q.380(-25) 

Q.403(-20) 

0.403(-20) 

0 702(4-39) 

0408(-19) 

0.327(-35) 

0383(-24) 

0 541(4-7) 

0.354f-30) 

J 2 

0529(-24) 

0 485(-30) 

0.618H1) 

0.488(-30) 

0.560

214 

matrix of modal control. This is one of the important benefits of the proposed modal control scheme. 

The numerical results show that the motion of the structure was effectively suppressed by merely 

controlling a few lowest modes, although resulting responses varied greatly depending on the choice of 

measurements available and weightings. The modal controller A and V achieved significant reductions 

in the responses. The modal controller Ayj, A J2 and V J3 achieve reductions (39%, 30%, 30%) in 

evaluation criteria Ji, Jo and */3, respectively, resulting in the lowest values of all cases considered here. 

The modal controller AJT and VJT fail to achieve any lowest value of evaluation criteria, but have 

competitive performance in all evaluation criteria. Based on these results, the proposed modal control 

scheme is found to be suited for use with MR dampers in a multi-input control system. Further studies 

are underwav to examine the influence of the number of controlled modes on the control performance. 


This research was supported by the National Research Laboratory Grant(No. : 2000-N-NL-01-C-251) 

in Korea. The financial support is gratefully acknowledged. 

REFERENCES 

Brogan, W.L. (1991). Modern Control Theory, Prentice Hall, Englewood Cliffs, New Jersey. 

Dyke, SJL, Spencer Jr., B.F., Sain, M.K. and Carlson, J.D. (1996). "Modeling and Control of 

Magnetorheological Dampers for Seismic Response Reduction," Smart Materials and Structures, Vol. 

5, pp. 565-575. 

Inaudi, J.A. (1997). "Modulated Homogeneous Friction: A Semi-active Damping Strategy," 

Earthquake Engineering and Structural Dynamics, Vol. 26, No. 3, pp. 361. 

Jansen, L.M. and Dyke, SJ. (2000). "Semi-active Control Strategies for MR Dampers: Comparative 

Study," Journal of Engineering Mechanics, Vol. 126, No. 8, pp. 795-803. 

Leitmann, G. (1994). "Semiactive Control for Vibration Attenuation," J. of Intelligent Material 

Systems and Structures, Vol. 5 September, pp. 841-846. 

McClamroch, N.H. and Gavin, H.P. (1995). "Closed Loop Structural Control Using Electrorheological 

Dampers," Proc. oftheAmer. Ctrl. Conf. 9 Seattle, Washington, pp. 4173-77. 

Meirovitch, L. (1990). "Dynamics and Control of Structures," John Wiley & Sons. 

Sack, R.L., Kuo, C.C., Wu, H.C., Liu, L and Patten, W.N. (1994). "Seismic Motion Control via 

Semiactive Hydraulic Actuators." Proc. of the U.S. Fifth National Conference on Earthquake 

Engineering, Chicago, Illinois, Vol. 2, pp, 311-320. 

Spencer Jr., B.F., Dyke, S. J., Sain, M. K., and Carlson, J. D. (1997). "Phenomenological Model off 

Magnetorheological damper," Journal ofEngrg. Mech., ASCE, 123(3), pp. 230-238.

Proceedings of the Intel-national Conference on 



EFFECTIVENESS OF SMART BASE ISOLATION SYSTEM WITH 

MR DAMPERS IN PROTECTING STRUCTURES 

IN NEAR-FAULT EARTHQUAKES 

Satish Nagarajaiah 1 , Sanjay Sahasrabudhe 2 and Yuqing Mao 2 

1.Associate Professor, Civil & Env. Eng. , Rice University, Houston, TX 

2 Engineer, McDermott, Houston, TX, Formerly, Graduate Research Assistant, Rice University 

Abstract 

This paper presents an analytical study of smart base isolated structures with vanable MR dampers, 

which extends the results of prior analytical and experimental studies by the authors. New analytical models 

of the nonlinear base isolated structures with MR dampers are developed. A Lyapunov based controller is 

developed and used to perform simulations. Numerical simulations and parametric studies are performed 

using several near-fault ground motions. The computed results are verified using shake table test results of 

scaled base isolated structural model with MR dampers. It is demonstrated that base isolated structures with 

MR dampers are more effective than base isolated structures with passive nonlinear dampers, in reducing 

the response in wide range of near-fault earthquakes. 

Introduction 

Base isolation systems are effective in protecting structures in strong earthquakes. Recent studies have 

shown that base isolated structures tend to have larger base displacements in near-fault long period earthquakes. 

This may lead to large isolation gap or in cases—with insufficient isolation gap—pounding against 

the retaining wall of the building, and damage to the superstructure. Nonlinear passive dampers have been 

implemented in base isolated buildings to counter some of the effects of near-fault earthquakes (Taylor et 

al. 2001). Although supplemental nonlinear passive dampers can limit base displacements, the isolation 

forces, superstructure drifts and accelerations will be higher. Several researchers have recently studied the 

capability of Magneto-rheological (MR) dampers and other variable dampers in effectively countering such 

limitations (Gavin et al. 2001, Madden et al. 2000, Nagarajaiah et al. 2000, Spencer et al. 2000, Yang et 

al 2002, and Yoshika et al. 2002). MR clampers, which can be used to vary the nonlinear damping in the 

isolation systems, can reduce base displacements in near-fault earthquakes, further than passive nonlinear 

dampers, while maintaining the same or lower level of isolation forces, superstructure drifts and accelerations, 

as compared to passive nonlinear dampers (Sahasrabudhe et al. 2000,2001). Spencer et al. (2000) 

and Yoshioka et al (2002) have shown the effectiveness of base isolated structures with MR dampers in 

both moderate/strong and in near/far-fault earthquakes by analytical and experimental studies. This paper 

presents an analytical study of smart base isolated structures with variable MR dampers, which extends the 

results of prior analytical and experimental studies by the authors (Sahasrabudhe et al. 2000, 2001). It is 

demonstrated that base isolated structures with MR dampers are more effective than base isolated structures 

with passive nonlinear clampers, in reducing the response in wide range of near-fault earthquakes. 

Analytical Model of Smart Base Isolated Structure with MR Dampers 

The equations of motion for the smart base isolated structure shown in Figure 1 are as follows 

M C U C + C C U C + K C U C + F c = B E u s (1) 

c 

f U \ ("

216 

Figure 1: Smart Base Isolated Structure with Sliding Isolation Bearings, Recentering Springs, and MR 

Dampers 

R — the matrix of earthquake influence coefficients, U = modal displacement vector relative to the base 

(modal transformation U =

217 

where gjî = g(X fcTl , F^+ît 

formulation iterations are avoided. 

) (4) 

.); Hence, the method is implicit, needing iteration. In the incremental 

Modeling of Base Isolation Bearings and MR Dampers 

Sliding bearings are modeled using the following equation: 

fb (t) = pN. S (t) (5) 

where N is the normal force at the sliding bearing, the coefficient of friction /x = /max — (/âx - / mm ) • 

e-aK! wj t k u b the velocity of the sliding base, / max the coefficient of sliding friction at high velocity, / m i n 

is the coefficient of sliding friction at essentially zero velocity of sliding, and a is the coefficient controlling 

the dependency of friction on velocity of sliding and z (t] is a Bouc-Wen hysteretic dimensionless quantity 

governed by the equation: 

Y-Z + T \u b \ • z \z\ n ~ l + p-u b -z n -Au b = Q (6) 

where Y is the yield displacement, 7, /3, n, and A are dimensionless parameters of the model that govern 

the hysteretic behavior, and lib is the base displacement. 

Magnetorheological (MR) dampers are semi-active control devices that use MR fluids to produce controllable 

damping. In this paper an extension to the MR damper model by Spencer et al (2000 ) is developed. 

The Bouc-Wen model hysteretic element with spring and variable dashpot in parallel are used to model the 

MR damper, as shown hi Figure 2. 

Figure 2. MR Damper Model 

The governing equation of the model shown in Figure 2 is given by 

where k is the stiffness of the accumulator, the variable damping coefficient, c = c a -f c& with c a being 

constant and c b — GI • hi(v), a = a a 4- a b with a a being constant and a& » aj. • h^(v}^ with &i(v), /ia(u) and 

f(v) being functions of voltage, v, applied to the MR damper, f(u) accounts for the variation in the force 

generated by the damper under different voltages. The hysteretic variable z of Bouc-Wen Model is governed 

by equation (6), where Y is the yield displacement and 7 = 0.9, / = 0.1, n = 2, .4 - 1. 

Equations (5), (6) and (7) are solved using the unconditionally stable semi-implicit Runge-Kutta method 

suitable for stiff differential equations. Equations (3), (4), (5) and (6) are solved using an efficient predictorcorrector 

algorithm developed by the authors.

218 

Lyapunov Controller for the System with MR Damper 

For MR damper, a Lyapunov Based Controller is developed as Mows. The equation (3) can be rewritten 

in absolute coordinates for a rigid superstructure as follows 

= AX a (i) + B/(t) 4- Bu + Bk b u a = g(X a ,/, u g ) (8) 

i a = the absolute displacement of rigid 

mass m fc , / = the aonlinear force at the isolation level, u = c(t) u bl c(i) is the tune varying damping 

coefficient of the MR damper, u b — relative displacement of mass with respect to the ground, and u g = 

ground displacement. Based on a appropriate Lyapunov function V and V being negative or minimum the 

following switching control algorithm is derived. 

C (+\ 

( t ) _ j 

^ 

— \ o or 

where C min is the minimum damping coefficient for one volt, C max is the maximum damping coefficient 

for 4 volts, pi and p> are constants. 

Figure 1 shows a smart sliding isolated two-story building model with MR Damper, a restoring spring, 

and four sliding bearings, considered in this analytical study. A corresponding scaled model has been tested 

by authors (Sahasrabudhe et al. 2000,2001). The 1:5 scale model is 1.47m in length and the height is 

1.48m, with 0.74m height of each floor. The base mass is 5.54 N-s 2 /cm, and the mass of the first and second 

floor is 5.92 N-s 2 /cm each. The total stiffness of recentering springs connected between the base of the 

building and the shaking table is 720 N/cm. The condensed superstructure stiffness matrix in the fixed base 

condition (including joint rotations), and corresponding clamping matrix are considered. The fundamental 

period of the superstructure in the fixed base condition (2DOF) is 0.15 sec. The fundamental period of the 

model sliding base isolated structure (3DOF) is nearly 1.0 sec (2.24 sec at prototype scale). Simulations 

are performed at both the prototype and model scale with proper consideration being given to the scaling 

issues. For the superstructure (2DOF), in the fixed base condition, the damping coefficients are & = 2.56% 

and g> = 1.48%. Simulations are performed with MR clamper off—constant zero volts—low damping, MR 

damper on—constant four volts—high damping, and controlled cases where the voltage is switched between 

one and four volts. 

Results 

The response of the prototype smart sliding isolated rigid structure (SDOF) with an isolation period of 2 

sec to a cosine pulse with a period of 2 sec, representative of the Rinaldi fault normal earthquake, is shown 

in Figure 3. It is evident from the response in Figure 3 that in the controlled case the base displacement as 

well as the total shear force at the isolation level are reduced further than the passive high damping case; 

the switching of the MR damper dissipation force, from one volt level to the four volt level, in the controlled 

case is clearly evident. It is clearly evident that the smart isolation system performs better than the passive 

high damping case. 

The response of the prototype smart sliding isolated rigid structure (3DOF) with an isolation period of 

4 sec to Newhall fault normal earthquake, is shown in Figure 4. It is evident from the response in Figure 4 

that the controlled case clearly reduces the base displacement as well as the total shear force at the isolation 

level. In the total isolation force-base displacement loops shown in Figure 4 the difference in loops of the 

three cases is primarily due to changes in MR clamper forces—switching of the MR damper, between one 

and four volt levels, in the controlled case is clearly evident in the force-displacement loops. The ability of 

the smart MR damper to reduce the response further than the passive high damping case is clearly evident. 

Next the simulations of the smart sliding isolated model structure (SDOF) are performed for the following 

earthquakes: (1) El-Centro SOOE Earthquake (May 18,1940), peak acceleration: 0.34 g (100%); (2) Newhall 

Channel 1 90 Deg. Fault Parallel (Jan. 17, 1994), peak acceleration: 0.608 (105%)g; (3) Newhall Channel

219 

3 360 Deg. Fault Normal (Jan. 17, 1994), peak acceleration: 0.59 g (100%); (4) Sylmar Channel 1 90 Deg 

Fault Parallel (Jan. 17, 1994), peak acceleration: 0.461 g (80%); (5) Sylmar Channel 3 360 Deg. Fault 

Normal (Jan. 17, 1994), peak acceleration: 0,84 g (100%); (6) Kobe NS (Jan. 17, 1995), peak acceleration: 

0816 g (100%); (7) Kobe EW (Jan. 17, 1995), peak acceleration: 0.62 g (100%). The records are time 

scaled by a factor of 2.236 to satisfy similitude requirements. The peak accelerations are scaled from the 

original recorded values. Results are presented in Table 1 to Table 3. Table 1 shows peak values of relative 

base displacement and normalized peak shear force at the isolation level (total of friction force, MR damper 

force, and spring force normalized by total weight). Table 2 shows the peak values of first floor and second 

floor inter-story drifts. Table 3 shows the peak values of first floor and second floor accelerations. 

In Tables 1 to 3 in case of Sylmar, Newhall, and Kobe earthquakes, the relative base displacement 

reductions range from 5 to 40 % in the passive high damping (4 volt) case when compared to the passive 

low damping (0 volt) case. This occurs due to the increase in energy dissipation in high damping case, with 

nearly 5 to 30 % increase in the peak total force at the isolation level when compared to the low damping case. 

The semiactive controlled case gives an additional 5 to 15 % reduction in the base displacement response 

over the passive high damping (4 volt) case. The controlled case also reduces or maintains the same level 

of total force at the isolation level when compared to the high damping case. Hence, the controlled case 

reduces the base displacement further, when compared to 0 volt and 4 volt cases, with no further increase 

in force as compared to 4 volt case. The energy dissipation occurs more efficiently in the controlled case 

revealing the potential of smart damping. In most cases the interstory drifts in the controlled case are the 

smaller than the passive high damping case—which indicates the effectiveness of the smart isolation system, 

The accelerations in the controlled case remain bounded by the high and low damping cases. As evident 

from Table 1 in case of Sylmar 360 earthquake the peak relative base displacement, the largest amongst all 

earthquakes, is reduced by 42% in the passive high damping (4 volt) case when compared to the passive low 

damping (0 volt) case. The passive high damping (4 volt) case also reduces the total force at isolation level 

by 4% when compared to the passive low damping (0 volt) case. The controlled case gives additional 7% 

reduction in the base displacement response over the passive high damping (4 volt) case; the control case 

also maintains the same level of total force at the isolation level as compared to that of the passive high 

damping case. Thus based on the results in Tables 1, 2 and 3 it can be concluded that the controlled case 

reduces the base displacement further and maintains the total force at the isolation level within bounds in 

most earthquakes-the one exception being El Centro. In addition the controlled case maintains the interstory 

drifts and acceleration response of the two-story model within bounds. Thus the newly developed 

Lyapunov based control algorithm and the smart isolation system with MR damper is effective in reducing 

the base displacements, without further increases in the total force at the isolation level, interstory drifts 

and accelerations. 

Table 1 Peak Relative Base Displacement and Normalized Peak Shear Force at the Isolation Level of the 

Two-story Smart Sliding Isolated Structure 

Earthquake Rel. Base Disp.(cm) Force at Isolation Level/W 

El Centro(100%) 

Sylmar 90 (80%) 

Sylmar 90(100%) 

Sylmar 360(105%) 

Newhall 90(100%) 

Newhall 360(100%) 

Kobe NS(80%) 

Kobe NS(100%) 

Kobe EW 

0 Volt 

0.938 

2.293 

3.00 

5.48 

1.815 

3.867 

2.511 

3.362 

2.578 

4 Volt 

0.85 

1.649 

2.15 

3.183 

1.614 

3.373 

2.45 

2.825 

2.24 

Control 

0.77 

1.484 

2.1 

3.02 

1.276 

3.294 

2.105 

2.543 

2.22 

0 Volt 

0.16 

0.224 

0.254 

0.354 

0.21 

0.293 

0.204 

0.235 

0.236 

4 Volt 

0.244 

L 0.263 

0.292 

0.343 

0.28 

0.349 

0.29 

0.312 

0.305 

Control 

0.243 

0.26 

0.279 

0.34 

0.261 

0.29 

0.28 

0.3 

0.278 

Comparison with Experimental Results 

Comparison of analytical and experimental results are presented in Table 4 for three earthquakes: (1)

220 

Sylmar 90, (2) Newhall 90, and (3) El Centro. As evident from Table 4 the comparison is satisfactory 

indicating'the accuracy of the analytical model. In addition it is worth noting in Table 4 that in the 

experimental results of Sylmar 90, Newhall 90, and El Centro the base displacement is reduced further by 

15 to 20 % in passive high damping case (4 volts) when compared to passive low damping (0 volts) case; 

this occurs due to increase in peak total force at the isolation level in the high damping case by 15 to 45 

% as compared to the low damping case. In the controlled case with smart damping the base displacement 

is further reduced by 10 to 15% while the peak total force at the isolation level is less than that of the 

high damping case. The controlled case reduces the base displacement, when compared to passive low and 

high damping cases, with no further increase in force as compared to passive high damping (4 volt) case. 

Also interstory drifts in the controlled case, presented in Table 4, are smaller than the passive high damping 

case, while the accelerations remain bounded. From Table 4 it is clearly evident that the reductions in base 

displacement and total isolation force in the experimental case are better than the analytical case. 

Table 2 Interstory Drifts of the Two-story Smart Sliding Isolated Structure 

Earthquake First Story Drift (cm) Second Story Drift (cm) 

El Centro 

Sylmar 90 (80%) 

Sylmar 90 

Sylmar 360 

Newhall 90 

Newhall 360 

Kobe NS(80%) 

Kobe NS 

KobeEW 

OVolt 

0.111 

0.109 

0.121 

0.149 

0.129 

0.137 

0.134 

0.14 

0.12 

4 Volt 

0.136 

0.144 

0.157 

0.165 

0.155 

0.179 

0.158 

0.18 

0.153 

Control 

0.136 

0.126 

0.156 

0.15 

0.141 

0.159 

0.151 

0.158 

0.145 

OVolt 

0.15 

0.143 

0.148 

0.181 

0.162 

0.153 

0.162 

0.169 

0.137 

4 Volt 

0.166 

0.181 

0.201 

0.215 

0.222 

0.227 

0.205 

0.222 

0.191 

Control 

0.166 

0.157 

0.2 

0.19 

0.209 

0.187 

0.187 

0.2 

0.172 

Table 3 Acceleration Response of the Two-story Smart Sliding Isolated Structure 

Earthquake Base Ace. cm) First Fl. Ace. (g) Second Fl. Acc.(g) 

El Centro 

Sylmar 90 (80%) 

Sylmar 90 

Sylmar 360 

Newhall 90 

Newhall 360 

Kobe NS(80%) 

Kobe NS 

KobeEW 

0 V 

0.45 

0.48 

0.51 

0.55 

0.56 

0.47 

0.56 

0.633 

0.474 

4V 

0.76 

0.55 

0.61 

0.63 

0.75 

0.74 

0.752 

0.81 

0.64 

Cont. 

0.75 

0.51 

0.61 

0.6 

0.73 

0.66 

0.7 

0.736 

0.611 

0 V 

0.352 

0.333 

0.41 

0.55 

0.359 

0.46 

0.372 

0.47 

0.42 

4 V 

0.66 

0.435 

0.49 

0.564 

0.60 

0.49 

0.486 

0.524 

0.54 

Cont. 

0.66 

0.412 

0.489 

0.55 

0.60 

0.489 

0.485 

0.506 

0.518 

0 V 

0.497 

0.478 

0.5 

0.702 

0.551 

0.55 

0.59 

0.597 

0.494 

4V 

0.541 

0.576 

0.65 

0.679 

0.785 

0.79 

0.806 

0.789 

0.66 

Cont. 

0.54 

0.515 

0.645 

0.631 

0.774 

0.664 

0.72 

0.693 

0.6 


Sylmar90 

Sylmar 90 

Sylmar 90 

Newk0190 

Newhall 90 

Newhall 90 

El Centro 

El Centro 

El Centro 

Table 4 Comparison of Analytical and Experimental Results 

MRD Relative Base Total Force First Story Second Story 

Disp. (cm.) Isol. Lvl./Wt. Drift (cm) Drift (cm) 

Expt. Anal. Expt. Anal. Expt. Anal Expt. Anal. 

0.0 V 2.179 2.293 0.212 0.224 0.113 0.109 0.127 0.143 

40V 1.761 1.649 0.28 0.263 0.139 0.144 0.147 0.181 

Control 1.409 1.484 0.26 0.26 0.128 0.126 0.141 0.157 

0.0 V 1.914 1.815 0.214 0.21 0.123 0.129 0.122 0.162 

4.0V 1.613 1.614 0.28 0.28 0.185 0.155 0.192 0.222 

Control 1.411 1.276 0.241 0.261 0.163 0.141 0.159 0.209 

0.0 V 0.969 0.938 0.166 0.16 0.098 0.111 0.142 0.15 

4.0V 0.928 0.847 0.246 0.244 0.131 0.136 0.159 0.166 

Control 0.743 0.775 0.24 0.24 0.119 0.136 0.142 0.166

221 


Sylmar 90 

Sylinar 90 

Sylmar 90 

Newhall 90 

Newhall 90 

NewhaU 90 

El Centro 

El Centre 

El Centre 

MRD 

0.0 V 

40V 

Control 

0.0 V 

4.0V 

Control 

0.0 V 

4.0V 

Control 

Table 4 Continued 

Base Ace. (g) First Fl Acc.(g) 

0.38 0.296 0.384 0.478 

0.43 0.324 0.482 0.576 

0.40 0.32 0.481 0.515 

0.42 0.56 0.484 0.359 

0.565 0.75 L 0.544 0.60 

0.58 0.73 0762 0.60 

0.44 0.192 0.301 0.497 

0.40 0.308 0.364 0.541 

0.38 0.282 0.344 0.54 

Sec. Fl. Ace. (g) 

0.333 0.48 

0.435 0.55 

0.412 0.51 

0.359 0.551 

0.59 0.785 

0.609 0.774 

0.352 0.45 

0.66 0.76 

0.66 0.75 

Conclusions 

Analytical study of smart sliding isolated structure with MR dampers is presented. New detailed nonlinear 

analytical models for the smart sliding isolated structure and MR damper are developed and the solution 

procedure is presented. The analytical results are verified using shake table test results. The response to 

several near fault earthquakes is presented. From the presented analytical and experimental results T it is 

evident that the smart sliding isolation system with MR dampers is capable of reducing the base displacement 

response and the total isolation force further to the passive high damping case, while maintaining the 

interstory drifts and accelerations within bounds 

Acknowledgments 

Funding for this project provided by the National Science Foundation CAREER Grant (CMS-996290), 

with Dr. S. C. Liu and Dr P. Chang as program directors, is gratefully acknowledged 

References 

Gavin, H. and Aldernir, U. (2001). "Behavior and response of auto-adaptive seismic isolation," Proc. U. 

S. Japan Workshop in Urban Earthquake Disaster Mitigation, Seattle, WA. 

Maddan, G. J., Wongprasert, N., and Symans, M. D. (2000). "Adaptive seismic isolation systems for 

structures subjected to disparate earthquake ground motions,'" Proc. Structures Congress, ASCE, Philadelphia 

Ṅagarajaiah, S, Sahasrabudhe, S., and Iyer, R. (2000). "Seismic Response of Sliding Isolated Bridges with 

Smart Dampers Subjected to Near Source Ground Motions," Proc. Structures Congress, ASCE, Philadelphia. 

Sahasrabudhe, S., Nagarajaiah, S., and Hard, C. (2000). "Experimental Study of Sliding Isolated Buildings 

with Smart Dampers," Proc.Eng. Mech. Conf., ASCE, UT Austin. 

Sahasrabudhe, S., and Nagarajaiah, S. (2001)."Sliding Isolated Buildings with Smart Dampers: Shake 

Table Studies," Proc. Structures Congress, ASCE, Washington D. C. 

Spencer, B.F, Johnson, R. A., and Ramallo, J. C., (2000)."Smart isolation for seismic control," JSME 

Int. J. Series C-Mech. Sys. Machine Elem. And Manufacturing, 43 (7), 704-711. 

Taylor, D. P., and Constantinou, M. C. (2001). "Fluid Dampers for Applications of Seismic Energy 

Dissipation and Seismic Isolation" Taylor Devices Inc., also see http://www.taylordevices.com/. 

Yang, J. N., and Agrawal, A. K. (2002)."Semiactive hybrid control systems for nonlinear buildings against 

near-field earthquakes," Eng. Structures, 24 (3): 271-280. 

Yoshioka, H., Ramallo, J. C.; and B. F. Spencer Jr. (2002). "Smart Base Isolation Strategies Employing 

Magnetorheological Dampers," J. of Eng. Mechanics, ASCE, Vol. 128, No. 5.

222 

V 

•03 -W 

-0.3 .02 

Figure 3. Response of the Prototype Smart SHding Isolated Rigid Structure (SDOF) with an 

Isolation Period of 2 sec to a Cosine Pulse Excitation of 2 sec Period 

Figure 4. Response of a Prototype Smart Sliding Isolated Structure (3DOF) with an 

Isolation Period of 4 sec to a Newhall FN Earthquake Excitation




SEISMIC PROTECTION OF A BENCHMARK CABLE-STAYED 

BRIDGE USING A HYBRID CONTROL STRATEGY 

K. S. Park, H. J. Jung, K. M. Choi, and I. W. Lee 


Korea Advanced Institute of Science and Technology, Daejeon, Korea 

ABSTRACT 

This paper presents a hybrid control strategy for seismic protection of a benchmark cable-stayed bridge, 

which is provided as a testbed structure for the development of strategies for the control of cablestayed 

badges. In this study, a hybrid control system is composed of a passive control system to reduce 

the earthquake-induced forces in the structure and an active control system to further reduce the bridge 

responses, especially deck displacements. Lead rubber bearings and ideal hydraulic actuators are used 

for the passive and active control systems. An H 2 fLQG control algorithm is adopted as an active 

control algorithm. Numerical simulation results show that the performance of the proposed hybrid 

control strategy is superior to that of the passive control strategy and slightly better than that of the 

active control strategy. The proposed control method is also more reliable than the fully active control 

method due to the passive control part. Therefore, the proposed hybrid control strategy can effectively 

be used to seismically excited cable-stayed bridges. 

INTRODUCTION 

Structural control systems, such as passive, active, semiactive or a combination thereof, could provide 

an efficient means for seismic protection of cable-stayed bridges, but the control of such type of bridge 

is a new, unique and challenging problem because those structures are very large and flexible.

224 

Under the coordination of the ASCE Committee on Structural Control, Dyke et al developed the first 

generation of benchmark structural control problems for seismically excited cable-stayed bridges to 

investigate the effectiveness of various control strategies (Dyke et al., 2000). In this study, a hybrid 

control strategy for the seismic protection of a cable-stayed bridge is investigated by using this ASCE 

first generation benchmark bridge model. 

BENCHMARK PROBLEM STATEMENT 

This benchmark problem considers the cable-stayed bridge shown in Fig. 1, which is scheduled for 

completion 2003 in Cape Girardeau, Missouri, USA. In this benchmark study, only the cable-stayed 

portion of the bridge is considered, because the Illinois approach has a negligible effect on the 

dynamics of the cable-stayed portion of the bridge. Based on detailed drawings of the bridge, Dyke et 

al developed and made available a three-dimensional linearized evaluation model that effectively 

represents the complex behavior of the full-scale benchmark bridge. The stiffness matrices used in this 

linear model are those of the structure determined through a nonlinear static analysis corresponding to 

the deformed state of the bridge with dead loads. Because this bridge is assumed to be attached to 

bedrock, the effect of the soil-structure interaction has been neglected. A one-dimensional ground 

acceleration is applied in the longitudinal direction for seismically excited cable-stayed bridges. 

142.7m 350.6m 142.7m 57Dm 

Bout Pier 2 

Q Cable Number 

FIGURE 1 SCHEMATIC OF THE CAPE GIRARDEAU BRIDGE (Dyke et a/., 2000) 

Application of static condensation to the full model of the bridge as a model reduction scheme resulted 

in a 419 DOF reduced-order model, designated the evaluation model. Each mode of this evaluation 

model has 3% of critical damping, which is consistent with assumptions made during the design of 

bridge. The first ten frequencies of the evaluation model for the uncontrolled system are 0.2899, 

0.3699, 0.4683, 0.5158, 0.5812, 0.6490, 0.6687, 0.6970, 0.7102, and 0.7203 Hz. The deck-tower 

connections in this model are fixed (Le., the dynamically stiff shock transmission devices are present). 

On the other hand, another evaluation model should be formed in which the connection between the 

tower and the deck are disconnected to place control devices acting longitudinally. The first ten 

frequencies of this second model are 0,1618, 0.2666, 0.3723, 0.4545, 0.5015, 0.5650, 0.6187, 0.6486,

225 

0.6965, and 0.7094 Hz, which are much lower than those of the nominal bridge model 

Three historical earthquake records are considered as ground excitations for numerical simulations of 

seismic protective systems installed in the bridge: i) 1940 El Centre NS; ii) 1985 Mexico City; and iii) 

1999 Turkey Gebze NS. Eighteen criteria have been defined to evaluate the capabilities of each 

proposed control strategy. The first six evaluation criteria are peak responses of the bridge to consider 

the ability of the controller. The second five evaluation criteria consider normed responses over the 

entire simulation time. The last seven evaluation criteria consider the requirements of each control 

system itself. More details can be found in Dyke et al. (2000). 

SEISMIC CONTROL SYSTEM USING HYBRID CONTROL STRATEGY 

Control Devices 

Passive control devices 

In this study, conventional base isolation devices such as lead rubber bearings (LRBs) are used. The 

design of passive control device follows a general and recommended procedure (Ah" and Abdel- 

Ghaffar, 1995). In the design procedure, the design shear force level for the yielding of lead plugs is 

taken to be 0.1 OM, where M is the part of the deck weight carried by bearings. The asymptotic 

stiffness ratio of the bearings at the bent and tower are assumed to be 1.0. As the results, a total of 24 

LRBs are placed between the deck and pier/bent. Six LRBs are installed between the each deck and 

pier/bent. The properties of LRBs are shown in Table 1 and these LRBs are installed after removing 

the horizontal stiffness of beam element in pier 4. 

Nonlinear model proposed by Wen (Wen, 1989) is used to simulate the motion of nonlinear dynamics 

of the LRB. The restoring force of the model is composed of the linear and the nonlinear terms as 

(1) 

(2)

226 

where k 0 and a are the linear stiffness and its contribution to restoring force, x r and x r are relative 

displacement and relative velocity of nodes which LRBs are installed, respectively. And D v and 3; are 

the yield displacement of LRB and the variable, respectively, satisfying Eqn. 2. 

where A to y, f> and n are the constant that affect the hysteretic behavior. The values of A,=rc=l and 

CE=/3=0.5 are used to simulate the characteristic curve of the LRB in this study. 

TABLE 1 

THE PROPERTIES OF THE LRB 

Property 

Elastic stiffness, & e (N/rn) 

Plastic stiffness, Ap(N/m) 

Yield displacement of lead plugs, D v (cm) 

Design shear force level for the yielding of lead plugs, Q

227 

Two displacement sensors are positioned between the deck and pier 2 and two displacement sensors 

are located between the deck and pier 3. All sensor measurements are obtained in the longitudinal 

direction to the bridge and are assumed to be ideal, having a constant magnitude and phase (Dyke et al, 

2000). 

Control Design Model 

A reduced order model of the system is developed for control design, which is formed from the 

evaluation model and has 30 states. This model obtained by forming a balanced realization of the 

system and condensing out the states with relatively small controllability and observability grammians 

(Laub et al., 1987). The resulting state space system is represented as follows 

(5) 

(6) 

(7) 

where x^ is the design state vector , x s is the ground acceleration, u is the control command input, and 

z is the regulated output vector including evaluation outputs (Le,, shear force and moments in the tower, 

deck displacements, and cable tension forces, etc). 

Control Algorithm 

In this study, an #2/LQG control design is adopted for the active control part. For this design, x g is 

taken to be a stationary white noise, and an infinite horizontal cost function is chosen as 

(8) 

where R is an identity matrix of order 8, and Q is the response weighting matrix. Further, the 

measurement noise is assumed to be identically distributed, statistically independent Gaussian white 

noise process, and the ratio of autospectral density function is 25.

228 

In the optimal control such as LQG, obtaining the appropriate weighting parameters is very important 

to get well-performed controllers. In this study, the maximum response approach is used as follows: i) 

select the responses which could be considered as the important responses for the overall behaviors of 

the bridge; ii) perform the simulations in each parameter with varying the value of the parameter and 

determine the appropriate weighting parameters and combination; iii) perform the additional 

simulation in the combination of the weighting parameters selected in the previous step and finally 

select the appropriate values of each weighting parameters. Consequently, the following combination 

and values of weighting parameters are obtained through the abovementioned approach for active and 

hybrid control systems: 

229 

14% to 45% (under El Centre earthquake), 11% ~ 24% (under Mexico City earthquake), and 10% ~ 

57% (under Gebze earthquake) compared to the passive control system. The structural responses with 

the hybrid control system under El Centro earthquake are decreased by 1% - 26% compared to the 

active control system. All the structural responses except the peak shear at deck level (J 2 ) are 

decreased by 0.3% - 35% under Mexico City earthquake. J 2 is increased by 2%. In the case of 

Gebze earthquake, the structural responses are decreased by 4% - 24%, whereas the normed moment 

at deck level (Jio) is increased by 2%. Moreover, active and hybrid control systems are satisfied the 

actuator requirement (Le., Max force: 1000 kN, Max Stroke: 0.2 m, Max velocity: 1 m/s) given by 

Dyke et al (2000). 

(a) El Centro earthquake (b) Mexico City earthquake (c) Gebze earthquake 

FIGURE 2 RESTORING FORCE OF LRB AT PIER 2 

TABLE 2 

MAXIMUM EVALUATION CRITERIA FOR ALL THE THREE EARTHQUAKES 

Criterion 

J r peak base shear 

J 2- peak shear at deck level 

J 3- peak overturning mom 

J 4- peak mom. at deck level 

J 5- peak dev. of cable tension 

JCT peak deck displacement 

J 7- normed base shear 

Jg- normed shear at deck level 

Jg- normed overturning mom. 

JUT normed mom. at deck level 

IIP normed dev. of cable tension 

J 12- peak control force 

J 13- peak stroke 

JV peak power 

J 15- peak total power 

lie- no. of control devices 

Ji7- no. of sensors 

Jis~ no. of resources 

Dvke et aL 

0.4582 

1.3784 

0.5836 

1.2246 

0.1861 

3.5640 

0.3983 

1.4271 

0.4552 

1.4569 

2.2968e-2 

1.7145e-3 

1.9540 

7.3689e-3 

6.9492e-4 

24 

9 

30 

Passive Control 

0.5459 

1.4616 

0.6188 

1.2656 

0.2077 

3.8289 

0.4211 

1.5502 

0.4815 

1.4429 

2.2327e-2 

2.l611e-3 

2.0993 

- 

- 

24 

, 

- 

Active Control 

0.5071 

1.1576 

0.4485 

0.8792 

0.1474 

1.8023 

0.3755 

0.9510 

0.3563 

0.7618 

1.6176e-2 

1.9608e-3 

0.9886 

9.3311e-3 

8.7997e-4 

24 

9 

30 

Hybrid Control 

0.4854 

0.9360 

0.4471 

0.6719 

0.1462 

1.6629 

0.3723 

0.9169 

0.3336 

0.7799 

1.8215e-2 

LRB+HA: 

2.6438e-3 

LRB: 1.2246e-3 

HA: 1.9608e-3 

0.9118 

6.6678e-3 

8.4888e-4 

LRB-fHA: 

24-^-24 

9 

30

230 

CONCLUSIONS 

In this paper, a hybrid control strategy, which is composed of a passive control system to reduce the 

earthquake-induced forces in the structure and an active control system to further reduce the bridge 

responses, especially deck displacements, has been proposed by investigating the ASCE first 

generation benchmark control problem for seismic responses of cable-stayed badges. The proposed 

control design uses conventional base isolation devices such as LRBs and ideal hydraulic actuators for 

the passive and active control systems. The Bouc-Wen model is used to simulate the nonlinear behavior 

of these devices. An #2/LQG control algorithm is adopted for the active control part. The numerical 

results show that all the structural responses with the proposed hybrid control strategy except the peak 

shear at deck level under Mexico City earthquake and the normed moment at deck level under Gebze 

earthquake are decreased by 03% ~ 35% compared to the active control strategy. In the comparison to 

the passive control system, all the structural responses are decreased by 10% ~ 57% because of the 

additional active control devices in the hybrid control system. The hybrid control strategy is also more 

reliable than the active control method due to the passive control part. Therefore, the proposed hybrid 

control strategy can effectively be used to seismically excited cable-stayed bridges. 


This research was supported by the National Research Laboratory Grant(No. : 2000-N-NL-01-C-251) 

in Korea. The financial support is gratefully acknowledged. 

REFERENCES 

All H. M., and Abdel-Ghaffar A. M, (1995). Seismic passive control of cable-stayed bridges. Shock 

and Vibration 2:4,259-272. 

Dyke S. J., Turan G.., Caicedo J. M., Bergman L. A., and Hague S. (2000). Benchmark control 

problem for seismic response of cable-stayed bridges. http://wusseLcive.wustl.edu/quake/ 

Wen Y. K. (1989). Method for random vibration for inelastic structures. Journal of applied mechanics 

division 42:2, 39-52. 

Laub, A. J., Heakth, M. T., Paige, C C, and Ward, R, C. (1987). Computation of system balancing 

transformations and other applications of simultaneous diagonalization algorithms. IEEE Transaction 

on Automatic Control AC-32,17-32.




STRUCTURAL VIBRATION CONTROL USING PIEZOCERAMIC 

PATCH ACTUATOR 

G. Song 'and B. Xie 

Smart Materials & Structures Laboratory 

Department of Mechanical Engineering 

The University of Akron, 

Akron, OH 44325-3903 USA 

ABSTRACT 

This paper presents active vibration control of a 3-floor model building using piezoceramic patch 

actuators. Piezoceramic material possesses the property of piezoelectricity, which describes the 

phenomenon of generating an electric charge in a material when subjected to a mechanical stress (direct 

effect), and conversely, generating a mechanical strain in response to an applied electric field. This 

property prepares piezoceramic materials being able to function as both sensors and actuators. The 

advantages of piezoceramic include high efficiency, no moving parts, fast response, and being compact. A 

commonly used piezoceramic is the Lead zirconate titanate (PZT), which has a strong piezoeffect. PZT 

actuation strain can be on the order of 1000 £i strain. Within the linear range, PZT actuators produce 

strains that are proportional to the applied electric field/voltage. These features make them attractive for 

dynamic applications. PZT can be fabricated into different shapes to meet specific geometric 

requirements. PZT patches are often used as both sensors and actuators, which can be surface-bonded to 

various structures. In the research, surface-bonded PZT patches are used as both sensor and actuators. 

Prior to control design, experimental modal testing of the model building is performed to reveal the 

dominant modes and their corresponding modal shapes. This information is then used in the control 

design. Several controllers, positive position feedback, strain rate feedback, and sliding-mode control, are 

designed and implemented on the 3-floor model building. Increased vibration damping is observed in all 

control designs. A comparative study is conducted to reveal the advantages and disadvantages of each 

control design. 

1 INTRODUCTION 

Recent years have seen the emergence of the new field of structural control "smart structure" or "active 

structure" for application to active control of seismic-excited linear and nonlinear civil engineering 

structures. The term of "smart structure" or "active structure" is uses to describe a structure, which has the 

ability to sense and adapt to the changing operational condition according to the designed specifications. 

Smart structures are characterized by integrating sensors, actuators and micro-processors. Among the 

smart materials used in the smart structures, piezoceramics has proven to be one of the most promising 

ones in the active control applications. The advantages of piezoceramic include high efficiency, no 

The author to whom all correspondence should be addressed, Tel: (330) 972- 6715; Fax: (330) 972- 6027; Email:

232 

moving parts, and fast response, being compact and easy implementation. A commonly used 

piezoceramic is the lead zirconate titanate (PZT), which can produce a maximum actuation strain on the 

order of 1 000 \JL strain. Furthermore, PZT actuators produce strains that are proportional to the applied 

electric field/voltage within linear range, which have bandwidths beyond the frequency range of structural 

and acoustic control applications. These features make them attractive for active control applications. 

The controller design of smart structures has been based on the linear and nonlinear control theories. In 

the linear controller, positive position feedback (PPF) (Goh and Caughey (1985); Fanson and Caughey 

(1990); Agrawal and Bang (1994); Song et al, (2000)) is applied by feeding the structural position 

coordinate directly to the compensator, and the product of the compensator and a scalar gain positively 

back to the structure. Song et al (1998) experimentally demonstrated that PPF is insensitive to a varying 

modal frequency. Strain Rate Feedback (SRF) control is also used for active damping of a flexible space 

structure (Newman (1992), Song etal (2002)). 

In the nonlinear control of civil engineering application, Yang et al (1997) has used the sliding mode 

controller for wind and seismic response control application on ten-story and forty-story steel frame 

building. Allen et al (2000) has used this control algorithm to control a large flexible structure, 

In this paper, the objective is to examine the effectiveness using smart sensors and actuators for active 

vibration control on a 3-floor model building. Prior to control design and implementation, the 

experimental modal testing of the model building is performed to obtain the natural frequency and their 

corresponding modal shapes. PZT will be used as both sensors and actuators in this research. These three 

active vibration control methods, positive position feedback (PPF), strain rate feedback (SRF), and 

sliding-mode controller (SMC), are designed and implemented. 

2 VIBRATION SUPPRESSION METHODS 

For this research three vibration-suppression methods were identified viz. Positive Position Feedback 

(PPF), Strain Rate Feedback (SRF), and sliding mode control (SMC). 

2.1 Positive Position Feedback Control 

Positive Position Feedback (PPF) control was first proposed by Goh and Caughey (1985) as a robust 

control solution since it is insensitive to spillover phenomenon. It is not destabilized by the finite actuator 

dynamics and the stability is guaranteed by only considering the structures stiffness properties only. Later 

Song et al (2002) experimentally demonstrated PPF control in pultruded fiber-reinforced polymer I-beam. 

The PPF control is applied by feeding the structural position coordinate directly to the second order 

system. The position is then positively fed back to the structure. PPF offers quick damping for the 

controlled mode provided that the natural frequency and gain are known. The scalar equations governing 

the vibration of the structure in a single mode and the PPF controller are given as: 

Structure: £ 4- 2go> -h &r£ ~bu (1) 

Controller: 77 + 2q c O) c fj + 0777 = gO)^ (2) 

-^(fVl (3) 

where § , q and oo are the modal co-ordinate displacement, damping ratio, and natural frequency of the 

structure respectively. In the controller equation, r\, q ff , co c , g are the compensator co-ordinate 

displacement, damping ratio, natural frequency, and feedback gain of the controller. It can be proved 

(Friswell and Inman, (1997)) that the stability condition is satisfied if and only if

233 

g ^ w (4) 

Notice that the stability condition depends only on the natural frequency rather than the damping ratio of 

the structures. In order to achieve maximum damping effect in the PPF control, the natural frequency 

should be closely matched to that of the structure. 

2.2 Strain Rate Feedback Control 

Strain rate feedback control (SRF) is implemented by feeding the velocity co-ordinate to the compensator. 

The position co-ordinate of the compensator is then fed back with a negative gain to the structure. When a 

smart structure is involved using a collocated PZT actuator and sensor, this control is achieved by feeding 

the derivative of the voltage from the sensor, which is proportional to the strain rate, to the input of the 

compensator and applying the negative compensator output voltage to the actuator. The scalar equations 

governing the vibration of the structure in a single mode and the SRF controller are given as: 

-gQ} 2 7l (5) 

= a; t 2 f (6) 

The variables above are the same as those defined in the PPF control. The stability condition requires that 

the frequency in the second order compensator be greater than the controlled mode frequency of the 

structure. As compared with the PPF, SRF has a much wider active damping frequency region, which 

gives a designer some flexibility. Selecting a precise compensator frequency for SELF is not as critical as 

for PPF. As long as the compensator satisfies the stability condition, a certain amount of damping will be 

provided. SRF can also stabilize more than one mode given a sufficient bandwidth. A limitation of SRF is 

that the magnitude of the transfer function in the active damping region becomes extremely small very 

quickly. Therefore, the amount of damping provided SRF over a certain frequency range is limited. 

2.3 Sliding-Mode Based Robust Controller 

Sliding mode controller (SMC) technique is naturally robust with respect to the uncertainty in the 

structural parameters and external disturbances. In principle, SMC consists of the control law that 

switches with infinite speed to drive the system on a specified state trajectory, called the sliding surface, 

and has capability to keep the state on this surface. Clearly, this controller belongs to the non-linear robust 

controller. Among the design issues, the most important is to design the sliding surface and reduce the 

chatter phenomenon. The governing equation of the structure is the same as described in equation (1). The 

control force in the SMC framework is obtained by following the equivalent control method, which 

requires that the sliding surface r = £ + A£ and r = 0 , and the control force is called as the equivalent 

force. The parameter A is a positive constant. The control law can be expressed as 

u = -K D r~-pszt(ar) (7) 

where K D and r are positive constants. The parameter p called robust gain is maximum value of the 

nonlinear switching control action to account for the disturbance. And the sat function is a nonlinear 

saturation function imposing upper and lower bounds on a signal. The function is defined as 

Isgn(ar) 

otherwise 

(8)

234 

The purpose of the robust compensator (psat(ar)) is to reduce the undesirable so-called chattering 

phenomenon. Continuity of the control force between the upper bound and the lower bound of sliding 

surface reduces chattering. 

3 EXPERIMENTAL RESULTS OF ACTIVE VIBRATION SUPPRESSION 

3.1 Experimental Setup 

The control objective is to show the effectiveness of various vibration suppression strategies for a 3-floor 

model building by using smart sensors and actuators. The setup is shown in Figure 1. Two pairs of larger 

PZT patches are surface-bonded both support beams near their base ends. These PZT patches are used as 

actuators to excite the building and to enable active control of the model building vibration. There are also 

four smaller PZT patches surface-bonded on one the support beam of 

the model building. These four PZT patches act as sensors for the 

feedback of the signal in the active control algorithms. The sensor 

signal from the bottom floor (near the actuator pair) of the model 

building is used for feedback control. The sampling frequency for real 

time control was set to Ik Hz. 

Fig. 1 Experimental setup 

In the research, various vibration control algorithm are used to control 

vibration of the model building. The modal parameters of the first two 

modes have been determined by experiment before control design and 

implementation. The fundamental frequency and the corresponding 

damping ratio were found experimentally to be 7.81Hz and 0.0081 

respectively. The second natural frequency is 24.0Hz. In each test, the 

system is excited by a combination of a sinusoidal signal at the first 

modal frequency and a white noise for the initial 5 seconds. Various 

control methods are then applied to control the induced vibration. The 

uncontrolled response of model building is shown Figure 2. 

3.2 PPF Experimental Results 

The positive position feedback control was implemented using the experimental set up described in the 

previous section. The first mode vibration control is investigated first. Different values of gains were 

tested. The PPF controller-damping ratio q c was set to 0.4 and controller frequency co c was set at 49 

rad/sec (7.8Hz). The controller-damping ratio q c was chosen as a compromise between damping 

effectiveness and robustness. 

Figure 3 shows the corresponding PPF controlled vibrations of the model building. The PPF 

controlled vibrations of the model building damp out within 5 seconds after excitation. The PSD plot 

for 5-15 seconds is shown in Figure 4(a). A drop of 51dB as compared to the uncontrolled energy level is 

observed during this period of active control. The corresponding damping ratio was calculated and found 

to be 0.0499, as compared with the free vibration damping ratio of 0.0081. This represents an increase of 

516%. A number of tests were conducted with variation in the PPF gain while keeping c, c at 0.4 and co c at 

49 rad/sec. The results are shown in table 1, which shows that maximum damping is achieved at the gain 

of 15. 

PPF control is then extended to control several modes simultaneously. If the first two modes are under 

control, the modal selectivity is also achieved by tuning the compensator frequency co c on the targeted

235 

modes. An energy drop will be obtained in the second mode as the first mode. Figure 4(b) shows that the 

energy drops about 17 dB in the second mode when compared with the uncontrolled system response. 

with control 

time(s) 

Fig. 2 Free response of the building 

time(s) 

Fig 3 PPF controlled system response 

Table 1 PPF result for 1st mode 

Gain 1st modal dB drop Controlled damping ratio % Change in damping 

5 

10 

12.5 

15 

44.75 

50.9 

50.4 

48.8 

0.0313 

0.0467 

0.0499 

0.0519 

286.42 

476.54 

516.05 

540.74 

with control 

without control 



(a)PPF with first mode under control (b) PPF with first two modes under control 

Fig. 4 PSD plot comparison for PPF control

236 

3.3 SRF Experimental Results 

In this research, the strain rate feedback control is designed to control the vibration of the fundamental 

mode. The SRF controller-damping ratio q c was set at 0.10, controller frequency co c was set at 62.8 

rad/sec (lOHz), which is greater than the targeted mode of 7.62Hz and the effectiveness of SRF controller 

at various gains was tested. The results are shown in table 2, with maximum energy drop achieved at gain 

of 0.12. 

SRF 

Gain 

0.12 

0.15 

0.17 

0.20 

Modal dB drop 

j-ij sec 

52.66 

52.80 

50.35 

49.12 

Table 2 SRF Results 

Contn)Ued damping ratio 

0.040 

0.042 

0.043 

0.045 

% Change in damping 

387.7 

423.4 

435.8 

451.2 

The excitation in the SRF experiment is the same as that in the PPF experiment. The SRF controlled 

system response is depicted in Figure 5. The building vibrations disappear within 5 seconds after the 

excitation signal is stopped. When compared to the PPF controller, the results are almost as dramatic as 

that of PPF. The PSD plot for 5-15 seconds is shown in Figure 6. The SRF control was accompanied with 

a drop of about 50 dB in the energy level of the model building in the period of active control. The 

damping ratio achieved was 0.045. As compared with the undamped ratio of 0.008, this represents an 

increase of 451%. 

with control 

CD 

H,. 

"55 

with control 

— - without control 

>,.30 

O) 

time(s) 


Fig. 5 SRF control response of the model 

building 

Fig. 6 PSD plots of uncontrolled and SRF 

controlled model building for 5-15 sec 

3.4 SMC Experimental Results 

The sliding mode control algorithm is also applied to control the vibration of the 3-floor model building. 

Since the first mode dominants the free response, the sliding mode control will suppress the first mode 

vibration. Following the method discussed by Adhikari and Yamaguichi (1997)), the controller 

parameters are A = 0.5, p = 1, and a = 1, and thus the sliding surface is r = £ + 0.5§ =0. 

In the research, the excitation of the model building is the same as those in the PPF and SRF control. The 

SMC controlled response is shown in Figure 7 and the corresponding PSD plot for 5-15 seconds is 

depicted in Figure 8. An energy drop of 51dB as compared with the uncontrolled energy level is observed 

during this period of active control. The vibration control of the first mode is comparable to that of the

237 

PPF controller. The second mode energy experienced a small increase rather than a decrease. This is 

mainly due to the switching nature of the SMC. 

In the controller law, the parameter p is set to be a constant of 1. The maximum control voltage is 

assumed to be 100V. The phase plane plot of the closed-loop system is shown in Figure 9 and the 

corresponding control voltage is shown in Figure 10. It can be clearly seen from the phase plane plot that 

the system trajectory converges to the equilibrium point. The control objective is achieved. 

with control 

with control 

without control 

time(s) 

Fig. 7 Sliding mode control response 

10 20 30 40 


Fig. 8 PSD plot of uncontrolled and SMC 

controlled model building for 5-15 sec 

"5 

o 

Phase plane plot 

Control voltage 

Displacement(V) 

Fig 9 Phase plane plot of the SMC 

time(s) 

Fig 10 Control voltage 

4. CONCLUSION 

The paper experimentally demonstrated the vibration suppression of a 3-floor model building using the 

positive position feedback (PPF), strain rate feedback (SKF), and sliding mode control (SMC). Since the 

same computerized excitation was used all the tests, the energy level drop in the power spectrum density 

plots as compared with uncontrolled case is used as one index to examine the effectiveness of the control 

method. The first mode is the dominant mode and is the main concern for vibradon control. Increased 

vibration damping is observed in all control designs. A maximum increase in damping ratio of 540% was 

obtained and energy level drop of 51dB was achieved in the case of PPF control. The PPF control was the

238 

most effective control strategy. Multi-mode PPF control targeting at the first two modes was tested and 

achieved energy level drop in both modes simultaneously. Compared to SRF control, the PPF controller 

will have most likely to become unstable due to the natural frequency change of the system. Research 

work is currently being conducted using adaptive positive position feedback control Sliding mode control 

in the research achieves energy level as much as those in the PPF control The main disadvantage of the 

sliding mode control is the excitation of high modes and relatively large initial overshoot. 

5. Acknowledgements 

The first author would like to thank the support provided by NSF via a CAREER grant. 

6. REFERENCES 

1. Goh C J and Caughey T K (1985), On the stability problem caused by finite actuator dynamics in the 

collocated control of large space structure, International Journal of Control, 41:3, 787-802. 

2. Fanson J L and Caughey T K (1990), Positive position feedback control for large space structures, Am. 

Inst. Aeronaut. Astronaut J., 28:4, 717-724. 

3. Agrawal B N and Bang H (1994), Adaptive Structure for Large Precision Antennas, 45th Congress of 

the International Astronautical Federation, Jerusalem, Israel, 

4. Song G, Schmidt S P and Agrawal B N (2000), Active vibration suppression of a flexible structure 

using smart material and modular control patch, Proc Institute of Mechanical Engineers, 214 Part G 

217-229. 

5. Song G, Schmidt S P and Agrawal B N (1998), Experimental Study of Vibration Suppression of 

Flexible Structure Using Modular Control Patch, Proceedings of IEEE Aerospace Conference, 

Snowmass, Co. 

6. Won C C, Sulla L, Sparks D W and Belvin W K(1994), "Application of piezoelectric devices to 

vibration suppression", J. Guidance, Control, Dynamics, 17(6), 1333-1338. 

7. Song G, Qiao P, Sethi V. Active Vibration Control of a Smart Pultruded Fiber-Reinforced Polymer 1- 

Beam, Proceedings ofSPIE9th Smart Material & Structural Conference (San Diego, CA), 2002. 

8. Manning WJ et a/,(2000) "Vibration Control of a Flexible Beam with Integrated Actuators and 

Sensors", Smart Material and Structures, 9, 932-939. 

9. Yang J.N, et al (1997) Sliding Mode Control With Compensator For Wind and Seismic response 

Control, Earthquake Engineering and Structural Dynamics, 26, 1137-1156. 

10. Allen M., et al (2000) "Sliding Mode Control of a Large Flexible Space Structure", Control of 

Engineering Practice, 8, 861-871. 

11. Sugavanam S, et al (1998) "Modeling and Control of a Lightly Dampe T-Beam Using Piezoceramic 

Actuators and Sensors", Smart Material and Structures, 7, 899-906. 

12. Preumont A (1996) "Vibration Control of Active Structures—An Introduction", Kluwer Academic 

Publishers. 

13. Dorf R.C, Bishop R.H (1998) "Modern Control Systems", Addison-Wesley. 

14. Adhikari R, Yamaguichi H,(1997) "Sliding Mode Control of Building with ATMD", Earthquake 

Engineering and Structural Dynamics, 26, 400-422. 

15. Friswell M.I, Inman DJ. (1997) "The Relationship between Positive Position Feedback and Output 

Feedback Controllers", Smart Material and Structures, 8, 285-291.




RHEOLOGICAL BEHAVIOR OF MR 

MATERIALS IN FLOW MODE 

X. Wang 1 , F. Gordaninejad 1 , G. H. Hitchcock 1 , 

A. Fuchs 2 , M. Xin 2 and G. Korol 3 

1 Department of Mechanical Engineering, University of Nevada, Reno, NV 89557, USA 

Department of Chemical Engineering, University of Nevada, Reno, NV 89557, USA 

3 Visteon Automotive Systems, Dearborn, MI 48126, USA 

ABSTRACT 

A flow mode type rheometer is developed to evaluate the tunable rheological properties of magnetorheological 

(MR) materials. The proposed method is simple, and provides measurement of the 

apparent viscosity and yield stress of MR materials. In addition, a theoretical shear stress-shear strain 

rate relationship is developed based on a non-Newtonian fluid flow analysis. The rheological 

properties of a commercial MR fluid and a newly developed magneto-rheological polymeric gel 

(MRPG) are investigated. Experimental shear stress results are presented for a shear strain rate range 

of 20s" 1 to 20,000s" 1 . The results for apparent viscosity, dynamic and static shear yield stress under the 

influence of different applied magnetic fields are reported. 

INTRODUCTION 

Typically, the design of an original equipment manufacturer (OEM) damper consists of one or more 

orifices which either are located within the piston or are external by-passes. For most applications, a 

magneto-rheological (MR) fluid damper can have a similar orifice design characteristic to an OEM 

damper (Gordaninejad and Breese, 2000, Gordaninejad and Kelso, 2000). Available measurements 

for rheological properties of MR fluids are mainly in the shear-mode and low shear stain rates (Li, 

2000). The focus of this study is to determine the shear stress of a MR fluid in the flow mode for a 

wide range of shear strain rates. This may be important for the force estimation and the controller 

design of a MR fluid damper where the working fluid operates at high shear strain rates. 

Recently, a piston-driven slit channel rheometer has been developed to estimate the shear yield stress 

of MR fluids under a magnetic field by Nakano et al. (1999). The slit rheometer utilizes an applied 

uniform magnetic field normal to the slit channel flow direction where forced convection assists in 

heat dissipation at high shear strain rates. However, it is difficult to determine the velocity gradient for 

a non-Newtonian fluid in such a device. 

In general, although it is important in some MR fluid based applications, relatively little attention has 

been paid to the behavior of MR suspensions in pressure driven flows. In addition, conventional 

rotational rheometry encompasses shear strain rates only up to a few thousand I/sec, while shear strain

240 

rates in most MR fluid devices can reach up to 20,000s' 1 . 

properties of MR fluids at high shear strain rates is warranted. 

Therefore, study of the rheological 

Base on these observations, a piston-driven flow type rheometer has been designed and built at the 

Composite and Intelligent Materials Laboratory of the University of Nevada, Reno, to evaluate the 

tunable rheological properties of MR fluids in flow mode. Three different MR fluid samples are 

investigated in this study. The apparent viscosity, dynamic and static shear yield stress, under the 

influence of applied magnetic fields, are studied. 

METHODS AND THEORETICAL ANALYSIS 

Figure 1 presents the piston-driven flow-mode rheometer system for MR fluids with a rectangular 

cross section. An electromagnet provides a magnetic flux normal to the slit channel flow. The coil is 

activated by a power supply in constant current mode. By using a Gauss meter, the magnetic flux 

density inside the channel is measured. 

Piston (connected to 

Instron 882 Is) 

:ro magnet 

Accumulator 

Figure 1. Schematic of the MR channel flow rheometer. 

The MR fluid is confined in the well-sealed channel cell, and is pressurized to generate a flow through 

the channel between the two parallel magnet poles. There is an external groove in the test channel in 

order to align the electromagnetic field over the test section. The control of the hydraulic source piston 

velocity is achieved by connecting the piston-cylinder mechanism to an Instron 882IS servo hydraulic 

actuator. Two pressure transducers measure the pressure drop across the test section. The test section 

has dimensions of h= 1.0mm in height, w=10mm in width and L=14mm in length (Figure 2). The 

proposed MR fluid rheometer utilizes a nitrogen gas accumulator. The accumulator functions as a 

sink, when the piston is positively displaced, and a source, when the piston is negatively displaced, 

making it possible to measure the static and dynamic shear yield stress of MR fluids. 

Pressure P 

Magnetic field B 

Figure 2. Geometric dimensions of the MR rheometer channel.

241 

In order to determine the apparent viscosity, r\ aoo = —, one has to relate the shear stress r and shear 

r 

strain rate, y, with measured pressure drop, Ap, and volumetric flow rate, Q. If minor losses are 

small, the value of r at the wall can be determined by pressure drop across the channel, as follows: 

hAp f 7 

r = —£- for h«w (I) 

2/ 

It can be shown that for a one-dimensional, steady, non-Newtonian flow the shear strain rate can be 

expressed as: 

)=4g. 

h 2 wjsri 

where n = —. Equation (2) provides the relationship between t and y(r ) with the measured 

dlnO 

quantities Ap and 0. The term in the bracket is similar to the Rabinowitch correction used for 

conventional capillary viscometer (Zaman, 1998). For a Newtonian fluid n -I. For a non- 

Newtonian fluid, n can be determined from the experimental results of ln(r )V ) versus In(0//z 2 w). 

For a given flow rate, n is equal to the slope of the curve at that point. Linear regression methods can 

be used to obtain the best fit for \&(r w ) as a function of lu(Q/ h 2 w). 

EXPERIMENTAL PROCEDURE AND DATA ANALYSIS 

Experiment results are presented for three MR fluid samples that are: 1) A Hydrocarbon-based MR 

fluid (MRF-132LD, Lord Corp., USA), 2) A polyalphaolefm (PAO)-based Magneto-rheological 

polymer gel (MRPG) prepared by the Polymer Science Laboratory (PSL) at the University of Nevada, 

Reno, and 3) A Ferro fluid-based MR fluid prepared by PSL based on information in the literature 

(Ginder, 1996). 

Magneto-rheological polymer gels are a type of MR material, which are prepared by suspending iron 

particles in polymeric gels. These composite polymeric materials allows the control of viscosity, 

provide high shear yield stress and exhibit low particle settling behavior through different 

combinations of resins and crosslinkers (Wilson et al., 2002). PAO based MRPG is prepared using 

polyalphaolefin as the carrier fluid with polymer/carrier fluid ratios from 10:100 to 50:100. The PAO 

MRPG contains, 80% by weight, carbonyl iron particle with 99.5% purity. Ferrofluid-based MR fluid 

employs a magnetizable carrier fluid, such as a ferro fluid to obtain higher shear yield stress. 

All the tests were performed using INSTRON servo-hydaulic test rig Model 882IS at room 

temperature. Before conducting the experiments, the accumulator was pressurized to 300-900kPa 

(50~150psi) via a nitrogen tank. The displacement input was a double ramp profile to generate 

constant piston velocities; hence, constant flow rates. Each set of tests included different velocities, 

which covers shear strain rates from 20s" 1 to 20000s" 1 . The electromagnetic coil was energized with 

current for each velocity at 0 A, 0.5A, l.OA and 2.0A, corresponding to O.OmT, 120mT, 250mT and 

350mT magnetic flux densities (Bo), respectively. A ramp input displacement was used to keep the 

piston moving with a constant velocity, thus, the time variations of pressure drop across the measured 

channel were almost constant during each measurements. 

Typical results for the input and output profiles are presented in Figure 3. As can be seen in Figure 3, 

initially, an internal pressure about 300kPa (50psi) was applied by the accumulator. A pressure drop 

v ;

offset about 30kPa (6psi) across the channel test section was observed. For the first ramp, the piston 

pushes the MR fluid through the channel with a constant velocity with an externally applied magnetic 

field. As a result of the MR effect, the pressure drop (Ap=Pi-P 2 ) across the magnetically activated 

region increases significantly. As illustrated in Figure 3, we define this pressure drop as the dynamic 

pressure drop, since its value depends on the combination of the applied magnetic flux density and the 

input velocity of the piston. In this study, the dynamic pressure drop experimental data can establish 

the shear stress and shear strain rate relationship needed to obtain the apparent viscosity and dynamic 

yield stress of MR fluids. 

For the second ramp, the piston returns to its original position. Both the pressures P! and P 2 decreased 

while the accumulator pushes the MR fluid back. In addition, PI drops more than its initial pressure. 

This means that a vacuum may have been formed in the inlet chamber. The material resistance due to 

the MR effect may prevent the MR fluid returning back completely. A negative pressure drop is 

obtained at this phase. We refer to this pressure drop as the static pressure drop, as shown in Figure 3. 

The static pressure drop is highly dependent on the magnetic flux density and nearly is unaffected by 

the piston movement. 

When the piston is completely stopped, after returning to its initial position, the static pressure drop 

can be kept in equilibrium until the magnetic field is switched off. After a short period following the 

input electric current being turned off, the pressures PI and p2 recover to their original values, as 

shown in Figure 3. This would suggest that the static pressure drop is the minimum pressure that can 

induce MR fluid flow. Therefore, using Equation (1), the static shear yield stress of MR fluids could 

be evaluated from the measurement data of the static pressure drop. 

242 

2000.00 0.40 

-500.00 -1.60 

I I I ~ 

0.00 5.00 10.00 15.00 20.00 25.00 

Time (&\ 

Figure 3. A typical measured result of piston displacement and flow 

pressures across the rectangular channel.

243 

RESULTS AND DISCUSSIONS 

Figure 4 shows the measured apparent viscosity versus shear strain rate for Lord MRF-132LD fluid at 

various magnetic fields for the range of 20s" 1 to 20,000s" 1 . These results are compared to those 

obtained by a commercial parallel-plate MR fluid rheometer (Li, 2000). As can be seen from Figure 4, 

the results of both studies agree well in'the range of 20 s* 1 to 200 s" 1 where the experimental data of 

shear strain rates for these two different rheometers overlap. In addition, the experimental results show 

a linear relation in the log-log scale, which implies that there is a power-law dependence (i.e., 

rj app cc y~ m ) of the apparent viscosity, rj app , on the shear strain rate, /, for MR fluids. A shear thinning 

effect was observed as the apparent viscosity decreases with the increase of shear strain rate for all 

three MR fluid samples. At the "on-state" (magnetic field is activated), all exponent m values are in 

the range of 0.60-0.94. "m" increases with increasing the magnetic flux density. Similar phenomena 

were also reported for monodispers electro-rheological fluids (Halsey, et al., 1992) and monodispers 

MR fluids (Felt, et al., 1996). At the "off-state" (with no applied magnetic field), ferrofluid-based MR 

fluid behaves as a Newtonian fluid, when m is close to zero. The PAO based MRPG shows a weak 

shear thinning effect where its power law exponent m is much less than that of the "on-state". 

Typical results of the shear stress versus shear strain rate for the Lord MRF-132LD fluid are shown in 

Figure 5. The dynamic shear yield stress is obtained by extrapolating shear stress data to intercept the 

shear stress axis based on the Bingham plastic model. For an increase in magnetic flux density, a 

similar increase in the respective shear stress is observed. 

Figure 6 shows the experimental results of shear stress dependence on the magnetic flux density for 

three MR materials. The static shear yield stress was obtained by evaluating the experimental data of 

the static pressure drop employing Equation (1). Note that the dynamic shear yield stress exceeds the 

static case over an entire range of magnetic fields for all three samples. These results are consistent 

with Kordonski et al. (1999) and Li (2000) experimental studies where they examined MR fluid in 

shear mode. Volkova et al (2000) analyzed the existence of the two different shear yield stresses for 

two types of magnetic particle suspensions in the presence of a magnetic field. They explained that 

the dynamic shear yield stress (whic they referred to it as Bingham yield stress) is associated with the 

rupture of the aggregate which reform in the presence of the magnetostatic forces, while the static yield 

stress (which they referred to it as frictional yield stress) is associated with solid friction of the 

particles on the plates of the rheometer. Therefore, there is no logical reason for the static and the 

dynamic yield stress to have the same value. 

SUMMARY AND CONCLUSIONS 

A flow-mode rheometer is designed and built to evaluate the tunable rheological properties of MR 

materials in the presence of a magnetic field. This instrument can be used to study the apparent 

viscosity of MR fluids for a wide range of shear strain rates, from 20s" 1 to 20,000s" 1 . A Rabinowitch 

correction for the slit channel flow is developed to obtain the shear strain rate at the wall as a function 

of shear stress. By measuring the pressure drop as a function of flow rate, the apparent viscosity is 

determined. A unique design feature for the flow-mode MR fluid rheometer used in this study allows 

measurements of both the dynamic and static shear static yield stress of MR materials.

244 

1.00&05T — 

1.00E+04-S 

o. 100E^03-E 

w 

o 

» 1QQE+Q2-: 

« From U, Bmrf=370mT 

• Channel Rheometer, BO=350mT 

A From U, Bmrf=270mT 

X Channel Rheometer, BO=250mT 

• From U,Bmrf=170mT 

• Channel Rheometer, BO=120mT 

£ 

8. 1oofr-01 

CL 

1.00E+OQ-: 

I OOE-01 

1.00E-K30 1.00E-K31 100E+02 1.00E+03 1.00E-K34 1.00E+05 

Shear Strain Rate (1/s) 

Figure 4. Log-log plot of the apparent viscosity of Lord MRF-132LD fluid 

versus shear strain rate for various magnetic fields. 

2000 4000 6000 

8000 10000 

Shear Strain Rate (1/s) 

Figure 5. Shear stress versus shear strain rate measurea ai rour different magnetic fields 

for Lord MRF-132LD. The straight lines represent the curve fit of the 

experimental data with using the Bingham plastic model

245 

100 150 200 250 300 

Magnetic Flux Density, (mT) 

350 400 

Figure 6. Comparisons of shear yield stress versus magnetic field for three MR 

materials. Hollow marks are the dynamic shear yield stress and 

solid marks are static shear yield stress. 

The performance of the device is compared to a rotational shear-mode rheometer. The results indicate 

that the measurements obtained by the flow-mode device are consistent with rotational shear-mode 

rheometer. Three different MR materials are examined under various magnetic flux densities. The 

apparent viscosities, dynamic and static shear yield stresses are determined. All materials examined 

show a strong shear rnrnning effect. A power-law relationship was obtained for the apparent viscosity 

versus shear strain rate. The dynamic shear yield stress exceeds the static shear yield stress over the 

entire range of magnetic flux densities for all the three MR materials. 


This study in part is supported by the U.S. Army Research Office and Visteon Corporation, USA. The 

authors at the University of Nevada, Reno are thankful for their support. 

REFERENCE 

Ginder, J. M. (1996). Rheology Controlled by Magnetic Fields. Encyclopedia of Applied 

Physics, 16, 487-503. 

Felt, D. W. 9 Hagenbuchle M., Liu J, and Richard, J. (1996). Rheology of a magnetorheological 

fluid. /. Int. Mat Sys. Struct 7 :5, 589-593. 

Gordaninejad, F. and Breese, D. G. (2000). "Magneto-rheological fluid dampers." U.S. Patent 

No. 6,019,201.

246 

Gordaninejad, F. and Kelso S. P. (2000). Fail-safe magneto-rheological fluid dampers for offhighway, 

high-payload vehicles. J Int. Mat. Sys. Struct., 11:5, 395-406. 

Halsey, T. C, Martin, J. E. and Adolf, D. (1992). 

Physical Review Letters, 68 :10, 1519-1522. 

Rheology of Electrorheological Fluids. 

Kordonski, W. I. and Goiini, D. (1999). Fundamentals of magnetorheological fluid utilization 

m high precision finishing. J Int. Mat Sys. Struct. 10 :9, 683-689. 

Li, W. H. (2000). Rheology of MR Fluids and MR Damper Dynamic Response: Experimental 

and Modeling Approaches, Ph.D. Dissertation, School of Mechanical and Production Engineering, the 

Nanyang Technological University, Singapore. 

Nakano, M., Yamamoto, H., Jolly, M. R. (1999). Dynamic viscoelasticity of a 

magnetorheological fluid in oscillatory slit flow. International Journal ofMordern Physics B 13:14- 

16, 2068-2076. 

Volkova, O., Bossis, G., Guyot,M., Bashtovoi, M. and Reks, A. (2000).Magnetorheology 

of magnetic holes compared to magnetic particles. J. Rheology 44:1, 91-104. 

Wilson, M., Fuchs, A. and Gordaninejad, F (2002). Development and Characterization of 

Magnetorheological Polymer Gels. J. Appl. Polym. Scf,, 84:14, 2733. 

Zaman, A. A. (1998). Techniques in Rheological Measurements: Fundamentals and 

Applications, NSF Engineering Resource Center for Particle Science & Technology, University of 

Florida, Gainesville, Florida.




INNOVATIVE APPROACHES FOR STRUCTURAL HEALTH 

MONITORING OF INTELLIGENT INFRASTRUCTURE SYSTEMS 

Raymond W. Wolfe 1 , Sami F. Masri 2 , and John Caffrey 2 

Supervising Bridge Engineer, California Department of Transportation, Sacramento, CA, USA 

2 Department of Civil Engineering, University of Southern California, Los Angeles, CA, USA 

ABSTRACT 

Intelligent infrastructure components (e.g., MR dampers) possessing adaptive features 

are being seriously considered for deployment in civil infrastructure systems in highly 

seismic regions throughout the world. Due to the long service life and harsh 

environments that such components may be subjected to, it is essential to have a 

reliable and efficient procedure for condition assessment of these components based 

on the analysis of their vibration signature. 

This paper presents an overview of some promising approaches based on nonlinear 

system identification techniques to detect slight changes in the characteristics of 

nonlinear dampers of the type commonly encountered in structural control 

applications involving large civil infrastructure systems such as bridges. By 

characterizing the restoring-force surface of the damper in a nonparametric form, and 

subsequently analyzing the higher-order statistics of the coefficients defining such 

surfaces, it is found that the associated probability density function of the identified 

coefficients furnishes a sensitive indicator of the underlying damper parameters. Both 

simulation results as well as experimental measurements from a prototype nonlinear 

viscous damper are presented to illustrate the approach and discuss its range of 

validity. Sensor requirements for field implementation are also discussed. 


Large-scale civil infrastructure design and construction present an interesting class of 

problems related to the large dynamic demands from wind and seismic events. 

Engineers are searching for new ways to handle the large forces and displacements 

generated from such events. Strength design concepts are yielding to more elegant 

and efficient energy dissipative devices such as large-scale viscous dampers. The 

reliance on these devices to dissipate energy stemming from wind and seismic 

occurrences makes them integral components to the success of the design to withstand 

such events. Failure of a viscous damper can portend potentially catastrophic 

localized or large-scale system failure, as the adjoining members are sized based on 

the energy absorbed by the dampers. Given the criticality of the damper elements to 

the success of the design strategies being implemented on these large structures, a 

means of evaluating their in-situ health is imperative.

248 

INNOVATIVE APPROACHES TO STRUCTURAL HEALTH MONITORING 

Scope 

The focus of this paper is to investigate the applicability of on-line condition 

assessment of structural damper components. A relatively simple approach developed 

by Masri and Caughy [1] allowing the identification of a broad class of dynamic 

models was in this effort. Complex issues typically associated with nonparametric 

identification methods, such as greater mathematical complexity, convergence 

difficulties, excessive computational effort, restrictions on the nature of the dynamic 

systems to be evaluated, and restrictions on the allowable system excitations are 

addressed through the application of regression techniques coupled with orthogonal 

polynomials. Results of analytical and experimental studies are presented in 

succeeding sections to substantiate the identification algorithm's robustness and 

applicability to real-time in-situ system health monitoring. 

2. SIMULATION STUDIES 

2.1. Overview of Nonlinear System Model 

Figure 2.1 Noise polluted damper simulation response, random excitation

249 

WOLFE, MASRI, CAFFREY 

Synthetic data was developed utilizing a nonlinear single-degree-of-freedom (sdof) 

Duffing oscillator. Manipulating the equation of motion yields the oscillator 

restoring force as 

f(x,x)=m[2£aK + G) 2 (x + £x 3 )] (1) 

Unit mass and system period were prescribed for this oscillator. The natural 

frequency was set at In. These values yielded the system stiffness, damping 

coefficient and nonlinear term of 39.48, 1.26 and 1.23, respectively. Noise pollution 

was incorporated into the simulated data by adding stationary, zero-mean, 0.10 

standard deviation noise. Potential noise sources in real-world applications include 

instrumentation susceptibility, cabling interference, acquisition hardware, etc. The 

system response to a stationary random excitation is depicted in Figure 2.1. 

2.2. Identification Procedure 

The restoring force method [1] was utilized to identify the system parameters from the 

system response. This method expresses the estimated restoring force in terms of 

two-dimensional orthogonal polynomials as 

where the T's are defined as Chebyshev polynomials, taking advantage of their equalerror 

approximation within the interval of interest. Chebyshev polynomials are 

defined as 

(3) 

satisfying the weighted orthogonality property 

0 

(4) 

The normalized displacement and velocity values are defined as 

-x_)/2]

250 


The normalized Chebyshev coefficients are generated as a result of fitting the 

computed restoring force from the prescribed data set containing the system 

displacement and velocity to the measureoVsimulated force response. One advantage 

of this method is that upon a relatively straight forward transformation of Equation 2 

to a Power Series expansion [3], the coefficients can be correlated to recognizable 

physical systems. This property allows a simple verification of the identified results 

against a known standard and facilitates understanding of the root cause of identified 

system degradation. 

23. Sample Results 

The identification algorithm presented above is easily adapted to statistical studies 

drawn from large ensemble data sets, enabling the establishment of identification 

results under system parameter uncertainty or variability. 

55155 11231 1684S5 22462 55262 110524 165.786 221.048 

(a) 

(b) 

Figure 2.2 Chebyshev coefficient (2,1), (a) 5% stiffness degradation with 10% noise 

pollution, (b) 5% epsilon degradation with 10% noise pollution 

The robustness of the identification algorithm in detecting subtle changes to system 

parameters was evaluated by implementing a Monte Carlo approach. System 

parameters were varied from the prescribed system mean values noted above. 3000 

simulations were then performed with a zero-mean random excitation. After 

computing the mean and standard deviation statistics of these runs, plots of individual 

parameters highlighted excursions from prescribed values, such as in Figure 2.2.

251 

WOLFE, MASRI, CAFFREY 

Transforming the computed Chebyshev polynomial coefficients to an equivalent 

Power Series representation allows direct comparison of individual parameter shifts. 

Figure 2.3 illustrates the mean shift of the identified system stiffness parameter when 

the simulated stiffness and Duffing epsilon values were degraded by 5% individually. 

As expected, the identification algorithm discnminates the system change, noting 

mean values of approximately 42.4, 40.0, and 42.5 for the reference, degraded 

stiffness and degraded epsilon cases, respectively. These values yield a stiffness 

degradation of 5.8% for the case wherein the stiffness parameter was degraded by 5%, 

and a 0.1% change in the identified stiffness parameter when the nonlinear Duffing 

term epsilon is degraded by 5%. These results confirm the ability of the algorithm to 

detect and discriminate system change even under noise pollution. 

0.03 

0.02 

0.01 

-10 15 40 65 90 

Figure 23 Comparison of degraded stiffness and epsilon mean shift to reference case 

(undegraded)

252 


3. EXPERIMENTAL STUDIES 

3.1. Test Description 

Validation of the results obtained through the above and additional simulation studies 

requires application to experimental data. To accomplish this, a test facility was 

developed to test a 10-kip nonlinear viscous damper with a 12-inch stroke and a 

velocity rating of 70 ips. The system was excited through an 11-kip MTS Systems 

actuator fitted with a 90 gpm servo valve to accommodate high velocities. Figure 3.1 

presents a digital image of the test set-up. The damper/actuator connection was 

fabricated with linear bearings to only allow motion along the axis of the damper, 

thereby avoiding out-of-plane bending. 

Figure 3.1 Digital photo of test set-up (damper is approx. 4 feet long) 

3.2. Sample Measurements 

Figure 3.2 depicts the measured system response of the damper excited by a 4-inch 

peak broadband signal. The highly irregular system response captured in the 

velocity/restoring force phase plot in the transition range reveals dead-space 

fluid/orifice dynamics internal to the damper. 

It is important to note that the damper component tested experimentally incorporates a 

relatively minor contribution of the stiffness parameter to the measured response. In 

contrast, the Duffing oscillator defined in the analytical studies included a significant 

stiffness to the restoring force. The simulations were conducted with a sdof 

descriptive model, interpreted as a model containing an elastic member and a damper

253 

WOLFE MASRI,CAFFREY 

in parallel [2] The tested damper element incorporates stiffness only in the 

seal/piston interface and sikcone fluid compressibility, with the fluid passing through 

internal orifices lending the primary system force response Hence, the test setup 

isolates the damper element, resulting in a damping dominated restoring force instead 

of the stiffness controlled force defined in the simulation efforts This difference is 

readily observed from a detailed review of the phase plots in Figures 2 1 and 3 2 

While the simulated Duffing oscillator system response is clearly heavily dependent 

on the stiffness term, the measured system response from the tested damper is largely 

influenced by the damping parameter contribution 

0 06 

displacement (inches) 

Figure 3*2 Measured nonlinear damper response, broadband excitation 

4200 

2800 

1400 

2800 

4200 

02 03 03 04 

Figure 33 Comparison of measured and identified system force

254 


CONCLUSIONS 

The application feasibility of system identification tools to condition assessment 

problems has been demonstrated in this paper. The restoring force algorithm based on 

orthogonal Chebyshev polynomials was proven successful in identifying and 

quantifying subtle system changes with both artificially noise-polluted simulation data 

and experimental data. A more in-depth treatment of the subject supporting these 

conclusions is detailed in [3]. 

jdfl* 

:- 

i 

u r 

£ ,B, 

•1, 

-2 - 

Figure 3.4 Identified restoring force surface 

REFERENCES 

1. Masri, S.F., and Caughey, T.K. (1979). A Nonparametric Identification 

Technique for Nonlinear Dynamic Problems. ASME Journal of Applied 

Mechanics 46:1, 433-447. 

2. Wolfe, R.W., Masri, S.F., and Caffrey, J. (2002). Some. Structural Health 

Monitoring Approaches for Nonlinear Hydraulic Dampers. Journal of 

Structural Control. 

3. Wolfe, R.W., (2002). Analytical and experimental studies of structural health 

monitoring of nonlinear viscous dampers. Ph.D. Dissertation. University of 

Southern California, Los Angeles, California, USA.

STRUCTURAL ANALYSIS AND DESIGN




DRIFT-BASED SEISMIC ASSESSMENT OF BUILDINGS 

IN HONG KONG 

A.M. Chandler 1 , R.K.L. Su 1 and M.N. Sheikh 1 

1 Department of Civil Engineering, The University of Hong Kong, 

Pokfularn Road, Hong Kong SAR, China 

ABSTRACT 

Large magnitude earthquakes generated at distances exceeding 100km are typified by low frequency 

(long period) seismic waves, since the high frequency components have greatly diminished in amplitude 

as a result of energy absorption along the wave travel path. The peak ground acceleration (PGA), or 

response spectral accelerations (RSA), from such distant earthquakes can be very low and yet the 

induced motion can be highly destructive due to its high displacement (drift), and possibly high velocity, 

shaking characteristics. This paper introduces a methodology for predicting the response spectral 

displacement (RSD) using the Component Attenuation Model (CAM). Hence, the critical drift demands 

from long distance earthquakes affecting tall buildings in Hong Kong have been estimated. It is shown 

that soil (site) resonance effects play a critical role in determining the level of seismic drift demand. 


Large magnitude, distant earthquakes represent the critical design events for the prevalent medium-rise 

and high-rise construction in Hong Kong (Lam et al, 2002). This paper firstly introduces attenuation 

relationships to predict the velocity and displacement (drift) demands from long distance earthquakes, 

based on stochastic simulations for the crustal conditions of South China. The Component Attenuation 

Model (CAM) procedure, so developed, is based on separating the "source", "path" and "site" effects. 

Such separation is intended to address the fact that attenuation of seismic waves is strongly dependent 

on regional properties of the earth's crust, through which seismic waves are propagated and modified. 

This regional crustal influence is particularly significant for seismic waves propagating over a long 

distance (site-source distance R> 100km). Response spectrum formats that can effectively represent 

such effects are shown in Fig's 1.1 a and Lib, and the parameters of interest are accordingly RSD max 

and RSV rtulx . Existing empirical response spectrum attenuation relationships from high seismic regions 

are typically based on strong motions recorded in the near-field. Thus, they cannot be relied upon to 

model distant earthquakes. The recorded ground shaking on different sites, and in different regions, can 

vary significantly even for very similar moment magnitude (M) and site-source distance, R (km). By

258 

separating the above three effects, such variability can be addressed in stages so that ground shaking 

prediction becomes easier to generalise. 

RSD 

(a) 

Miaximum Displacement (Drift) Demand RSD m 

RSD ^RSV^—^S L 

applies to rock spectrum for 

moderate-large events. 

suggested 

T=5secs 

Natural Period T^T 2 = TS for soil spectrum 

LogRSV 

Maximum Velocity (Kinetic Energy) Demand RSV inax 

ct 

(b) 

T,orT s T=5secs Velocity Demand (RSV) 

Log Natural Period T 

Fig. 1.1: Response spectra and response spectral parameters 

2. PATH AND SITE EFFECTS MODELLING 

The M-dependent source properties a(M) of the CAM model have been dealt with by Lam and 

Chandler (2002). The attenuation of seismic waves is significantly more regionally dependent than 

source properties; thus, it is inappropriate to propose a generalised relationship. Ideally, attenuation is 

best modelled by correlating locally recorded strong motion with distance, using a large representative 

database. A direct approach is often not achievable, due to the chronic lack of strong-motion data in 

regions of low to moderate seismicity such as South China (including Hong Kong). An alternative 

viable approach is to estimate the effects of individual wave modification mechanisms based on known 

geophysical parameters (which do not require the recording of strong motion for their determination). 

The combined path effects (which exclude local site modifications, dealt with below) can be expressed 

as the product of the following component factors: (i) geometrical attenuation factor G(R,D), (ii) 

anelastic attenuation, or energy absorption, factor $R,Q) 9 (iii) the mid-crustal amplification factor 

JmcWwee), and (iv) the upper crust modification factor y uc ; where the associated parameters are the 

crustal depth (D, in km), the shear wave velocity gradient of the earth's crust and the wave 

transmission quality factor (Q). The component factors as summarised in Lam and Chandler (2002) 

were derived by curve-fitting results from stochastic simulations of regional seismological parameters. 

Shaking from distant earthquakes can also induce very significant soil amplification effects, as a result 

of the robust transmission of high period energy (in the velocity and displacement-sensitive parts of the 

response spectrum) over long distances. The 1985 Mexico City and 1989 Loma Prieta, California 

earthquakes offer well-known recent examples of this phenomenon. The site response component

259 

factors briefly described in Table 2.1 have been based on the "frame analogy soil amplification" (FASA) 

model, as described by Lam et al (2001). The term Vîs the shear wave velocity of the bedrock (ra/s). 

The basis of FASA is that the amplitude of the soil spectrum (soil RSV max ) can be obtained by scaling 

from RSVnuu of the corresponding rock spectrum at the natural period of the site (T s ). This modelling 

concept is supported by results obtained from non-linear wave analysis and by Seld observations. 

Demand Parameters 

RSD max 

RSV max 

H 

1.2 

1.2 

TABLE 2.1 

SITE FACTOR S(T S v s( > p. ox t 

a 

X 

4-6 

4-6 

0.65 + -i~< 10 

10000 

V 

0.65 + — *£—

260 

200 300 400 

SHAKE Computed RSD^ and RSV^ 

Fig. 2.1. Comparison of FASA and SHAKE in predicting maximum response spectral velocity and 

displacement at the soil surface (Hong Kong sites) 

Eqn.(2.1) is also known as the Component Attenuation Model (CAM), initially developed by the 

authors in Lara et al (200Qa,b). The attenuation relationships are for predictions of the mean. 

Log Natural Period (log T) 

Fig. 2.2. Soil and rock response spectra modelled by CAM and FASA procedures

261 

3. GROUND SHAKING PREDICTIONS FOR HONG KONG 

Based on results presented in Lam et al (2002) and Chandler et al (2002b), Table 3 1 summarises the 

5% damped response spectrum predictions for design-level earthquakes affecting Hong Kong, 

comprising critical very far-field events (R=280km) The results have been based on CAM model for 

rock sites and FASA scaling model for soil sites 

TABLE 3.1 

RECOMMENDED PARAMETERS FOR SEISMIC RESPONSE SPECTRA FOR HONG KONG 

Return 

Period 

(years) 

PE 

750 

years 

Rock 1 

RSV max 

(mm/s) 

Rock 2 

RSD mx 

(mm) 

S-Factor 

for Soil 

RSV raax 

Soil 3 

RSV^ 

(mm/s) 

S-Factor 

for Soil 

RSDan 

Soil 4 

RSDo^ 

(mm) 

70 

500 

1000 

2500 

50% 

10% 

5% 

2% 

34 

70 

84 

100 

12 

25 

30 

36 

65 

57 

55 

54 

220 

400 

460 

540 

29 

26 

24 

24 

35 

64 

73 

86 

1 for periods T falling between CAM bedrock response spectrum comer periods TI and ^T 2 (refer Fig. 1.1) 

2 for periods T = 5 0 sec (Fig. 11) 

3 for penods T falling between TI and site period T s , TI = corner period on Soil RSV response spectrum (Fig 2 2) 

4 results correspond to period T = site period T s , values quoted are for T s = 1.0 sec and can be scaled in proportion to T 5> 

provided T s < X m T 2 : X m = 1 - (M-5)/6 [Lam and Chandler (2002)] 

4. DRIFT PREDICTIONS FOR HONG KONG BUILDINGS 

In the recently established displacement-based approaches for seismic assessment of buildings, for 

example Priestley (1995), the overall drift demand at effective height of building (or at roof level) is 

compared with the corresponding drift capacity curve in order to assess the degree of damage as well 

as the factor of safety of the building This approach cannot however distinguish whether the drift is 

uniformly distributed at each storey or is concentrated at a particular storey that may be caused, for 

example, by the presence of a soft storey or by higher mode effects It is important to recognize that 

local deformation associated with high inter-storey drift would cause significant damage to the nonstructural 

(NS) components and high internal forces to develop in particular structural elements such as 

columns and coupling beams In this section, which starts with the estimation of overall lateral seismic 

drift demand, a prediction formula for maximum inter-storey drift demand of buildings with due 

consideration of higher mode effects and potential soft storey effects will be presented. 

The average building drift angle # a v g defined as the ratio between the roof seismic displacement and the 

height of building can be estimated by the equation [Chandler et al (2002b)] 

=165 RSD l 

(41) 

The maximum seismic inter-storey drift angle #max (due to the combined seismic vibration modes) can 

be related to the effective first-mode displacement RSD\ using the following generic expression:

262 

where H b is the building height and l max is the maximum dynamic drift factor determined by response 

spectrum analyses employing, for example, Fig. 2,2 and Table 3.1. Chandler et al (2002b) investigated 

19 RC buildings with a wide range of height, and established the following relationships for A max 

For rock sites: ^ = 2.72 + 0.497; < 9.5 (43) 

For soH sites: ^ = 2.72 + 0.67/ 5 < 9.5 (44) 

Both RSDi and A max depend on the building's fundamental lateral period T\. Su et al (2002) estimated 

the fundamental lateral period of typical dynamic computer models for buildings in Hong Kong and 

found the following estimation formula for fundamental lateral period of bare frame models: 

r 6/1 = 0.025tf, (45) 

The initial fundamental lateral period of real buildings (or denoted by full frames) with due 

consideration of non-structural components and the variation of mean and design elastic modulus of 

concrete was measured by ambient vibration tests [Su et al, 2002]. The following period prediction 

equation for full frames was suggested 

7^=0.013^ (4.6) 

Chandler et aL (2002a) further considered the reduction of initial stiffness of the NS components and 

the inelastic deformation of structural components under seismic dynamic deformation and proposed 

the following period prediction formula with consideration of period shift at large-amplitude vibration: 

Ti=T f 

(4.7) 

where p (=1.0) represents the ratio of lateral stiffness fromNS components and structural components, 

n\ (

263 

are assumed to be founded on a deep soil site, with seismic action having 2% PE/50 years. The interstorey 

drift predicted by eqn's (4.2) and (4.8) (denoted as predicted real buildings) using an iterative 

procedure is also displayed in Figure 4. lb for comparison. The period shift factor £ is found to vary 

with building height from 1.04 to 1.17 with mean value of around 1.09. The maximum demands of drift 

angles 0 a v g and i9 max are 0.24% and 0.45%, respectively, for real buildings of height -75m. 

0.5 

04 - 

0.3 - 

0.2 - 

0.1 - 

0 

A Bare frame models 

* Fullframemodels 

- -' Predicted real buildings 

0.8 - 

0.7 - 

0.6 - 

0.5 - 

0.4 - 

0.3 - 

0.2 - 

0.1 - 

0.- 

A A * 

A Bare frame models 

* — .. Fullframemodels 

1 1 

Predicted real buildings 

0 50 100 150 200 250 300 350 400 450 0 50 100 150 200 250 300 350 400 450 

Figure (a) 

Height^ _ _ Height^ 

Figure (b) 

Fig. 4.1. Variations of drift angles against building heights of deep soil site with 2% PE/50years 

(a) 0 avg and (b) # max 

TABLE 4.1 

RELATIONSHIP BETWEEN PERIOD SHIFT FACTOR Q AND 0 m 

O max (%) 

0.0 

0.2 

0.5 

1.0 

1.5 

€ 

1.00 

1.03 

1.19 

1.64 

2.95 

5. CONCLUSIONS 

1. Long distance attenuation relationships for the maximum response spectral displacement (RSD max ) 

and the maximum response spectral velocity (RSVma*) for rock and soil sites have been developed 

using the CAM-FAS A framework. 

2. The input parameters to CAM and FASA can be obtained from seismological monitoring, as 

opposed to capturing strong motion data. Thus, reliable distant earthquake attenuation relationships 

that account for regional characteristics can now be obtained for low to moderate seismic regions 

such as South China (including Hong Kong). 

3. The maximum demands of drift angles # avg and 6^ of buildings founded on deep soil sites in Hong 

Kong with consideration of partial damage of structural and non-structural components are found 

to be 0.24% and 0.45%, respectively, for buildings with height of around 75m.

264 

6. ACKNOWLEDGEMENTS 

The work described in this paper has been supported by the URC (University Research Committee) of 

The University of Hong Kong through Seed Funding for a Basic Research Grant (2002-2003), and also 

by the Research Grants Council of Hong Kong (Project No.'s HKU 7023/99E and HKU 7002/OOE), 

whose support is gratefully acknowledged. The research collaboration with Dr Nelson Lam of The 

University of Melbourne on topics covered in this paper is also acknowledged with thanks. 

7. REFERENCES 

Chandler, A.M., Su R.K.L. and Lam, N.T.K. (2002a). Review of Displacement-Based Seismic 

Assessment of Building Structures, Joint Internal Report, Department of Civil Engineering, The 

University of Hong Kong and The University of Melbourne, Australia. 

Chandler, A.M., Su R.K.L. and Lee, P.K.K. (2002b). Seismic Drift Assessment for Hong Kong 

Buildings. Proceedings of the HKIE Structural Division Annual Seminar, "Recent Developments in 

Earthquake Engineering", 1-15. 

Chandler, A.M., Lam, N.T.K. and Sheikh, M.N. (2002c). Response Spectrum Predictions for Potential 

Near-Field and Far-Field Earthquakes Affecting Hong Kong: Soil Sites. Journal of Soil Dynamics & 

Earthquake Engineering (in press). 

Lam, N.T.K. and Wilson, J.L. (1999). Estimation of Site Natural Period from a Borehole Record. 

Australian Journal of Structural Engineering SE1:3, 179-199. 

Lam, N.T.K., Wilson, J.L, Chandler, A.M. and Hutchinson, G.L. (2000a). Response Spectral 

Relationships for Rock Sites Derived from the Component Attenuation Model. Journal of Earthquake 

Engineering and Structural Dynamics 29:10, 1457-1490. 

Lam, N.T.K, Wilson, J.L, Chandler, A.M. and Hutchinson, G.L. (2000b). Response Spectrum 

Modelling for Rock Sites in Low and Moderate Seismicity Regions Combining Velocity, Displacement 

and Acceleration Predictions. J. Earthquake Engineering and Structural Dynamics 29:10, 1491-1525. 

Lam, N.T.K., Wilson, J.L. and Chandler, A.M. (2001). Seismic Displacement Response Spectrum 

Estimated from the Frame Analogy Soil Amplification Model. J. Eng. Structures 23, 1437-1452. 

Lam, N.T.K., Wilson, J.L, Chandler, A.M. and Hutchinson, G.L. (2002). Response Spectrum 

Predictions for Potential Near-Field and Far-Field Earthquakes Affecting Hong Kong: Rock Sites. 

Journal of Soil Dynamics & Earthquake Engineering 22, 47-72. 

Lam, N.T.K. and Chandler, A.M. (2002). Prediction of Displacement and Velocity Demands of Long 

Distance Earthquakes. Proc. 12 th Euro. Conf. on Earthquake Engineering, London, 10pp. (in press). 

Priestley MJ.N. (1995). Displacement-based seismic assessment of existing reinforced concrete 

buildings, Proc. 5th Pacific Conf. on Earthquake Engineering, Melbourne, Australia, 225-244. 

Sheikh, M.N. (2001). Simplified analysis of earthquake site response with particular application to low 

and moderate seismicity regions. M.Phil thesis, Dept. of Civil Eng., The University of Hong Kong. 

Su, R.K.L., Chandler, A.M., Lee, P.K.K., To, A.P. and Li, J.H. (2002). Dynamic testing and modelling 

of existing buildings in Hong Kong, Trans. ofH.K. Institution of Engineers (submitted Apr 2002).




UNDERSTANDING OF THE SEISMIC PERFORMANCE OF 

ASYMMETRIC R/C BUILDING STRUCTURES 

Junwu DAI 1 Yuk-Lung WONG 2 and Minzheng ZHANG 1 

1 Institute of Engineering Mechanics, China Seismological Bureau, China 

Department of CSE, The Hong Kong Polytechnic University, Hong Kong 

ABSTRACT 

This paper is composed by the following investigations: 1) Using the time-history response of the base 

shear-torque of 3 asymmetric RC single-story-building systems under a set of earthquake simulation 

tests on shake table, both of the advantages and shortcomings of the base shear-torque (BST) surface 

(1C. De La Llera and A.K. Chopra) established on the assumption of idealized force-displacement 

relationship are checked. 2) Based on the experimental results and the assumption of the bi-linear 

force-displacement relationship, a more reasonable and effective analytical model, the state space of 

base shear-torque is developed for estimation of the seismic performance particularly the torsional 

effects of the RC asymmetric structures. 

KEYWORDS 

earthquake simulation test, torsional, asymmetric, seismic response, inelastic, shear-torque relationship 

BACKGROUND AND OBJECTIVES 

It has been observed repeatedly in strong earthquakes that the presence of asymmetry in the plan of a 

structure makes it more vulnerable to seismic damages. There are reports of extensive damages to 

buildings that are attributed to excessive torsional responses caused by asymmetry in earthquakes such 

as the 1972 Managua earthquake, the 1989 Loma Prieta earthquake, the 1995 Kobe earthquake and the 

1999 Chi-Chi earthquake. Although torsion has been recognized as a major reason for poor seismic 

performance of asymmetric buildings and many studies have been done on this topic, the analytical and 

experimental studies on the inelastic seismic response of asymmetric buildings do not have a long 

history. Recently, De La Llera and Chopra (1995) developed a simple model for analysis and design of 

asymmetric buildings based on the assumption of idealized elasto-plastic force-displacement 

relationship. Each story of the building is represented by a single super-element in the simplified model. 

In this method, the story shear and torque interaction surface (Kan and Chopra 198L Palazzo and 

Fraternali 1988, De La Llera and Chopra 1995) is used as an important component. The story shear and 

torque (SST) surface is basically the yield surface of the story due to the interaction between story shear 

and torque. Each point inside the surface represents a combination of story shear and torque that the 

story remains elastic. On the other hand, each point on the surface represent a combination of shear and 

torque that leads to the yielding of the story. How realistic is such a model to represent the behavior of 

ductile buildings in seismic regions is a subject that requires further investigation.

266 

The objectives of this paper presented here are: 1) to check the limitations of the story shear-torque 

surface based on the assumption of the idealized force-displacement relationship with experimental 

results of earthquake simulation tests for 3 asymmetric single-story structures and 2) to develop it to be 

more realistic three state spaces based on the bi-linear force-displacement relationship. 

SHEAR-TORQUE RELATIONSHIP 

There are three reinforced concrete single-story-building structures are designed to cater for the research 

objectives. All experiments are carried out on a single-axial shake table installed by MTS in HK 

Polytechnic University. Design detailing, instrumentations and ground excitations used in experiments 

have been described in reference [2]. 

& i 

} I 

iv 

! 

': D1/A1 on ground, D2DID4/A2A3A4 on roof | 

^03 Wl iDUD2) D4 « 

T* 3 T AUA^ A4 

X 

600mm 

' 

AY 

^ W1^ * 

•02 ; 1 

; I W3 I 

DI/A1 on ground, D2D3b4/A2A3A4 on roof | 

WD3 ADKD2) D4 9 

' 

A3 

IAKA2) 

A4 

X 

3 

& 

I W2 1 

: 

• ci t „ !„.. . 

"4 l^wVirn " M ^ O^^TUTI ^' 

i r 

• 

(D U Input (3) 

®~" 

1 

I W2 1 

MjCI | | 

1 

A 

•^ mnnm J L innnm 

£ If *" 

Cu U input (D 

Specimen 1/3 & Instrumentation 

Fig.l Specimen Configuration & Instrumentation 

Specimen2 & Instrumentation 

Verification of the shear-torque relationship 

Based on the assumption of idealized force-displacement relationship of structural members, J.C. De Le 

Llera and A.K. Chopra (1995) developed a useful and effective method for understanding of 

fundamental issues on seismic torsional phenomena of asymmetric structures. The most important 

component of this approach is the use of the idealized yield base shear-torque (BST) surface of the 

studied system. As examples, abovementioned 3 model structures studied in earthquake simulation tests 

are used to examine both of the validity and the shortage of the idealized yield BST surface here for 

understanding of the real seismic behavior of asymmetric systems correctly. 

According to the method of reference [3] developed, the BST surfaces of the 3 asymmetric model 

structures are founded respectively based on the assumption of idealized force-displacement relationship 

while only single-directional ground excitation is considered, shown in fig. 8 for convenience of 

comparison. The displacement-based approach that T. Paulay (2000) established is used in the 

determination of the srif&iess of structural members. From fig.2, obviously, it can be found that the BST 

surface of specimen2 expanded uniformly on the basis of specimenl while for that of specimenj, it is 

skewed and extended to the 1 st and 3 rd quadrant relatively. In fact, due to the strength of concrete used in 

the fabrication of specimen2 is higher than that of in specimenl, the strength of structural members in 

specimen2 is uniformly increased. Together with the wall wl being arranged at the outer edge of 

specimen! comparing with specimenl, the entire shear-torque resistant capacity is correspondingly 

increased in specimen2 than mat of specimenl. The result is the space covered by the BST surface of 

specimen2 is larger than that of specimenl. Similarly, the higher strength of concrete and the 

reinforcement ratio used in specimenj particularly in walls than that of in specimenl, intensified the

strength eccentricity and simultaneously increased the entire shear-torque resistant capacity of 

specimens. The associated consequence is that the BST surface of specimens is skewer and larger than 

that of specimen 1. This figure illustrates the different seismic capacity among different model structures 

directly. From the area covered by the BST surface of each specimen, it can roughly predict that the 

seismic behavior of each specimen will be changed differently in a certain way. 

267 

Fig.2 BST surfaces of 3 model structures 

b) Specimen2 

Time-history BST surface 

response 

c) Specimens 

Fig.3. Verification of the BST surface with time-history response of specimens 

in earthquake simulation tests

268 

The comparison between abovementioned theoretical results (solid line signifies the idealized BST 

surface of each specimens) and the experimental response (dotted area) in terms of the time history of the 

base shear-torque of each specimen is shown in fig.3a)~c). Each sub-figure dictates one case of ground 

excitation. For example, A0.14& E0.30g, L0.37g and N0.91g signify the time history response of base 

shear-torque in the case of ground excitation with intensity of artificial wave 0.14g, El Centro-NS 

component 0.30g, Loma Prieta record 0.37g and North Ridge record 0.91g, respectively. 

From fig.3 a) ~c), the following conclusions can be inferred: 

1) For most cases of the intensity of ground motions are not higher than about 0.47g, the time history 

response of the base shear-torque does not or only lightly exceed the area the BST surface encircled. But 

it exceeds the boundary of the BST surface greatly in cases of the intensity of ground motions higher 

than about 0 47g for all 3 model structures. These facts tell the truth that the assumption of the idealized 

force-displacement relationship is only roughly reasonable for evaluation of the most common concrete 

structures in region with potential lower to moderate seismic intensity. However, for other relatively 

more important structures particularly those located in region with higher potential seismic intensity, 

more reasonable models such as bi- or tri-linear force-displacement relationship based strength/stiffness 

degradation/harden model should be used case to case. Correspondingly, the idealized BST surface 

should be developed further to be more realized bi- or tri-state spaces model to estimate the seismic 

performance (torsional effects) of asymmetric structures in a more general sense. 

2) With the increase of the intensity level of the ground excitations, the time history response of the base 

shear-torque tends to be developed to the longest inclined side of the BST surface. From the definition of 

the BST surface according to reference [3], these two longest inclined sides signify the mechanism of 

other lateral resistant planes (the left column even the central wall planes) are damaged when the 

strongest right wall plane remains linear elastic. The experimental observation proved the fact that the 

damage sequence of all 3 specimens are started from the left column plane to the right wall plane at last 

although there are obvious difference among the seismic damage type of them. These experimental 

results testified that the theoretical results are reasonable. Therefore, the idealized BST surface can be 

used to estimate the seismic damage sequence of asymmetric structures roughly. 

3) On the other hand, the trends of the time history of the base shear-torque lean to the longest sides of 

the BST surface illustrate effects of the change of the stiffness eccentricity to the seismic behavior of the 

model structure. That is, at the onset of the damage of the left column and central wall plane, the stronger 

right wall plane still remains linear elastic. Inevitably, the stiffness center shifts to the stronger right wall 

plane, intensified the stiffness eccentricity of the model structure. This phenomenon explains that 

although the stiffness eccentricity does not affect the shape of the BST surface, it really influences the 

developing trends of the time history response of the base shear-torque. In other words, the larger of the 

stiffness eccentricity is, the smaller of the possibility that the stronger members near the stiffness center 

are damaged will be and, the larger of the possibility that the weaker members far away from the 

stiffness center are damaged will be. The consequence is the larger of the possibility of the serious 

unbalanced damage occurring will be. 

4) In most experimental cases, there is no pure torsional damage observed from the responses of the base 

shear-torque, i.e., only in few cases of modell and mode!3, the torque resistance are exceeded lightly. In 

model2, such phenomena are never found. This fact illustrates, on one hand, the possibility of the 

occurring of the pure twist damage will be very small even for structures with larger eccentricity such as 

the model structures here studied. On the other hand, it shows that the location change of the central wall 

wl improves the entire torque resistance directly and makes the coupling response more uniform and 

simultaneously, it tells the fact of the effects of the structural members in the orthogonal direction is 

un-negligible when the seismic resistant capacity in one principal direction of the structure is studied and 

vice versa.

269 

State space of base shear-torque relationship 

In fact, the degradation type of bi- and tri-linear force-displacement relationship should be more suitable 

than the idealized elasto-plastic one for most cases of reinforced concrete building structures in seismic 

ductility design. This conclusion can be verified directly from the comparison of the abovementioned 

experimental response and the idealized BST surface shown in fig.3a)~c). Then, correspondingly, the 

single space surrounded by the BST surface can be developed to 2 or 3 state spaces signifying 

respectively the elastic, inelastic even collapsed state of the related structural system. In other words, for 

evaluating the seismic performance of asymmetric RC structures realistically, the ultimate BST surface 

in the idealized model should be developed to nonlinear space(s) of the base shear- torque responses. The 

first space signifies the elastic response while the second space represents the inelastic characteristics 

when yield damage occurred in structure, and the third space means that the shear-torque response 

exceeds the ultimate capacity of the system. The boundary between the first and the second space 

describes a kind of critical state that the yielding damage initiates in system. Whilst, the boundary 

between the second and the third space should theoretically signify that the system begins to collapse. 

As example, abovementioned specimen2 (bi-directional asymmetry with 3 lateral resistant planes along 

the y-direction of ground excitations) here is used to show how the 3 state-spaces model for evaluation of 

the seismic response of the base shear-torque is established. From the experimental data for the yielding 

strength f y of the reinforcement used in the model system shown in table 1 respectively, the yield 

boundary dividing the elastic and the inelastic response spaces can be determined according to the 

method that De La Llera and A.K. Chopra established, i.e., its* location is completely same to that of the 

BST surface in the idealized model but they have quiet different meaning. Using the similar way and the 

ultimate strength/, shown in table 1 , the eight vertices of the collapse boundary can be easily determined 

based on the following formulas* 31 : 

5 = — | 7 6 ~ — 2 i -j — —-} i % — — ^ 

y^ = -y\ ' y 6 = -^ > ^ = -^ > ^ = -^ 

( 1 ) 

where, in a more general sense, the symbol "*" signifies two kinds of critical state: the onset of yielding 

and reaching at the ultimate strength. If the structure system just starts to be damaged (reaches its yield 

strength), parameters in the above formulas have the same meanings with those of in the idealized BST 

surface. If the structural response reaches the maximum capacity of the system, these parameters can be 

determined based on the ultimate strength of structural members. As shown in fig.4, structural 

parameters controlling the response state space of the studied system are explained as, 

1). y* =VJV* Q is the normalized base shear along x-direction, while F r 

v 

0 = ^^ / vn and 

^/o = X- /«' are ê yi e^m§ an d ultimate lateral resistant capacities along the x-direction of the 

studied system respectively, in which, f yxl and f llxl are the corresponding yield and ultimate strength of 

the zth lateral resistant plane, and Mis the number of the lateral resistant planes along x-direction. If the 

earthquake actions along the x-direction (besides slong the y-direction) should be taken into account, i.e., 

j7 r =£0, the shape of the state space would be changed with the change of the value of V x . Otherwise, if 

only single-directional earthquake excitation is studied, i.e., F r =0, the effects of this item should be 

ignored. 

2). J/, V 0 = Y^ f m and V" Q = ^N f uyl are the lateral yield and ultimate capacity of the system along 

y-direction, f yyi and f liy , are the lateral yield and ultimate strength of the lateral resistant planes, 

respectively. N is the number of the lateral resistant planes along y-direction. The value of V^ and V^

270 

determine the span of the state space of base shear-toque on the shear axial V" Q - F/ 0 directly gives out 

the improvable space of the shear capacity of the structural system after yielding. 

3). V\ and V\ are the lateral yield and ultimate capacity of the resistant plane passing through the 

center of mass. Generally, this parameter represents the sum of the resistant capacity of all lateral planes 

closing to the center of mass. From fig.6, it is obvious that the increase of the value V u yc - V v yc means the 

increase of the possibility that the pure torsional mechanism will occur after yielding. 

and jZ ml \f y, respectively, define the 

torsional capacity of the system when pure twist yield and ultimate damage will be occurring. These two 

parameters control the maximum and minimum coordinates of the two critical boundaries along the 

torque axial. Moreover, the improvement of the torsional capacity of the system after yielding is also 

controlled by the value T 0 " - T Q V . 

5)- TL = 5^ fy**yi ^ -Ti "XTil^^'i are contr & u ti° n of the torsional capacity provided by the 

orthogonal lateral planes at the level of initiating yield and collapse respectively. The possibility of the 

pure shear mechanism should be increased with the increase of the value 7^ - 7 1 / particularly during the 

stage of inelastic response 

Base Torque 

Collapse Boundary 

-V: 

Base Shear 

Yield Boundary 

Elastic Space 

6 y 7 y 

Fig.4 Structural parameters controlling the state space of base shear-torque 

, IV lv 

y* y* determine the strength eccentricity of the system at 

two damage levels: before yielding and at the initiating of collapse, respectively. They signify 

independently the degree of the strength eccentricity on the two critical levels. Here, x v is considered to 

be a constant during the elastic stage while x u p is also defined as a critical value at the instant of initiating 

collapse although the strength eccentricity is varied from x y p to its' maximum value, and then decreased 

to x" p with the increase of the intensity of the ground excitations gradually during the inelastic stage.

271 

7)- C = Z;;i 2 (/ w *i/*,|) and V u yu ^^(f^x, /|x,|) , at two levels: before yielding and at the 

initiating of collapse, respectively, define the degree of the unbalanced strength distribution at the two 

sides of the center of mass. The variation of the value V^ ~ V^ shows the change of the possibility of the 

occurring of the pure shear mechanism. Corresponding with the change of the strength eccentricity, this 

value should represent the state of some weaker structural members being damaged on one side of the 

center of mass while remaining elastic on the other side of the CM. Fig.4 does not show the peak value 

directly 

From the determination of the two critical boundary of the torsional-lateral coupling system, it can be 

inferred that the seismic response of the base shear-torque is naturally divided into 3 regions that 

represent respectively the linear elastic state, nonlinear state and the destroyed collapse state. The two 

critical boundaries, which signify the initiating yield and collapse, can be established independently 

based on the yield strength and the ultimate strength of structural members. This model can be directly 

used to estimate both of the overall seismic capacity and the developing trends of the time history 

response of the system base shear-torque to strong earthquakes even without complicated nonlinear 

analysis. 

«Collapse boundary of the state space 

Yielding Boundary 

* Dotted area: Time-history responses of the base shear-torque in earthquake simulation 

test for specimen 2 

Fig.5 comparison of the theoretically established state space of the base shear-torque with experimental results 

Fig.5 shows the comparison of the theoretical state space of the base shear-torque with the experimental 

results. The time-history shear-torque response of the bi-directional asymmetric single-story RC system 

in 19 cases of ground excitations illustrates highly consistency with the state space established based on

the assumption of the bi-linear force-displacement relationship. Such consistency is unreachabie only by 

using the BST surface based on the idealized elasto-plastic assumption. 

272 


In summary, the shake table tests provided useful data references for theoretical studies on seismic 

torsional effects of asymmetric building structures. Through the comparison between the experimental 

and the theoretical results, the following conclusions can be obtained: 

1) Although the BST (base shear-torque) surface established on the assumption of idealized 

force-displacement relationship can be roughly used to estimate the seismic performance of 

asymmetric systems, it will bring obvious error on the prediction of the nonlinear response when the 

studied structure suffered strong earthquakes. 

2) Therefore, based on the assumption of the bi-linear force-displacement relationship and the 

displacement-based method to determine the stiffness, the state space signifying the base 

shear-torque relationship is developed to describe the entire seismic capacity of the asymmetric 

systems. It shows high consistency with the experimental results. Therefore, it is no doubt a reliable 

model that can be directly used to estimate not only the entire seismic capacity but also the potential 

seismic damage may occur in the studied systems, even without relying on any complicated 

numerical analysis. It can be seen, this model will be very convenient and useful in evaluation of the 

lateral-torsional coupling effects of asymmetric structures particularly in the preliminary stage of 

design, retrofitting and research. 

It should be noted that abovementioned conclusions are obtained on the basis of the experimental results 

of single story building structure. Their reliability and universality should be checked further with real 

seismic records or earthquake simulation tests for multi-story and high-rise building. 

References 

1. Dai Junwu, Yuklung Wong and Minzheng Zhang, Torsion Response of Reinforced Concrete Building under Shaking 

Table Test. Advances In Structural Dynamics, 2000, Vol. II, pp859-865. 

2. Dai Junwu, Research on nonlinear seismic performance of asymmetric RC building structure^ Ph.D. thesis, Institute 

of Engineering Mechanics, China Seisraological Bureau, 2002, 

3. De la Llera, J.C., Chopra, A. K., Understanding the Inelastic Seismic Behaviour of Asymmetric Plan Buildings, J. 

Earthquake Engineering & Structural Dynamics, 1995, v.24, pp.549-572. 

4. De la Llera, J.C., Chopra, A. K., Inelastic Behaviour of Asymmetric Multistory Buildings, L Structural Engineering, 

1996,v.l22,no.6,pp.597-606. 

5. Guo Xun, Yuk-lung Wong and Yuan Yifan, Investigation of Seismic Response of Soil site in HK, HKIE Transactions, 

2000, Vol. 7, No. 3. 

6. Kan, C.L. and Chopra, A.K., Torsional coupling and earthquake response of simple elastic and inelastic systems, J. 

structural Division, ASCE, 1981, v.107, pp.1569-1588. 

7. Minzheng Zhang, Problem study on the application of the similitude law in earthquake simulation test, Earthquake 

Engineering and Engineering Vibration, 1997.6, Vol. 17, No.2. 

8. Palazzo, B. and Fraternali, F., Seismic ductility demand in buildings irregular in plan: A new single story nonlinear 

model, Proceedings of the 9 th World Conference on Earthquake Engineering, Tokyo-Kyoto, Japan, 1988, Vol.V, 

V-43 to V-48. 

9. Paulay, T., Displacement-based design approach to earthquake-induced torsion in ductile buildings, Engineering 

Structures, 1997, VoL19,No,9, pp.699-707. 

10. Paulay, T., Torsional mechanisms in ductile building systems, Earthquake Engineering and Structural Dynamics, 

1998, Vol.27, pp.1101-1121. 

11. Paulay, T., Understanding torsional phenomena in ductile systems, Bulletin of the NZ Society for Earthquake 

Engineering, 2000 S Vol.33, No.4, pp.403-420.




SEISMIC RESISTANCE OF VERY HIGH STRENGTH 

HIGH RISE RC BUILDINGS FOR URBAN DEVELOPMENT 

A.S. Elnashai 1 and B. Laogan 2 

1 Mid-America Earthquake Center, Department of Civil and Environmental Engineering, 

University of Illinois, Urbana, Illinois, USA 

2 ATR JEngineering, Manila, Philippines 

ABSTRACT 

High rise buildings have increased in number and geographical spread. This in turn dictates that higher 

strength materials are utilized, to keep member sizes at manageable proportions. Whereas the increase 

in strength of high strength steel and concrete (steel above 500 MPa yield strength and concrete above 

70 MPa compressive strength) results in commensurate increase in member capacity, this does not 

necessarily follow when considering the deformational capacity of members hence structures. 

Moreover, the increase in cost associated with high strength materials is not necessarily offset by the 

reduction in quantities and construction time. Issues of ductility of high strength materials have been 

explored by the writer and his co-workers using testing and analysis on the material, member and 

structure levels for a number of years. In this paper, after a brief mention of previous studies, ten high 

rise frame-wall RC buildings are designed with various combinations of steel and concrete strengths 

and equivalence criteria (defined as criteria that render two different designs equivalent in terms of 

their fulfilment of a design premise). These structures are then analyzed up to collapse when subjected 

to a number of earthquake ground motions. The results for the different strengths and also the different 

equivalence criteria are compared from a structural performance as well as from the cost of 

construction points of view. It is concluded that there is a limit to the use of high strength materials 

dictated by the ductility capacity, energy absorption potential and economics of construction. These 

limits are quantitatively established for the class of structure investigated. 

INTRODUCTION 

The need for higher buildings (due to financial centres and population expansion) naturally leads to the 

conclusion that high strength construction materials will be increasingly used in the future, in order 

that column sizes are maintained at current dimensions and more effective use is made of floor areas, 

especially in the lower stories of high rise structures. Two other performance criteria lend weight to the 

necessity of the use of high strength concrete: increased wind and traffic vibration susceptibility 

dictates that the modulus of elasticity of the material is as high as possible, in order to delimit small 

amplitude elastic displacements. Moreover, the need for rapid construction requires early age strength 

gain, a feature that may be offered readily by high strength concrete.

274 

There are several outstanding examples of high rise buildings using high strength materials around the 

world, both existing and under construction or consideration. In Seattle, Washington, the Pacific First 

Centre (44 storeys) and Two Union Square (62 storeys) buildings employ concrete with compressive 

strength of 115 MPa; mainly used for its high modulus of elasticity of 50,000 MPa. These are extreme 

examples. More commonly used high strength concrete mixes are represented by the Two Prudential 

Plaza (281 m) and 311 South Wacker Drive (295 m) buildings (both in Chicago, Illinois), which use a 

mix with compressive strength of over 80 MPa. Outside the US, the Miglin-Beiter building (Frankfurt, 

Germany) has a 610 m height and concrete of 97 MPa compressive strength. In addition, the Petronas 

Towers (Kuala Lampur, Malaysia) (450 m) employ concrete of compressive strength of 80 MPa for 

the lower stories; while the BfG building (186 m), in Frankfurt, Germany, goes up to 85 MPa. 

Development of high yield steel for general construction lagged behind that of concrete. Only in the 

early 1990s was there widely available high yield steel, mainly from Japanese steel manufacturers, 

with yield strength above 1000 MPa being tested and used. This has had an effect on high strength 

materials construction economics because the decrease in column dimensions would lead to an 

increase in required steel area, unless high yield steel is used. 

REVIEW OF PREVIOUS STUDIES 

Whereas several research projects have been concerned with the experimental behaviour of reinforced 

concrete members with high strength concrete and high yield steel, very few have applied loading and 

boundary conditions relevant to earthquake response. Furthermore, none of the previous projects 

comprehensively and systematically addressed the range of concrete compressive strength, 

longitudinal steel yield stress, transverse steel yield stress and confining steel spacing. Though, it is 

noted that such early studies were most valuable in establishing trends and alerting designers to serious 

issues which may affect seismic safety. In Table 1, the ranges covered by tests conducted by various 

researchers are presented: the light shading indicates testing under axial force only (including eccentric 

loading); the darker shade indicates combined axial-flexural loads; and blank areas were not 

previously tested. 

TABLE 1 

RANGE OF TEST PARAMETERS FOR STEEL AND CONCRETE 

(Letters allude to references, below Table 2) 

f côr f y -> (MPa) 

60-80 

80-100 _, 

100-120 

>120 

300-500 

a,c,d,f,l,m 

cJOr 

b,c,d,jq,r 

b,r 

500-700 

d 

k,p 

b,d,p 

b 

700-900 

' d 

k 

d 

900-1100 

m 

>1100 

With regard to confining steel characteristics and spacing tested, as identified from the published 

literature, these are indicated in Table 2.

275 

f c i or f v -* (MPa) 

60-80 

80-100 

100-120 

>120 

TABLE 2 

RANGE OF TEST PARAMETERS FOR CONFINING STEEL 

(Letters allude to references, given below) 

300-500 

c,e,f,l 

c,e,l,q 

b,c,d,e,g,j,q 

b,e,g,h 

500-700 

f 

o,p 

o,p 

h 

700-900 

d,f 

k,q 

dj,q,r 

r 

a. Ahmad and Shah (1982) b.AI-Hussainie/a/(1993) 

d. Nagashima etal (1992) 

e. Muguruma et al (1993) 

g. Razvi and Saatciogiu (1994) h. Razvi and Saatciogiu (1996) 

j Azizinammi and Kuska( 1993) LKimura etal (1996) 

m.Tanakae/a/(1994) 

n. Sugano(1996) 

p. Galeota and Giammatteo (1996) q. Muguruma and Watanabe (1990) 

900-1100 

a 

h 

h 

>1100 

a,d,l,m 

1 

d,n 

e t n 

c.Bjerkeli etal (1990) 

f. Cusson and Paultre (1993) 

i. Sun etal (1996) 

L Liet al (1994) 

o. Bayrak and Sheikh (1996) 

r. Nishiyama et al (1993) 

It is clear from the above, that there are considerable gaps in testing results of relevance to seismic 

response; which require attention prior to attempting to derive comprehensive design guidance. This is 

further emphasized by the observation that some of the above tests were conducted under eccentric 

axial load to represent combined axial-flexural testing. Notwithstanding other objectives that the 

researchers may have had in mind, this type of testing is considered by the authors to be not strictly 

relevant for seismic assessment purposes. There are several behavioral considerations supporting the 

latter statement; amongst which is that levels of axial load at maximum moment are so high that the 

mobilized confining stresses are rather unrealistic and unrepresentative of seismic response of 

structures. A comprehensive testing program on beam-column members was recently completed at 

imperial College (Elnashai et al [1998], Goodfellow [1998] and Elnashai and Goodfellow [2002]). The 

testing range was concrete strength of 60 MPa to 130 MPa and steel yield strength of 500 MPa to 1300 

MPa, with two stirrup spacing and two axial load levels under cyclic and monotonic transverse 

loading, leading to 92 test specimens. 

It was concluded (in the experimental studies above) that values of plastic hinge length and 

displacement ductility of reinforced concrete members, constructed from concrete up to 130 MPa and 

steel up to 1300 MPa, are on the whole similar and lower, respectively, than the normal material values 

commonly expected. However, no indication was given that ductility-based seismic design cannot 

make use of these materials that may be dictated on the project, by column size requirements and wind 

vibration control It is clear though, that specific design guidance is needed and curbs on the level of 

steel yield in particular should be imposed. This is confirmed by the low values of deformation-related 

parameters obtained for the f c 130-f y !300 pairs. Until further work is undertaken, a limit of concrete 

strength and steel yield for ductile seismic design may be set at f c =100 MPa, with f y =700 MPa. These 

values started from the study reported above and therefore, are employed below. The study also gave 

values of secant stiffness at yield and ultimate for displacement-based design purposes. Finally, it is 

significantly concluded that the use of high strength confining hoops is not warranted, especially when 

employing high strength concrete. This is attributed to the low levels of dilation generally observed for 

high strength concrete. This is an important economic consideration in ductility-based design using 

high strength materials. 

There are very few studies on high strength materials at the structure level. The study by Kateinas 

(1997) (summarized in Elnashai, 1998) was one of the first detailed analytical investigations into high 

strength concrete buildings. It used the simplest approach possible for the analysis of a suite of

276 

buildings with different concrete strength and steel yield. This comprised keeping the member size 

constant and changing the material properties. Moreover, the higher the strength of concrete, the 

steeper was the third segment of the model. The above features of that study significantly affected the 

results obtained, hence, the conclusions drawn. By keeping the section dimensions and ratio of 

reinforcement constant, many of the structures ended up with response typical of 'strength design' as 

opposed to 'ductility design', as well as sections being over-reinforced. The consequence is to decrease 

the ductility of structures using high yield steel. The feature of using severely dipping descending 

branches for concrete lead to the structures behaving in a highly non-ductile fashion. It was noted in 

that study that high strength structures exhibit very low levels of ductility and behaviour factors 

nearing unity. In common with the current study though, the range of concrete and steel for viable 

seismic design was identified. Therefore, the above analytical study represents the expected behaviour 

of high strength concrete structures if the confined concrete behaviour is still non-ductile and the steel 

stress ratio (ultimate-to-yield) is near unity. Higher levels of ductility would have been obtained 

though, if the 'over-reinforced' sections were re-designed. 

SCOPE OF WORK ON ANALYTICAL ASSESSMENT 

The main objective of this paper is to summarize investigations of the seismic performance of high 

nse, high strength, concrete buildings designed according to existing code provisions and undertaken 

by the primary author and his co-researchers (Kateinas [1997], Laogan and Elnashai [1999], 

Goodfellow [1998] and Elnashai and Goodfellow [2002]) and to employ the conclusions of the 

experimental study in structural-level investigations. 

Ten, 24-storey structures with the same overall dimension 

and geometry, sized according to two equivalence criteria 

(Laogan and Elnashai, 1999), were analyzed elastically and 

designed according to the provisions of the Uniform Building 

Code (1994). The general layout of the frame-wall structures 

is shown in Figure 1. The material strength characteristics of 

the structures are given in Table 3. Every effort was made to 

have realistically designed and detailed structures. A 

comparative cost analysis of these structures was performed 

to determine their relative cost-effectiveness. Inelastic 

analyses were then performed using the program ADAPTIC 

(Izzuddin and Elnashai, 1989). Two sets of inelastic analyses 

were undertaken: first a static pushover analysis of the 

structure was performed, followed by a set of dynamic 

analyses using three artificial accelerograms, scaled to the 

design ground acceleration (PGA) and twice the design PGA. 

Moreover, dynamic analyses up to collapse, for purposes of FIGURE 1 

(R ° r q) ' ^ 24-STORY FRAME-WALL 

MODEL STRUCTURE

277 

TABLE 3 

SUMMARY OF STRUCTURES CONSIDERED 

Model 

N35 

N80 

N100 

N120 

N80R 

N100R 

N120R 

N80RS 

N100RS 

N120RS 

f c (MPa) 

35 

80 

100 

120 

80 

100 

120 

80 

100 

120 

Beams 

415 

415 

415 

415 

600 

700 

800 

600 

700 

800 

Columns 

415 

415 

415 

415 

415 

415 

415 

600 

700 

800 

f y (MPa) 

Shear Wall 

415 

415 

415 

415 

415 

415 

415 

600 

700 

800 

Hoops 

415 

415 

415 

415 

415 

415 

415 

600 

700 

800 

Remarks 

Basic Model Structure 

Equivalent Stiffness 



Reduced Stiffness 



One steel grade only 



In Table 3, ""equivalent stiffness" refers to an equivalence criterion whereby the stiffness of the 

structure is kept constant; taking into account that higher strength concrete will have higher stiffness. 

"Reduced stiffness" refers to a criterion whereby a limiting elastic drift value governs the equivalence 

of two structures (this is selected as 0.02/R W as recommended in Uniform Building Code, 1994). "One 

steel grade" refers to structures where all slabs, beams and columns utilize the same steel grade. The 

schedule of reinforcement and detailed drawings of the main lateral resisting elements, of the ten 

model structures, is given in Laogan and Elnashai (1999). 

RESULTS OF COST COMPARISON 

The assessment included not only the material costs, but also the cost of formwork. Considerable effort 

was expended in collecting realistic cost estimates from more than one source, but the significance of 

the study is more comparative than absolute. It was concluded that the most cost-effective structure is 

N80R. This is attributable to it using only a minimal amount of the expensive high yield steel (used in 

the beams only), while at the same time taking advantage of the large axial load carrying capacity of 

the high strength concrete to reduce the vertical steel reinforcement. Figure 2 shows a bar chart of the 

cost broken down into the different components.

278 

160,000 - 

140,000 

_^ 

d Concrete 

• Form works 

120,000 - -j DRebars 

! 

1 

1 

" 

" 

100,000 - 

s : 

60,000 - 

40,000 

— 

— 

- 

1 

- 

20,000 

n 

FIGURE 2 

COST OF THE DIFFERENT COMPONENTS OF THE STRUCTURES 

As a group, the structures designed using the equivalent stiffness criterion (NSO-N120) show an 

increase in the cost of the concrete component as the strength of the material increases. The cost of the 

formwork remains relatively stable. The cost of the steel reinforcement is almost similar among the 

high strength concrete structures and is substantially less than that of the normal strength concrete 

structure. The group designed using the reduced stiffness criterion seems to be the most economical. 

The cost of the concrete component increases with increasing concrete strength, but is generally less 

than the normal strength concrete structure. The cost of formwork (as in the previous group) remains 

relatively the same, while the cost of the steel component shows a slight increase with increasing 

concrete strength. The structures using only one grade of steel appear to be the least economically 

viable. The relative savings over T expenditures are given in Table 4. 

Model 

% Savings 

N35 

0.00% 

TABLE 4 

PERCENTAGE SAVINGS RELATIVE TO N35 

N80 

8 08% 

N100 

5.35% 

N120 

-0.99% 

N80R 

17.70% 

N100R 

12.99% 

N120R 

6 57% 

N80RS N100RS 

11.85% 6.22% 

N120RS 

-1.14% 

At this cost level, it is noticeable that even the structures using only one grade of steel become 

economically feasible. Moreover, the use of high strength concrete can result in substantial savings. 

This suggests that the cost effectiveness of a high strength structure depends to a large extent on the 

cost component attributable to the steel reinforcement. As this item drops due to wider international 

use, high strength RC buildings will become more economical than has hitherto been the case.

279 

STATIC AND DYNAMIC ANALYSIS RESULTS 

The yield point of the structure, defined based on an equivalent elasto-plastic system with reduced 

stiffness, evaluated as a secant through 75% of the maximum, is adapted for this study On the other 

hand, the ultimate limit state (ULS) is based on a global criterion of the attainment of 3% interstorey 

drift. At this level of drift, it is generally accepted that structures would have suffered major structural 

and non-structural damage. Further refinement of the limit states is not necessary since the primary 

objective is to compare the different structures. Figure 3 shows the force-displacement curve for the 

normal strength and the equivalent stiffness structures. It is noticeable that the initial stiffness of the 

structures is the same This feature is also exhibited in the reduced stiffness structure, thus validating 

the equivalence criteria used. This also confirms that imposing equivalent stiffness at the member level 

will give a structure with the same overall stiffness. 

10000 

8000 

* 6000 - 

W 

u 

K 

4000 

/ 

/ 

/• 

^ -~ ~ 

>- — 1 

500 1000 1500 

TOP DISPLACEMENT(mm) 

N35 1- 

---- NSO I 

Kttnn , 

2000 

FIGURE 3 

FORCE-DISPLACEMENT CURVE FOR N35, N80, N100 AND N120 

The level of overstrength is defined herein as the ratio of the capacity of the structure, based on the 

static pushover analysis, to the code-defined design base shear. The calculated overstrength for the 

different structures is given in Table 5. 

TABLES 

STRUCTURE OVERSTRENGTH 

Model 

Overstrength 

N35 

6.29 

N80 

4.51 

N100 

4.47 

N120 

4.42 

N80R 

4.20 

NIOOR 

4.67 

N120R 

4.93 

N80RS 

4.25 

N100RS 

4.77 

N120RS 

5 10 

For each of the ten structures, dynamic analysis was performed with three artificial records, scaled to 

the design ground acceleration (PGA) and twice the design PGA, corresponding to 0.4g and 0.8g, 

respectively. As a measure of global response characteristics, the displacement and total base shear 

time-history and the interstorey drift ratio from the analyses are examined. For the purposes of 

discussion, NSO, N100 and N120 constitute "E" group (based on the equivalent stiffness criterion); 

N80R, NIOOR and N120R constitute "R" group (based on the reduced stiffness criterion); and N80RS, 

N100RS and N120RS constitute"RS" group (based on the reduced stiffness criterion with one steel 

grade only). 

It was observed that for buildings in the same group, subjected to the same input motion, the shape of 

the displacement and base shear response was similar. This suggests that structures in the same group

280 

responded in a fairly similar manner, confirming the effectiveness of the criteria developed for 

generating the structures. The maximum top displacement of the structures, within the same group, 

shows a slight increase as the concrete strength increases. This is consistent with the load-displacement 

curves generated under static loading. The higher strength concrete experiences slightly more 

softening beyond a certain strain, in the initial ascending branch of the curve. This is partially 

attributable to the cracking of the concrete cover. In all instances, the maximum top displacement of 

the structures in the R and RS groups are greater than those structures of the same concrete strengths in 

the E group. 

N80R N100R N120R 

FIGURE 4 

SAMPLE PLASTIC HINGE LOCATIONS 

(SHADED AREAS ARE PLASTIC HINGING IN THE WALL) 

The plastic hinges formed in the different structures show that the number of column hinges generally 

increases with increasing concrete strength. This is attributable to the higher strain levels in the steel, 

due to the increase in strength and modulus of elasticity of the concrete. Although the strong-column 

weak-beam provision of the code was satisfied in the design, the 120 MPa concrete structures in the R 

group show substantially more hinging in the columns, as shown in Figure 4. The use of high yield 

steel in the beams and normal grade steel in the columns of the R structures has an adverse effect in 

terms of increased number of column hinges. The RS group, which used high yield steel in the 

columns, had fewer column hinges, even at high levels of loading. On the other hand, the spread of 

inelasticity in the shear wall is unaffected by the change in concrete strength. The use of normal grade 

steel with high strength concrete does not cause undesirable consequences in the wall. For the beams, 

at twice the design PGA, yielding has occurred in almost all cases. This is both expected and desirable 

at this level of loading. 

To study local effects, the curvature ductility demand of a number of beams at different locations in 

the structure, was calculated. The maximum curvature ductility demand in the beams is shown in 

Table 6.

281 

TABLE 6 

MAXIMUM CURVATURE DUCTILITY DEMAND IN BEAMS 

Model 

EC8-1 

EC8-2 

EC8-3 

0.4g 

3 11 

373 

2.33 

E- group 

0.8g 

7 15 

703 

6.09 

0.4g 

1 68 

1.63 

165 

R - group 

0.8g 

399 

383 

3.05 

RS - group 

0.4g 

1 71 

1 68 

167 

0.8g 

3.87 

3.90 

3.12 

It is interesting to note that the maximum curvature ductility demand, given in Table 7, occurred in the 

normal strength concrete structure and the equivalent stiffness HSC structures. The ductility demand at 

a PGA of O.Sg is already on the high side, but is within achievable ductility capacity of well-detailed 

members. For the R and RS structures, the ductility demand is significantly less compared to those of 

E group. Since the R and RS group structures use high yield steel in the beams, the strain that defined 

yield is much higher compared to normal strength steel; hence, a higher yield curvature and 

consequently, a lower ductility demand was observed. 

TABLE 

CALCULATED COLLAPSE PGA, YIELD PGA AND RESPONSE MODIFICATION 

Model 

N35 

N80 

N100 

N120 

N80R 

N1GOR 

N120R 

N80RS 

NIOORS 

N120RS 

EC8-1 

0.841 

0.870 

0.868 

0.893 

0.934 

0.958 

0.984 

0.963 

0.959 

0.997 

Mg) 

EC8-2 

0.950 

0.982 

0.954 

0.970 

0.864 

0.911 

0.939 

0.901 

0.918 

0.958 

EC8-3 

0.902 

0.871 

0.847 

0.869 

0.980 

0.898 

0.876 

0.980 

0.869 

0.853 

EC8-1 

0.150 

0.107 

0.106 

0.105 

0.100 

0.111 

0.117 

0.101 

0.114 

0.121 

Me) 

EC8-2 

0.150 

0.107 

0.106 

0.105 

0.100 

0.111 

0.117 

0.101 

0.114 

0.121 

EC8-3 

0.150 

0.107 

0.106 

0.105 

0.100 

0.111 

0.117 

0.101 

0.114 

0.121 

EC8-1 

5.61 

8.13 

8.19 

8.50 

9.34 

8.63 

8.41 

9.53 

8.41 

8.24 

Rorq 

EC8-2 

6.33 

918 

900 

9.24 

8.64 

8.21 

8.03 

8.92 

8,05 

7.92 

EC8-3 

6.01 

8.14 

7.99 

8.28 

9.80 

8.09 

7.49 

9.70 

7.62 

7.05 

For simplicity, the response modification factor (R in US practice and q in European practice) is 

defined as the ratio of the ground acceleration of the earthquake that caused the attainment of the 

ultimate limit state, to that of the earthquake corresponding to the yield limit state. Of all the structures 

considered, the normal strength structure has the lowest response modification factor. This is 

attributable to it having a much larger level of overstrength, compared to the other structures. The extra 

steel reinforcement, provided for gravity loading in the normal strength concrete, has affected its 

energy dissipation capacity in the inelastic range. The behaviour factor for structures in the E group is 

almost constant. On the other hand, the behaviour factor generally decreases with increasing concrete 

strength for both the R and RS structures (with very few exceptions). Comparison of the calculated 

behaviour factors suggest that, in terms of structural ability to respond in the inelastic range, the 80 

MPa concrete is the most effective for the given structural configuration.

282 

CONCLUSIONS 

In this paper, the seismic performance and cost effectiveness of high rise high strength RC buildings 

were discussed. Based on the analyses performed, the following conclusions are supported by the 

results: 

• Contrary to some of the studies performed on columns subjected to pure axial load, the optimum 

concrete strength that will give the most economical structure is not necessarily the highest 

concrete strength available. The concrete strength that can reduce the cost of the steel component, 

while at the same time limit the cost of the concrete component, will result in the most cost 

effective structure. 

• In the light of the common practice of using high strength concrete only in the columns of the 

structure, it was shown that the use of high strength concrete in the beams and slabs, where they 

are less effective, can still furnish substantial cost savings. 

• Under static loading, the high strength concrete structures have stable load-displacement curves. 

The shape of the curve is similar to that of normal strength concrete structures. For this type of 

loading, inelasticity developed mostly in the beams. The columns appear to be well protected from 

hinging by capacity design regulations. 

• The level of overstrength in high strength concrete structures, calculated based on the static 

pushover analysis, is less than that of the normal strength structure. In the normal strength 

structure, the additional steel reinforcement required to resist high axial loads provide extra lateral 

load capacity, thus increasing overstrength. 

• At the global level, based on the three response parameters (top displacement, base shear and 

maximum interstorey drift ratio) examined, the performance of the high strength concrete 

structures compares favourably with that of the equivalent normal strength concrete structure. 

• The selection of the grade of steel to use with the high strength concrete is very important. Results 

from the dynamic analysis indicate that using normal grade steel for concrete strengths of up to 80 

MPa is adequate. Beyond 80 MPa, the use of normal grade steel with high strength concrete 

resulted in significantly more hinging in the columns. The use of high yield steel in beams and 

normal grade steel in the columns should also be avoided. 

• The maximum curvature ductility demand at the design and twice the design earthquake is 3.73 

and 7.15, respectively. These are well within the achievable ductility capacity of seismically 

designed and detailed members. Moreover, the use of high yield steel with high strength concrete 

significantly reduced the ductility demand on the members to about 1.63 and 3.83. 

• The calculated behaviour factors suggest that the SOMPa concrete is the optimum concrete strength 

for the given structural configuration, in terms of energy dissipation capacity. In general, the higher 

strength concrete structures have a larger behaviour factor compared to the normal strength 

structure.

283 

REFERENCES 

Ahmad, S.H. and Shah, S.P. (1982). Stress-strain curves for concrete confined by spiral reinforcement. 

ACIJournal 79:46, 484-490. 

Al-Hussami, A, Regan P.E., Xue, H-Y. and Ramdane, K-E. (1993). The behaviour of high strength 

concrete columns under axial load. Third International Conference on Utilisation of High Strength 

Concrete, Lillehammer, Norway, pp. 83-90. 

Azizinamini, A. and Kuska, S. (1993). Flexural capacity of high strength concrete columns. Third 

International Conference on Utilisation of High Strength Concrete, Lillehammer, Norway, pp. 91-97. 

Bayrak, 0. and Sheikh, S.A. (1996). Confinement requirements for high strength concrete columns. 

llth World Conference on Earthquake Engineering, Acapulco, Mexico, Paper no. 463. 

Bjerkeli, L, Tomaszewicz, A. and Jensen, J.J. (1990). Deformation properties and ductility of highstrength 

concrete. High Strength Concrete - Second International Symposium, ACISP 121, pp. 215- 

238. 

Cusson, D. and Paultre, P. (1993). Experimental study of high strength concrete columns confined by 

rectangular ties. Third International Conference on Utilisation of High Strength Concrete, 

Lillehammer, Norway, pp. 136-145. 

Elnashai, A.S. (1998). Earthquake resistance of high strength reinforced concrete buildings. Seismic 

Design Practice into the Next Century., Booth (ed.), Balkema, Rotterdam, pp. 3-14. 

Elnashai, A.S. and Goodfellow, R.C. (2002). Seismic Response characteristics of high strength steel 

and concrete members. 7 th National U.S. Conference, Boston, Massachusetts, USA. 

Elnashai, A.S., Goodfellow, R.C. and Chana, P.S. (1998). Flexural ductility of high strength reinforced 

concrete structures. Proceedings of the Second Japan-UK Workshop on Implications of Recent 

Earthquake on Seismic Risk, Tokyo Institute of Technology, Japan, pp. 145-157. 

Galeota, D. and Giammatteo, M.M. (1996). Seismic resistance of high strength concrete columns, llth 

World Conference on Earthquake Engineering, Acapulco, Mexico, Paper no. 1390. 

Goodfellow, R.C. (1998). Ductility of RC members with high strength materials. PhD thesis, 

University of London, UK. 

ICBO, Uniform Building Code (1994). International Conference of Building Officials, 

California, USA. 

Whittier,' 

Izzuddin, B.A. and Elnashai, A.S. (1989). ADAPTIC - A program for adaptive large displacement 

eiasto-plastic dynamic analysis of steel and composite frames. Engineering Seismology and 

Earthquake Engineering Report No. 89/7, Imperial College, London, UK. 

Kateinas, I. (1997). Seismic response of R/C structures made with high strength material. MSc 

Dissertation, Imperial College, London, UK. 

Kimura H., Sugano, S. and Nagashima, T. (1996) Seismic behaviour of reinforced concrete columns 

using ultra-high-strength concrete under high axial load. Fourth International Symposium on 

Utilisation of High Strength/High Performance Concrete, Paris, France, Paper no. 184. 

Laogan, B.T. and Elnashai, A.S. (1999). Structural performance and economics of tall high strength 

RC buildings in seismic regions. The Structural Design of Tall Buildings, 8, 171-204.

284 

Li B, Park, R. and Tanaka, H. (1994). Strength and ductility of reinforced concrete members and 

frames constructed from high strength concrete. Research Report 94-5, Department of Civil 

Engineering, University of Canterbury, Christchurch, New Zealand. 

Muguruma, H., Nishiyama, M. and Watanabe, F. (1993). Stress-strain curve model for concrete with a 

wide range of compressive strength. Third International Conference on Utilisation of High Strength 

Concrete, Lillehammer, Norway, pp. 314-321. 

Muguruma, H. and Watanabe, F. (1990). Ductility improvements of high-strength concrete columns 

with lateral confinement. High Strength Concrete - Second International Symposium, SP-121, pp. 47- 

60. 

Nagashima, T., Sugano, S., Kimura, H. and Ichikawa, A. (1992). Monotonic axial compression test on 

ultra-high-strength concrete tied columns. 10th World Conference on Earthquake Engineering, 

Madrid, pp. 2983-2988. 

Nishiyama, M., Fukushima, I., Watanabe, F. and Muguruma, H. (1993). Axial loading tests on high 

strength concrete prisms confined with ordinary and high strength steel. Third International 

Conference on Utilisation of High Strength Concrete, Lillehammer, Norway, pp. 322-329. 

Razvi, S.R. and Saatcioglu, M. (1994). Behaviour of high-strength concrete columns confined with 

circular spirals', 10th European Conference on Earthquake Engineering, Vienna, Austria, pp. 1643- 

1648. 

Razvi, S.R. and Saatcioglu, M. (1996). Confinement of high-strength concrete columns for seismic 

applications, llth World Conference on Earthquake Engineering, Acapulco, Mexico, Paper no. 1855. 

Sugano, S. (1996). Seismic behaviour of reinforced concrete columns which used ultra-high-strength 

concrete, llth World Conference on Earthquake Engineering, Acapulco, Mexico, Paper no. 1383. 

Sun, Y.P., Oba, T., Tian, F.S. and Ikeda, T. (1996). Confinement effects of transverse hoops in highstrength 

concrete, llth World Conference on Earthquake Engineering, Paper no. 1363. 

Tanaka, H., Sato, Y., Park, R. and Kani, N. (1994). High-strength concrete columns with longitudinal 

reinforcement of mixed grades. Proceedings of AC1 International Conference on High Performance 

Concrete, Singapore, pp. 391-411.




PERFORMANCE-BASED SEISMIC DESIGN OF STEEL MOMENT 

FRAMES USING TARGET DRIFT AND YIELD MECHANISM 

Subhash C. Goel and Soon-Sik Lee 

Department of Civil and Env. Engineering, University of Michigan, Ann Arbor, MI, U.S.A. 

ABSTRACT 

This paper presents a seismic design procedure based on performance limit states using plastic 

mechanism method and the concept of energy balance. In this study, using the concept of seismic force 

reduction factor and the displacement amplification factor, the energy balance equation is modified for 

the design procedure. A new seismic lateral force distribution based on nonlinear dynamic analyses is 

also presented. This distribution is applied to the performance based plastic design procedure for steel 

moment frames. The results of nonlinear static and dynamic analyses of an example steel moment 

frame designed by using the proposed method are presented and discussed. The results show that the 

structure performed well under design level ground motions as intended, with story drifts within the 

target limit. 

INTRODUCTION 

Many studies have shown that building structures designed by modern seismic code procedures are 

expected to undergo large cyclic deformations in the inelastic range when subjected severe earthquake 

ground motions. Nevertheless, most seismic design codes are still based on elastic methods using 

equivalent static lateral design forces. This procedure can result in unpredictable and poor response 

during severe ground motions with inelastic activity unevenly distributed among structural members. 

Leelataviwat (1998) developed a new performance-based plastic design procedure using the concept of 

energy balance applied to a preselected yield mechanism, with adequate strength and ductility. This 

study is an extension of the previous study by Leelataviwat (1998). It is well known that force 

reduction and displacement amplification factors, intended to account for damping, energy dissipation 

capacity as well as overstrength, have important roles in seismic design. However, since the previous 

proposed design method (Leelataviwat, 1998) did not consider the above factors as influenced by the 

structure period, the method can result in conservative design for long period structures and 

unconservative design for low-rise, short period structures.

For simplicity, a linear distribution of the lateral design forces has been generally used in the codes 

However, many studies have shown that this distribution may not be applicable in the inelastic stage 

and may underestimate the story shears. It can also be too conservative for the design of columns in the 

performance based plastic design procedure. Moreover, this distribution does not satisfactorily 

recognize the higher mode effects for high-rise building structures. In this study, a new distribution of 

the lateral forces is presented and applied to a new performance based plastic design procedure. The 

results of nonlinear static and nonlinear dynamic analyses of an example steel moment frame designed 

by the new method are also presented and discussed. 

MODIFIED ENERGY BALANCE EQUATION 

286 

In the earlier study (Leelataviwat, 1998), the energy balance equation was written as: 

E = E.+E p (1) 

where E(= l/2MS^ is the elastic input energy, and E d and E p are the elastic and plastic components 

of work required to push the structure monotonically up to the target drift. Based on studies by 

Newmark and Hall (1982) and Uang (1994), Eqn. 5 can be modifies as: 

7E = (E e + E p ) (2) 

where 7 is a modification factor, which depends on the structural ductility factor ( ju & ) and the ductility 

reduction factor ( R^), 

Fig. 1 shows the relationship between the base shear ratio (C) and the drift ( A ), and Eqn. 2 can be 

written as: 

(3) 

Using the expression for drifts ( A ), Eqn. 3 can be rewritten as: 

where A tf and A inax from Fig. 1 are equal to R^ A v and // 4 A y , respectively. Substituting these terms in 

Equation 8, the energy modification factor y can be determined as:

287 

Figure 1 Structural Idealized Response-Application Principle of Energy Conservation 

(5) 

According to Eqn. 5, the modification factor is a function of the ductility reduction factor and the 

structural ductility factor Using different approaches, many investigators have attempted to estimate 

the ductility reduction factor (R M ) and structural ductility factor (// 4 ) (Miranda and Bertero 1994). In 

this study, the proposal of Newmark and Hall (1982) is used to estimate the ductility reduction factor 

and the structural ductility factor. 

The design input energy can be determined from the elastic design pseudo-acceleration spectra as 

given in the building codes. In this study, the design is based on the UBC (1997) design spectrum 

which, for elastic systems, is specified as: 

A = ZlCg = ag (6) 

where A is the design pseudo-acceleration, / is the importance factor, Z is the zone factor, C is the 

elastic seismic coefficient, g is the acceleration due to gravity, and a = ZIC. The energy balance 

. |W M.-2 

- Wi ** 3 

W\J n.= 4 

\y *»s 

Accln ; „ Velocity, ** 

Region "*["""* Displacement Regions 

Period (sec) 

Figure 2. Ductility reduction factors 

proposed by Newmark and Hall (1982) 

Figure 3. Modification factors for 

energy equation versus period

equation can be rewritten as: 

288 

l -rM(âg (7) 

Akiyama (1985) showed that the elastic vibrational energy can be calculated by assuming that the 

entire structure is reduced into a single-degree-of-freedom system, i e., 

where Vis the yield base shear and Wis the total seismic weight of the structure (W=Mg). Substituting 

Eqn. 8 into Eqn. 7 and rearranging terms gives: 

• "> 

NEW LATERAL FORCE DISTRIBUTION FOR PROPOSED DESIGN PROCEDURE 

Inelastic dynamic analyses were conducted to determine the distribution of maximum story shears. The 

nonlinear analysis program SNAP-2DX [Rai et al., 1996] was used to perform the analyses. The study 

frame was subjected to four selected earthquake records (Lee and Goel, 2000). Fig. 4 shows the 

example 5-bay, 9 story steel frame. The relative distributions of maximum story shears due to four 

selected earthquake records and the UBC static story shears are shown in Fig. 5. The ratio of the 

earthquake induced story shear at level / to that at the top level, n, is assumed to be of the form: 

where V t and V n , respectively, are the static story shears at level / and at the top level as computed 

from the design forces given by the UBC lateral force equations. The value of exponent b = 0.5 was 

derived from many analyses using different types of structures following the strong column- weak 

beam mechanism (Lee and Goel, 2000). 

Using the static story shears, V l and V n , at level i and the top story provided by UBC formulas, the 

factor can be written as: 

w,h, (i - o.or) + 0.077*2 

(ID

289 

W24N68 W24x68 W24x6S W24\6S WWvfiS 

\V27\84 * W27\S4 5 

3 

W27-v84 5 \V27\84 | W27\»4 

; 

W30x9

290 

DESIGN BASE SHEAR 

Equating the plastic energy term E p to the external work done by the lateral forces gives' 

Substituting Eqs. 9 and 14 into Eqn. 15 , and solving for V/W gives: 

V _-a + â" + 4yu 2 „„ 

W~ 2 

where a is a dimensionless parameter, which depends on the stiffness of the structure, the modal 

properties and the intended drift level, and is given by: 

Plastic Design of Moment Frames 

The distribution of beam strength along the height should closely follow the distribution of story shears 

induced by ground motion. Then, the required beam strength at each level can be determined as 

follows: 

fi M )h,F. (18) 

where \f p b r is the reference plastic moment of beams and the only unknown variable in the above 

equation. The plastic moment of the first-story columns to prevent the story mechanism in the first 

story can be taken as: 

where Vis the total base shear, h/ is the height of the first story, and the factor 1.1 is the overstrength 

factor to account for possible overloading due to strain hardening (Leelataviwat, 1998). 

In order to ensure the selected strong column-weak beam plastic mechanism at the ultimate drift level, 

it is important that columns are designed assuming that all beam plastic hinges are fully strainhardened 

when the drift is at the target ultimate level The detailed procedure for design of columns 

can be found in the previous study (Leelataviwat, 1998).

SEISMIC EVALUATION OF THE STUDY FRAME 

291 

The example frame was designed by the proposed design procedure. Fig. 6 shows the resulting 

member sizes of the frame. In this study, the frame was designed by selecting an assumed 2% target 

drift, which is consistent with the drift limit imposed by the UBC, and all members with specified yield 

strength of 50 ksi. The members were designed by using AISC-LRFD specification (AISC, 1994). 

Design parameters for the example frame are shown in Table 1. 

\V33vll8 "^ 

W3H\W 

~k 

W33\118 ^ 

W3li

292 

by using the computer program SNAP-2DX (Rai et al., 1996). Strain hardening and damping values of 

2 % were used for all members. 

Fig. 7 shows the base shear versus roof displacement plot of the frame obtained from die static 

pushover analysis. As can be seen, the yield drift and the design base shear of the frame are very close 

to the values assumed in the design. Fig. 8 shows the envelopes of maximum story drifts of the frame 

due to the four selected ground motions. The envelopes of maximum story drifts show that the story 

drifts are generally within the target design limit of 2%, as expected. 

Conclusion 

A new seismic design procedure based on modified energy balance equation and plastic design concept 

was presented and discussed. A new modification factor for energy was derived, which depends on the 

structural ductility factor f ju^) and the ductility reduction factor (R M ). A new design lateral force 

distribution based on nonlinear inelastic dynamic analysis results was used. In this proposed design 

method, the story drift is specified as a design parameter and, therefore, no explicit check for ultimate 

drift is required. 

The example 9-story frame was designed by the proposed design method. Nonlinear static and 

dynamic analyses of the frame were conducted to verify the proposed method. The results show that 

the proposed method can produce structures that meet preselected performance objectives in terms of 

yield mechanism and target drift. 

References 

Lee, Soon-Sik and Goel, S.C., U A New Lateral Force Distribution for Seismic Design of Steel 

Structure," Proceedings of U.S.-Japan Workshop on Seismic Fracture Issues in Steel 

Structures, San Francisco, CA, February 28-March 1, 2000. 

Leelataviwat, S. (1998), Drift and Yield Mechanism based Seismic Design and Upgrading of 

Steel Moment Frames, Ph.D. Thesis, Department of Civ. & Env. Engrg., University of 

Michigan, Ann Arbor, MI, USA. 

Miranda, E. and Bertero, V.V. (1994), "Evaluation of Strength Reduction Factors for 

Earthquake-Resistant Design," Earthquake Spectra, Vol. 10, No. 2, 1994 

Newmark, N.M. and Hall, W.J. (1982), Earthquake Spectra and Design, Earthquake Engrg. 

Res. Inst, El Cerrito, CA. 

Rai, D.C., Goel, S.C., and Firmansjah, J. (1996), U SNAP-2DX: A General Purpose Computer 

Program for Nonlinear Structural Analysis," Report EERC 96-21, Dept. of Civ. & Env. 

Engrg., University of Michigan, Ann Arbor, ML 

Uang, C.-M. and Maarouf, A. (1994), "Deflection Amplification Factor for Seismic Design 

Provisions," J. Struc. Engrg., Vol. 120, No. 8,2423-2436, ASCE 

Uniform Building Code (UBC) (1997), Int. Conf. of Bldg. Officials, Whittier, Calif.

Proceedings of the International Conference on ««« 



A HYBRID OPTIMIZATION ALGORITHM: 

GENETIC ALGORITHM - SIMPLEX 

HAN Wei 1 ' 2 and LIAO Zhenpeng 1 

'institute of Engineering Mechanics, China Seismological Bureau, 

Harbin, China 

"Harbin Institute of Technology, Harbin, China 

ABSTRACT 

Genetic algorithm is a global optimization algorithm that attracts much attention, and has been used in 

many fields recent years. Although it can search widely in model space, its local searching ability is 

poor, especially for some complex optimization problems in earthquake engineering. In general, the 

local optimization algorithm is excellent in searching ability. To cope with the problem of low 

searching efficiency and premature of genetic algorithm, a hybrid global - local optimization 

algorithm, genetic algorithm-simplex, is proposed in this paper. Taking the advantages of effective 

searching, easy implementing and proper combining of simplex, the proposed algorithm can overcome 

the premature of genetic algorithm, and the calculating efficiency of genetic algorithm is also 

improved by using simplex to enhance the searching ability, providing new models and increasing the 

variety of the population. Numerical experiments were carried out by seven test functions that can 

rigorously examine the searching ability of algorithm in various aspects. Compared with several other 

global optimization algorithms, the searching ability of the proposed genetic-simplex is verified. 


Genetic algorithm (GA) is a global optimization algorithm which uses simple encoding technology to 

express complex parameters. By genetic operation and natural selection, it can search model space 

intelligently. Because of its population searching mode, genetic algorithm can search many locations in 

the solution space synchronously. Moreover, natural selection and simple genetic operation make 

genetic algorithm easy to use and is not restricted by some additional conditions and information. The 

widely used genetic algorithm is simple genetic algorithm (SGA) with elitist strategy. It employs 

binary encoding and roulette wheel selection, and uses just two genetic operations: crossover and

294 

mutation. Elitist strategy means that the best individual has been found will always be kept in the 

population. The simple genetic algorithm with elitist strategy is convergence^. 

For many multi-parameters optimization problems, some algorithms have to calculate the grad of the 

function. But in nonlinear optimization problems, the grad of function can not be expressed 

analytically, and can only be calculated approximately. This will not only increase the calculating cost 

but also depress the precision of the solution. Simplex is a multi-dimensions local optimization 

algorithm which does not calculate the grad of the objective function. It has the advantages of easy 

implement, good searching ability in local area and simplicity of operation on computers. 

Simple genetic algorithm is one of the widely used optimization algorithms. But has the disadvantage 

of low searching efficiency problem, especially for multi-dimension and high precision demanded 

problems. Many efforts have been put by researchers to speed up the convergence of the genetic 

algorithm. For example, Cui 3 has obtained some improvement by combing GA with simulated 

annealing algorithm. In order to improve the search capability, a hybrid optimization algorithm was 

developed in this paper which combines the genetic algorithm and the simplex. In the proposed 

algorithm, the convergence is guaranteed by genetic algorithm, and simplex can speed up the 

optimization. Further more, the validity of this hybrid algorithm was checked by seven test functions 

and compared with other two algorithms. The result shows that the proposed hybrid algorithm has high 

searching efficiency. 

2. GENETIC ALGORITHM - SIMPLEX 

Premature is the main factor that affects the speed of convergence of GA. To cope with it, genetic 

algorithm-simplex is presented, and the main procedure is shown in Fig 1. 

2.1 Objective Function and Initial Population 

In many actual engineering optimization problems, observed data are used to identify the parameters. 

For example, as the observed data are Fc(i), i=l,2,-,M, M is the number of the parameters that to be 

identified in the model, the objective function can be defined as: 

in which, F L (i) is the theoretic value. If the model space is B, its size is K, the objective of optimization 

is to find a special model, and its M parameters meet the condition, 

(2.1) 

$ ^ s (2.2) 

where s is convergence precision, is a pre-chosen, small positive number.

295 

Genetic algorithm starts from a group of individuals, namely population, and its size is noted as N. 

Then the primal parameters of model can be given by: 

S, = S imin + ^(Staax - Si rai n) (i=l,2,...,M) (2.3) 

in which, S imax and Si min are respectively the upper and lower limits of parameter S, that need to be 

identified. ^ is a uniform distributed random number between 0 and 1. 

2.2 Genetic Operation 

2.2.1 Selection 

After the initial population was obtained, the objective function of every model can be calculated by 

(2.1). Objective function is the direct express of the quality of the model, and is the basic standard to 

distinguish the individuals. There are many methods to express the quality of individual in genetic 

algorithm, and the relative fitness used in the paper is 

(2.4) 

where T is a constant, £ is a random distributed variable, and £e[0,l]. If ^\|/, <

296 

2.3 Premature Judgment and Combined with Simplex 

The premature standard in this paper is that the best fitness in the population does not change m 20 

successive generations. 

I Initial Population X, . t=Q | 

| Objective function of every individual in X, 

Convergence 

Fitness of every individual in X, 

Fitness-proportionate selection of N individuals from X, 

I Crossover between

297 

genetic algorithm is N, then the demand of the hybrid algorithm is N>M+1 During the optimization 

process, if the GA is in the premature status, the hybrid algorithm will select M-t-1 individuals from the 

population as the beginning of simplex. Then the simplex begins its optimization calculation. If the 

simplex finds the solution satisfies (2.2), the optimization ends. Otherwise, simplex can find some 

better solutions generally after a number of calculations. Then the hybrid algorithm will replace M+l 

individuals selected randomly in the population with new better solutions found by simplex, and the 

genetic algorithm continues. Use GA and simplex alternately until the best solution is found. The flow 

chart of genetic algorithm - simplex is show in Fig. 1. 

3. TEST FUNCTIONS AND HYBRID ALGORITHM TEST 

In order to test the validity of the hybrid algorithm, seven complex test functions were selected. In the 

test, the objective function is defined as O = 

best value - function value 

best value 

. For the first four test 

functions, only two-dimension cases were calculated, and for other three functions, both 

two-dimension and multi-dimension cases were calculated. 

3,1 Test Function 

(1) Fi: Shubert function 3 (-10

298 

(3) F 3 : De Jone's F 5 3 (Shekel's Foxholes) function (-32

299 

3.2 Test and Results 

In the test, three algorithms were compared: hybrid algorithm, genetic-simulated annealing algorithm 

developed by Cui 3 (SAGA), and simple genetic algorithm with elitist strategy (SGA). The evaluation 

standard is the calculation number of test function by algorithms. In the test, if an algorithm can find 

the global optimization solution in 20 successive calculations, the algorithm is considered valid, and 

the average of these 20 calculations is shown in Table 3.3, If one algorithm can not find the solution 

using 10 times calculation number of other algorithms, it is considered "divergent". In the test, the 

population size is 50, the crossover probability is 0.6, the mutation probability is 0.05, and the 

convergence precision is shown in Table 3.2. The test results are shown in Table3. The number in the 

bracket is the root-mean-square deviation. 

TABLE 3.2 CONVERGENCE PRESISION s 

Test Function 

Num of Dim. 

£ 

Test Function 

Num of Dim. 

s 

F, F 2 

2 2 

2.03xlO" 3 

F 4 

2 

F 3 

F 5 

2 

2 5 10 15 

2.71x10° 1 00x10° 5-OOxlO" 3 1.00x10° 1 00x10° 1.00x10 

3 1.00x10° 

F 6 

2 

l.OOxlO" 3 5 

1 OOxl O" 3 10 

1 OOxl O' 3 15 

1.00x10° 

F 7 

2 

5 

10 

5.00x1 Q- 2 4.00x10' 2 2.80x1 0" 2 

TABLE 3.3 

Test 

Function 

Fi 

F, 

F 3 

F 4 

F 5 

F & 

F 7 

TEST RESULTS 

Dim. 

M 

2 

2 

2 

2 

2 

5 

10 

15 

2 

5 

10 

15 

2 

5 

10 

Hybrid Algorithm 

16052(9896) 

2816(1137) 

12550(10483) 

16851(17565) 

6813(4636) 

29129(4373) 

369000(141868) 

9975064(1823640) 

475(129) 

3854(1094) 

124386(52570) 

1625996(1537333) 

^_ 419(262) 

6344(3623) 

135385(86500) 

Calculation Number 

SGA 

14180(17313) U) 

2816(1137) 

12550(10483) 

- 

- 

_ 

- 

- 

475(129) 

3854(1094) 

284960(1 61 227) {2) 

- 

419(262) 

6344(3623) 

151883(144991) 

SAGA 

1 8608(26048) 13) 

2816(1137) 

12550(10483) 

(4) 

_ 

_ 

- 

- 

475(129) 

3854(1094) 

287574(1 66246) 15) 

- 

419(262) 

6344(3623) 

107984(51991) 

Note 

- : Divergent 

(1): 10 convergence, the number is average of the other 10 converged calculations; 

(2): 13 divergence, the number is average of other 17 converged calculations; 

(3): 13 divergence, the number is average of other 7 converged calculations; 

(4): seldom convergent; 

(5): 5 divergence, the number is average of other 15 converged calculations.

300 

3.3 Test Conclusions 

From the test results, some conclusions can be obtained. First of all, for some test functions (e.g. F 2 , F 3 

and some cases of F 6 , Fy), three algorithms are equivalent. The reason is that genetic algorithm has 

strong searching ability for some optimization problems, and can find the solution directly Secondly, 

hybrid algorithm has the best searching ability among three algorithms. It can find solutions in all test 

functions. Furthermore, the hybrid algorithm has the highest searching efficiency which requires the 

least amount of calculations. 

4. CONCLUDING REMARKS 

Generally, an efficient global optimization algorithm should has local searching ability, and can 

transfer from a local optimization status to another. When genetic algorithm and simplex are combined 

together, simplex not only can enforce the local searching ability, but also provided new individuals for 

GA. The test results showed that the hybrid algorithm is an efficient optimization method. 

References 

1. PAN Zhengjun, KANG Lishan, CHEN Yuping. (1998). Evolutionary Computation, Tsmghua, China. 

2. HAN Wei, LIAO Zhenpeng. (1999). A Note on the Convergence of Genetic Algorithms. Journal of 

Earthquake Engineering and Engineering Vibration 19:4, 13-16. 

3. CUI Jianwen. (1998). The Global Optimization Method for Inverting Velocity Profile with Rayleigh 

Surface Wave Data, Institute of Engineering Mechanics, China 

4. Scrrwefel H.P. (1981), Numerical Optimization of Computer Models. John Wiley, UK.




SEISMIC SIMULATION OF PRESTRESSED CONCRETE 

BRIDGES 

Chyuan-Hwan Jeng, Y.L. Mo and Thomas T.C. Hsu 

Department of Civil and Environmental Engineering, University of Houston 

Houston, Texas, USA 

ABSTRACT 

There are three mam subjects m this research: finite fiber element analysis (FFEA) 

for reinforced concrete and prstressed concrete (RC/PC) frame structures, application of 

artificial neural networks (ANN) to seismic evaluation of prestressed concrete (PC) 

bridges, and development of finite-element (FE) software using object-onented 

programming (OOP), By employing force-based FFEA, nonlinear static and dynamic 

analyses of RC/PC frame structures were studied with investgation into the effect of the 

material constitutive models on the FFEA analysis and the solution algorithm of the 

FFEA procedures. Multiplayer perceptron networks (MLP) were applied to model the 

earthquake excitation - cntical structural response of PC bridges based on the database 

constructed from the analytical results of the FFEA. An augmented form of MLP was 

proposed to improve the modeling accuracy of the classical MLP. 

INTRODUCTION 

It is well known that reinforced concrete (RC) structures designed according to 

current design codes will respond inelastically to the maximum expected earthquake. The 

capability of RC structures to resist earthquake attacks relies heavily upon the nonlinear 

load-carrying capacity demanded implicitly by design codes through such indices as 

ductility and formation of plastic hinges. The majority of the seismic energy imparted on 

an RC structure is dissipated by these nonlinear mechanisms, and the proposition of 

contemporary structural engineering that structures are allowed to be moderately 

damaged without collapse under severe earthquake attacks is thus assured. 

This paper presents the study and applications of state-of-the-art finite element (FE) 

analysis of reinforced concrete (RC) and prestressed concrete (PC) frame structures and 

the investigation of the effect of material constitutive models on fee FE analysis. The 

objective was to accomplish a complete set of finite element software for nonlinear 

analysis of RC/PC frame structures subjected to cyclic static load or earthquake

30; 

excitation. Also, the software was built following the object-oriented (OO) programming 

paradigm by using Or computer language. 

To trace and simulate the overall behavior of the structures subjected to earthquake 

excitation, various types of non-linear finite element analysis for reinforced concrete 

frame structures have been pursued and investigated over the last thirty years. Most of 

them can be divided into three categories: local models (microscopic or macroscopic 

finite element models), global models (member models), and semi-local models (fiber 

element models) (Miramontes et al. 1996, Meyer et al. 1991). 

Semi-local models stand between global models and local models (Miramontes et al. 

1996). A simplified kinematic hypothesis was adopted and equations of equilibrium are 

solved only for nodal points, like global models. By employing a cyclic constitutive 

relationship of the constituent materials, stress and strain were calculated at local level 

and consequently no hysteretic model at section or member level was needed, like local 

models. Fiber models exploited in this research fall into the category of semi-local 

models and have been investigated by many researchers (Taucer et al. 1991, Rubianobenavides 

1998). Despite many merits, fiber models had been considered to be too 

expensive for practical structural analysis. As personal computers have been getting 

cheaper and more computationally efficient, however, fiber models may become a 

rational and feasible tool for practical structural analysis. In this research, force-based 

finite fiber elements for RC frame structures and its applications were investigated. 

In addition to the precise nonlinear FE analysis of the overall structural responses, 

larger-scaled quick estimation of critical structural responses are convenient tools for preearthquake 

damage evaluation and emergent post-earthquake repair. In this research, the 

application of artificial neural networks (ANN") to estimate the critical structural 

responses was explored. The application of multiplayer perceptron (MLP) networks was 

studied and an augmented form of MLP was proposed to improve the modeling accuracy 

of classical MLP. 

CYCLIC STRESS-STRAIN RELATIONSHIP OF REINFORCED CONCRETE 

MATERIALS 

Reinforced concrete is actually a composite material, in which its constituent 

materials, reinforcing steel and concrete, are subjected to quite a different loading 

condition than that of the tests for the individual materials. Hence, another approach to 

the investigation of RC materials is to derive the average (smeared) behavior of steel and 

concrete of reinforced concrete directly from experiments on larger reinforced concrete 

specimens, such as the tests on panel element conducted at the University of Houston 

(Hsu 1993, Mansour et al. 2001). Such constitutive relations can be defined as reinforced 

concrete material models (Meyer et al. 1991). 

This study was aimed at analyzing RC/PC frame structures subjected to cyclic static 

loading or dynamic earthquake excitation. With this goal, five material models (Jeng 

2002) were employed for the finite element analysis to attain sufficient accuracy and to 

maintain computational efficiency as well. The first four models are shown in Figs. 1 to 4.

303 

Except the yield stress, the fifth model (SteelJ02) is the same as the fourth model 

(SteelJOl). In the fifth model (SteelJ02) the average yield stress proposed by Hsu (1993) 

was used. 

Fig. SConcreteJOl 

Fig. 4 SteelJOl 

rMPLEMENTATION 

The proposed finite element program uses the smeared crack model for the main 

body of a reinforced concrete element in a bridge. The smeared crack model is suitable 

because a concrete bridge can be conveniently divided into small elements, each 

behaving in an approximately uniform manner. The average stress-strain relationships of 

concrete and steel at the element level should capture the primary load-deformation 

characteristics of the whole concrete bridge. 

Using OpenSees (Open System for Earthquake Engineering Simulation) (Fenves et al. 

2001) as a framework, a program (NRCFrame) for nonlinear flexural response analysis of 

concrete frames has been developed at the University of Houston (UH) (Jeng 2002) that 

can be used to perform linear and nonlinear dynamic response analysis. This program 

was validated through comparisons with results from shake table tests on prestressed

304 

concrete frames as documented in literature (Kowalsky et al. 2000). The program was 

then used to evaluate the ability of concrete bridges to withstand seismic disturbance. 

VALIDATION 

The dynamic tests of a prestressed lightweight concrete bridge model specimen were 

conducted at the University of California - San Diego (Kowalsky et al. 2000). The 

dimensions and configuration of the specimen are shown in Fig. 5. The columns were 

circular RC columns with 203 mm -diameter cross section. The girder was 381 mm x381 mm 

square PC girder. A total prestress force of 445 kN was applied through four 15 mm 

Dywidag bars. It was a shake table experiment; the specimen was subjected to a 

predetermined ground motion, A mass block of 11643 kg was attached to the girder. 

The fiber discretization of the cross sections of the fiber element model is shown in 

Fig. 6. Since only two-dimensional analysis is needed, concrete of the cross section is 

divided into layer fibers. Seven and six control sections for girder and columns are used 

respectively. The material parameters used in the analysis are as indicated in Tables 1 and 

2. The prestress forces were simulated with two horizontal concentrated nodal forces at 

the two end nodes of the girder. The prestressing tendons were modeled by the offset 

stress and strain with offset yielding strength due to the pretension. It was reported in 

(Kowalsky et al.2000) that the strength of the steel was increased by strain rate and an 

average yield stress enhancement of 13% was applied to their section analysis with better 

agreement with experiment. Thus the average yield stress enhancement of 13% of steel 

was also Mowed in the fiber element analysis. 

Table 1 Analytical material parameters of steel 

/,(MPa) 

Hardening Ratio 

Es(MPa) 

dsu 

Main Bar (DIORebar) 

514.15 

0.006 

200000 

0.09 

Tendon (offset values) 

419.88 

0.0123 

186000 

— 

Table 2 Analytical material parameters of concrete 

/'c(MPa) 

do 

/«,(MPa) 

dm 

Unconfined 

35.5 

0.004 

15 

0.02 

Confined 

36 

0.0048 

16 

0.023

305 

2794 r 

f" 

A 

Fig. 5 Outline of the UCSD bridge bent. 

As =12-010 

A.A 

B-B 

Fig. 6 Fiber discretization of the cross sections. 

The 1978 Tabas earthquake was chosen as the reference ground motion. The testing 

procedure involved subjecting the specimen to the selected input motion at various levels 

of excitation. The testing runs corresponding to service state and survival state are 

analyzed in this study and are referred to as Case A and Case B subsequently. A 

comparison of the analytical and experimental displacement history for Case A in the 

service state is shown in Fig. 7. The comparison shows good agreement. 

Fig. 7 Comparison between the analytical and experimental displacement history 

of Case A

30€ 

NEURAL NETWORKS 

Development 

In the case study, the UCSD PC bridge bent analyzed previously was selected as the 

reference structure. A nonlinear finite fiber element model of an enlarged PC bridge is 

composed as the target structure to be modeled, employing the nonlinear fiber frame 

element of the developed software. A total of 86 cases of the FE model subjected to the 

ground motion of the 1940 El Centro earthquake were analyzed using the developed 

nonlinear FE software. The analytical results were then used as the training and testing 

data for the ANNs. 

The magnitude and orientation of the horizontal ground motion (1940 El Centro 

earthquake) were selected as the input parameters of the ANN. The maximum biaxial 

bending moments of columns and girder and the maximum horizontal displacement of 

the girder were selected as the outputs of the ANN. Seven MLP networks with different 

number of hidden layers and hidden neurons were first tried. An augmented form of MLP 

was used to improve the modeling accuracy. 

The magnitude of the earthquake was represented by the scale factor for the ground 

acceleration record of El Centro earthquake. Scale factors ranging from 1.0 to 3.1 were 

used, which corresponds to a PGA level ranging from 0.32g to 0.99g. The orientation of 

the earthquake was represented by the inclination angle between the longitudinal axis and 

the incident direction of the ground acceleration, as the a angle. Considering the 

symmetry of the structure, a range of a angle from 0° to 90° was investigated. Thus, 86 

cases of fiber element analysis were conducted, the outcomes of which were summarized. 

The bi-axial maximum member-end moments of the piers and the girder and the 

maximum horizontal node displacements of all 86 cases were also summarized 

It was found that the bending moment at the bottom end of the piers was generally 

greater than that at the top end of the piers. Hence, only the bending moment at the 

bottom of the piers was chosen as one of the output parameters of the ANNs. Also, the 

maximum bending moments of the two piers were very close; thus, only one set of the 

maximum bi-axial bending moments M y , M z of the two piers was chosen as the 

representative critical bending moments of the bridge piers. 

Similarly, a pair of the maximum bi-axial bending moments of the girder was 

chosen from the maximum bending moments at the two ends of the girder; however, 

since the minor-axis bending moment M y of the girder was generally very small, only the 

major-axis bending moment M 2 was chosen to be investigated by the ANNs. The 

maximum value of the displacements at the two nodes was selected as the representative 

maximum horizontal displacement of the girder. Therefore, there were 4 output 

parameters altogether of the ANNs, that is, the maximum bending moments M y and M 2 of 

the piers, the maximum bending moment M z of the girder, and the maximum horizontal 

displacement of the girder.

307 

MLP Networks 

Seven fully-connected multiplayer perception (MLP) networks were constructed to 

model the critical structural response of the target PC bridge subjected to the ground 

motion of the 1940 El Centro earthquake of various magnitudes and orientations. To 

improve the predicting performance, an augmented type of MLP was used in the next 

section. 

Augmented MLP Networks 

To improve the modeling accuracy of the classical MLPs mentioned above, an 

augmented form of MLP was used. The augmented MLP was formed by adding two 

additional exponential input nodes to the classical MLP. The exponential input nodes 

were derived from the incident angle of the horizontal ground motion a. 

Four augmented MLPs were constructed to model the critical structural response of 

the target PC bridge. Each of the four augmented MLPs was trained using various values 

of learning rate and momentum. The outcome with best accuracy for each of the 

augmented MLPs was used to demonstrate the comparison between the networkcomputed 

values and the desired values for the training set as well as the testing set. The 

network configuration, network training parameters, and degree of predicting accuracy of 

all the four Augmented MLPs are determined. 

It was found that the augmented MLPs generally provided considerably better 

modeling accuracy than the classical MLPs. In terms of minimum ensemble total mean 

square error, the augmented MLP (0.006886) can be considered 26% more accurate than 

the classical MLP (0.00929). 

CONCLUSIONS 

The research presented in this paper basically demonstrates the feasibility of the 

employed methodology, the application of which can be extended to other concrete 

structures. The following conclusions may be drawn based on this research: 

1. The inclusion of the Baushinger effect and reduced average strength effect in the 

stress-strain relationship of mild steel is the dominant factor for the improved 

performance of the finite fiber element analyses. The effect of concrete model on the 

finite element analysis is less sensitive. 

2. It was found that the use of the proposed offset stress-strain curve for prestressing 

steel in the finite element analysis generated analytical results that were reasonably close 

to the experimental data. 

3. The application of artificial neural networks to quick estimation of the critical 

response of PC bridges is feasible. The proposed augmented multilayer perception 

networks were effective in enhancing the modeling accuracy. 

4. The merits of the application of object-oriented programming to finite element 

analysis were acknowledged.

308 


The research reported in this paper was founded by the National Science Foundation 

under Grant No. CMS-9711084, USA and the Sinotech Engineering Foundation, Taiwan. 

REFERENCES 

Fenves, G. L., McKenna, F., Scott, M. H., and et al. (2001), OpenSees User and 

Developer Workshop, Workshop Handout, Pacific Earthquake Engineering Research 

Center, University of California, Berkeley, CA, August 20-23. 

Hsu, T. T. C. (1993), Unified Theory of Reinforced Concrete, CRC Press, Inc., Boca 

Raton, Flonda, pp.205-217. 

Mansour, M. Y., Lee, J.Y., and Hsu, T. T. C. (2001), "Cyclic Stress-Strain Curves of 

Concrete and Steel Bars in Membrane Elements," Journal of Structural Engineering, 

s Vol. 127, No. 12, December, pp. 1402-1411. 

Jeng, C.H. (2002). Nonlinear Finite Fiber Element Analysis for Static and Dynamic 

Simulation of Concrete Frames: Modeling, Implementation, and Application Ph.D. 

Dissertation, Department of Civil & Environmental Engineering, University of Houston, 

Houston, Texas. 

Kowalsky, M. I, Priestley, M. J. N., and Seible, F. (2000), "Dynamic Behavior of 

Lightweight Concrete Bridges," AC! Structural Journal Vol. 97, No. 4, JuL-Aug., 

pp.602-618. 

Meyer, C, de Borst, R., Bicanic, N, Filippou, F.C., and Maekawa, K. (1991), "Chapter 6, 

Dynamic Loading," Finite Element Analysis of Reinforced Concrete Structures II, 

Proceedings of the International Workshop, ASCE, June 2-5. 

Miramontes, D., Merabet, O. and Rynouard, J. M. (1996), "Beam Global Model for the 

Seismic Analysis of RC Frames," Earthquake Engineering and Structural Dynamics, Vol. 

25, pp. 671-688. 

Taucer, F., Spacone, E., and Filippou, F. C. (1991), A Fiber Beam-Column Element for 

Seismic Response Analysis of Reinforced Concrete Structures, Research Report No. 

UCB/EERC-91/17, Earthquake Engineering. Research Center, University of California, 

Berkeley, CA.




3Qg 

PROGRESS IN THE SEISMIC DESIGN AND RETROFIT OF 

BRIDGES IN KOREA, A MODERATE SEISMICITY REGION 

Jae Kwan KIM ! , Ick Hyun KIM 2 , Gui-Hyun JUHN 3 and Dae Yeon CHO 4 

1 Associate Professor, Department of Civil, Urban and Geosystem Engineering, 

Seoul National University, Kwanak-ku, Seoul, 151-742, Korea 

E-mail: jkwankim@plaza.snu.ac.kr 

2 Assisstant Professor, Department of Civil and Environmental Engineering, 

University of Ulsan, Ulsan, 680-749, Korea 

3 Research & Development Division, Korea Infrastructure Safety & Technology Corporation, 

Ilsan-ku, Koyang, Kyounggi-do, 411-410, Korea 

4 REITs and Venture Team, Research Center, Korea Highway Corporation, 

Seongnam, Kyounggi-do, 461-703, Korea 

ABSTRACT: New performance-based seismic design concept had been developed in 1997 and 

included in 2000 edition of Korea Highway Bridge Standards. The limited ductility design approach 

appropriate to Korea is under active development. In this year the manual for the seismic assessment 

of highway bridges was drafted and approved. The manual for retrofit will be drafted and published 

soon. Because it is required by law, many old bridges will be retrofitted in near future. In this seismic 

assessment manual a new philosophy is introduced in the determination of design earthquake for 

retrofit of bridges. In this paper, this new concept is introduced briefly. In addition the Korean efforts 

in the seismic design and retrofit of highway bridges are briefly reported. 

INTRODUCTION 

Even though Korea is located rather far away from the active plate boundary many historic records and 

recent seismic activities indicate Korea belongs to a region of moderate seismicity. In spite of seismic 

hazard the earthquake resistant design was introduced to Korea for the first time in 1992 for highway 

bridges. But the energetic study and research to develop new seismic design concept has been 

activated since the 1995 Kobe earthquake. As a result performance-based design concept was 

developed in 1997 and included in 2000 edition of Korea Highway Bridge Standard partially. The 

details of this standard are basically similar to those of the AASHTO code of USA. However the 

seismic response characteristics of structures in moderate seismicity regions has not been fully taken 

into account in the present code. It is clearly recognized that the limited ductility design will be 

appropriate in the moderate seismicity regions. Currently research efforts are under progress to 

develop the details for the implementation of limited ductility design for the highway bridges. In this 

paper the research progress and design details of landmark bridges is reported.

310 

In order to ensure the entire social safety against seismic hazard the sufficient seismic performance has 

to be secured for the existing structures as well as new ones. In Korea many multi-span continuous 

bridges had been constructed in which earthquake load was not considered adequately. They can be 

very vulnerable to the earthquake ground motion especially in longitudinal direction. Hence many 

alternative retrofit methods are considered and/or implemented such as isolation bearings and shock 

transmission units. The project to develop the seismic performance assessment method has been 

initiated by the Ministry of Construction and Transportation of Korea (KMOCT). In this year the 

manual for seismic assessment of bridges was drafted and approved. It will be distributed soon. In this 

manual a new philosophy is introduced in the determination of design earthquake for the retrofit of 

bridges. In this paper, this new concept is introduced briefly. And several retrofit examples are 

reported. 

SEISMIC DESIGN OF BRIDGES IN KOREA 

Seismic Design Code of Bridges 

The history of seismic design in Korea is not as long as that of countries in high seismicity region. It 

has been introduced to Korea since 1986 for high-rise buildings, since 1992 for highway bridges and 

even earlier than that for nuclear power plants. The basic idea in these codes has been borrowed 

heavily from those developed for the high seismicity region. However the public did not aware the 

necessity of seismic design in general until Kobe earthquake yielding severe damage and huge 

casualties in 1995. Taking opportunity of this earthquake KMOCT decided to develop new 

performance based seismic design code system that can be commonly applicable to most types of 

facilities. As the first step, the high-level criteria of performance-based seismic design had been 

completed to regulate seismic design loads and seismic performance level of structures. Based on this, 

the low-level criteria and technical standards specific to each own type of facility have been under 

development. The 2002 edition of Korea Highway Bridge Standard (KHBDS) adopted this design 

concept. 

In Korea, the seismic design has been carried out in two ways. The one is relying on the ductility of 

column bents, and the other is adoption of various types of earthquake protection systems such as 

isolation systems. For the former, the identical reinforcement details to those in zone II of AASHTO 

code have been adopted as shown in TABLE 1. It is remarkable that the lap splice of longitudinal bars 

in the plastic hinge zone is implicitly allowed. Due to this lap splice detail same amount of transverse 

reinforcement as specified in high seismicity zone is required to attain satisfactory ductility. Unlike 

high seismicity zone, the limited ductile behavior is expected in the moderate seismicity regions. In 

Korea the sectional dimension of piers designed conventionally is as large as 3.5m diameters. 

Therefore they can have considerable amount of inherent lateral load resistant capacity. However the 

current codes requires relatively larger amount of transverse rebar than necessary. Therefore, it is 

necessary to develop limited ductility design that is appropriate to the seismicity of Korea and the 

characteristics of columns. This design approach should take advantage of the inherent seismic 

capacity of conventional designed piers.

311 

TABLE 1. 

COMPARISON OF SEISMIC DETAILS BETWEEN AASHTO CODE AND KOREAN CODE 

Desc. 

Zone 

Seismicity 

(A) 

Lap Splice 

Transverse 

reinforceme 

nt 

I 

A < 0.09 

No 

consideratio 

n of seismic 

forces : 

Convention 

al design 

n 

0.09

312 

Examples of Seismic Design 

In Korea, as in other moderate seismic regions, using the strength and ductility of columns is sufficient 

to meet performance criteria in many types of bridge system. But earthquake protection devices such as 

LRB (Lead Rubber Bearing), STU (Shock Transmission Unit) and viscous damper are indispensable in 

multi-span continuous bridge, cable-stayed bridge and cable-suspended bridges. As an example of 

seismic design Seohae Bridge can be supposed, which consists of 3 different types of bridges: cablestayed 

bridge (990m), FCM bridge (500m), and PSM bridges (5820m). It was designed with 

earthquake load of only 6% of self-weight and constructed. According to the current code, the 

earthquake force can be as high as 35% of self-weight. So, Korea Highway Corporation decided to 

retrofit these bridges. Strictly speaking the methods introduced here was applied as retrofit methods. 

But the same methods can be implemented to new bridge system as alternative seismic design concept. 

The 97 spans of PSM bridges are grouped as 16 bridges: 3 to 10 continuous span bridges. If the bridges 

with flexible piers can withstand the earthquake force by increasing the number of fixed piers concrete 

block and/or restrainers was implemented as shown in Fig 1. On the other hand, this method was not 

applicable to the bridge with stiff piers. The viscous fluid dampers were used, which placed between 

the superstructure and the pier as shown in Fig. 2. In 5-span continuous cables-stayed bridge, of which 

center span length is 470m and side span length is 60m, large earthquake load would concentrate on 

the fixed pylon by the seismic analysis. So, LUD (lock-up device; it is called Shock Transmission Unit 

also) was adopted to distribute this force to the other pylon as shown in Fig. 3. It can provide a 

temporary rigid link between bridge structural members under seismic, braking, or other dynamic load, 

while permitting slow thermal movements. It was the first application in Korea. 

FIGURE 1. 

CONCRETE BLOCK AND RESTRINER 

FIGURE 2. 

VISCOUS DAMPER

313 

FIGURE 3. 

LOCK-UP DEVICE 

SEISMIC RETROFIT 

Progress of Seismic Retrofit in Korea 

When the necessity of seismic design began to get support from the public after Kobe earthquake 

almost all resources have been concentrated on the research related to the development of new seismic 

design concept. Therefore the retrofit of existing structures did not receive due attention. But the 

seismic retrofit should be also treated as an urgent task in order to ensure uniform social safety against 

seismic hazard. In Korea only less than 1% of bridges are designed seismically. Fortunately it is 

realized by the preliminary studies that the seismic performance of existing bridges is adequate even if 

the ductility is not so satisfactory. But multi-span bridges were widely constructed in Korea entering 

1990's. This type of bridge has only one fixed bearing pier in the longitudinal direction in general. 

Therefore it is very vulnerable to the earthquake ground motion since a great earthquake load is 

concentrated on the single pier. The enhancement of ductility of fixed pier or the anti-seismic devices 

can be considered as a seismic retrofit method. Realizing the urgency for retrofit the MOCT has 

initiated a serious of project to develop the method for seismic assessment and seismic retrofit of 

existing bridges. As a result the manual for seismic assessment for existing bridges has been drafted in 

this year. The development of the seismic retrofit manual is on progress. 

Basic Concept of Seismic Assessment 

The manual for seismic assessment is to assist the engineers for the easy and efficient assessment and 

retrofit design. The check of capacity/demand ratio is adopted as the basic idea. Once the earthquake 

load is determined seismic performance is checked at each member/component level such as pier, 

bearing, abutment, foundation and so on. Considering the assessment efficiency it is performed through 

two- stage progress. In the first stage, the bridge considered is grouped to decide its priority by 

seismictiy, vulnerability and impact into 4 categories. The detailed assessment is carried out in the 

second stage for the bridges dropped into the critical category, the important and the observed grade. In 

this detailed assessment new philosophy is introduced in the determination of design earthquake for 

retrofit of bridges. As shown in Eq. (1) it is determined by the idea that the possibility of existing 

bridges to experience the seismic design earthquake equates to that of new ones. Therefore the

314 

earthquake load can be reduced with increase of the age of bridges. As the result many exiting bridges 

can be exempted from seismic retrofit. But if the owner wants to extend the lifetime of bridges under 

consideration, the appropriate level of seismic design earthquake should be applied. 

(The yearly occurrence probability of earthquake for seismic retrofit x target lifetime of existing bridge) 

= (The yearly occurrence probability of seismic design earthquake x design lifetime) (1) 

In fact the above idea was introduced on the background to secure urgently the seismic performance of 

a lot of existing bridges with the limited economic resource. Therefore it is still controversial due to the 

lack of validity backed up by sufficient studies. But we hope this idea can be embodied as a realistic 

and reasonable alternative to determine the necessity of retrofit by the further intensive studies. 

Examples of Seismic Retrofit 

The manual for the seismic retrofit bridges has not been completed yet. But many analytical and 

experimental studies have been performed on performance of retrofitted bridges. The dominant failure 

modes of existing bridges are the failure of bearings and the premature bond failure of lap spliced 

longitudinal bars in the fixed piers. Therefore, research efforts are concentrated on the enhancement of 

ductility by wrapping plastic hinge zone of piers with steel plates, carbon sheet, and glass fiber sheet 

etc.. In parallel the implementation of the various earthquake protection devices has been studied. 

Based on the results of these primary studies quite a few importance bridges have been retrofitted 

already prior to completion of the manual. Once the manual for retrofit is completed the retrofit of 

existing bridges will be carried out actively. 

The Figure 4 and 5 present the example of seismic retrofit of existing 4-span continuous bridge. The 

seismic assessment showed that it was very vulnerable at the bearings and the fixed pier. The bearings 

were protected by shear keys; steel brackets in the transverse direction and concrete blocks in the 

longitudinal one respectively. And the fixed pier was retrofitted by wrapping with the steel plate to 

enhance ductility. 

a) STEEL BRACKETS b) CONCRETE BLOCKS 

FIGURE 4. 

PROTECTION OF BEARINGS BY SHEAR KEY

315 

FIGURE 5. 

RETROFIT OF FIXED PIER BY STEEL PLATE 

CONCLUSIONS 

Korea belongs to a region of moderate seismic hazard. Earthquake resistant design was introduced to in 

1992 for highway bridges. New performance-based design concept was developed in 1997 and 

included in 2000 edition of Korea Highway Bridge Standards. Based on the research results on the 

performance of bridges designed conventionally the limited ductility design approach appropriate to 

Korea environment is under active development. In addition, various anti-seismic devices are actively 

implemented to the multi-span continuous bridge, cable-suspended bridge and cable-stayed bridges. In 

this year the manual for seismic assessment of highway bridges was drafted in Korea. The manual for 

retrofit also will be drafted soon. The existing bridges found to have very poor ductility due to the lap 

spliced longitudinal bars. Therefore several important bridges have been retrofitted already prior to 

accomplishment of manual. In near future many old bridges will be strengthened according to the 

manual. 

REFERENCES 

Chai, Y.H., Priestley, M.J.N. and Seible, F. (1991). Seismic retrofit of circular bridge columns for 

enhanced flexural performance. ACI Structural Journal 88: 5, 572-584. 

Kim, J.K., Kim, I.H., Lim, H.W., and Lee, J.H.(2000). Seismic upgrading of existing circular RC pier 

with steel jacket. Proc. ofEESK Conference 4:1, 341-349(in Korean). 

Kim, J.K., Kim, I.H., Lim, H.W., Lee, J.H. and Lee, J.H.(2001). Cyclic loading test of bridge pier 

models without seismic detailing. Proc. ofEASEC-8.

Proceedings of the Intel national Conference on 


Engineei mg Research Hong Kong Volume 

317 

SEISMIC BEHAVIOR OF WOODEN HOUSE 

USING DISTINCT ELEMENT METHOD 

J Kiyono 1 and A Furukawa 1 

1 Graduate School of Civil Engineering, Kyoto University, Kyoto, Japan 

ABSTRACT 

Collapsing process of wooden houses during an earthquake has been analyzed using the Distinct 

Element Method (DEM) DEM is a numerical analysis method in which the positions of the elements 

are calculated by solving the equations of motions step by step Not only individual but also group 

behavior can be simulated using this method The structure is modeled as an assembly of distinct 

elements and these elements are connected by virtual springs and dashpots Damage to human bodies 

has been considered as well The human body is modeled as a circle and put on the floor The 

maximum impact acceleration experienced by a human body during an earthquake is calculated 

Damage to human bodies in the house is discussed based on the index of Chest-G and HLC 

INTRODUCTION 

During the 1995 Hyogo-ken Nanbu earthquake, many wooden houses were destroyed More than 6400 

people were killed during this quake, and 80% of them died due to the structural failure To reduce 

casualties due to earthquakes, it is very important to know the way structures collapse and how people 

get injured because of it The purposes of this study are firstly to simulate collapsing process of 

wooden houses using DEM, secondly to estimate the impact acceleration applying to the people m the 

house, and finally to assess the damage level of people 

Rigid Body 

(a) rectangular element 

(b)circular element 

Fig.3.1 Contact model of elements

318 

METHOD 

DEM is used in this study. The DEM is a numerical analysis method that can compute the position of 

each element by solving the equation of motion step by step. All elements are assumed to be rigid. 

Each element has a spring and dashpot in its normal and tangential direction. When it comes in contact 

with others (Fig.3.1), the contact force is generated by the connected springs and dashpots. By solving 

the equations of motions for all elements step by step, the behaviors of all elements can be traced. The 

forces applying to the element are the external force (/,,/.) and the total sum of contact forces 

between elements (F x1 F z J e ). The element acceleration is obtained as follows: 

/m 

I m 

(3.1) 

(3.3) 

(3-2) 

where m is the element mass, and I 9 is the inertia moments around the center of the gravity. 

From the above equations, the velocity and the displacement of the next time step is calculated as: 

uT-A/ 

Ar (3.4) 

(3.5) 

where Ar is the time interval of the computation. 

ANALYTICAL MODEL 

Modeling of the structure 

The analyzed structure model is a two-story house as shown in Fig.4.1. Both the height and the width 

are 10m. The structure consists of four columns, two ceiling beams and a roof. The roof and the ceiling 

beam are connected by three support beams. In this study, we deal with wooden houses. All elements 

are 10cm width. The mass and inertia moments of all elements are calculated from the density of 

Japanese cypress (0.34g/cm 3 ). 

In usual DEM, elements affect only contact force each other because joint elements are not considered. 

But elements of real structures are connected at joints and behave as a continuum until joints are

319 

broken. Therefore, joint elements which connect different elements are introduced and behaviors are 

examined. 

Case A 

In Case A, joint elements are not considered. Only contact forces are taken into account and joints 

cannot resist tension forces. 

CaseB 

In Case B, joint elements are considered. So, not only contact forces but also tension forces are taken 

into account. During the 1995 Hyogo-ken Nanbu earthquake, many old wooden houses were collapsed 

because the first floor was smashed by the strong seismic motion. In many cases, second floor 

remained without being broken. We can presume that the joints of the first floor were too weak to 

sustain the strong shock of the earthauake. Therefore, we assumed that the joint strength of the first 

fl 

Table 4.1 Contact parameters 

o. _.._ n no 

Fig.4.1 Two-story structure model 

A* 

K n (N/m) 

K^N/m) 

C fl (N-sec/m) 

K n (Nlm) 

Human body 

0.277 

7.949xl0 4 

3.974xl0 4 

2.095xl0 3 

0.0 

Structured/kg) 

0.677 

3.944xl0 4 

3,944xl0 4 

1.256xl0 3 

1.256xl0 3 

Case C 

In Case C, joint elements are considered. If all joints were strong enough to sustain the earthquake, 

houses could avoid collapse and many people could survive. Therefore, we assumed that the joints of 

the first floor are as strong as those of the second floor. 

Modeling of the human body 

The human body is modeled as a circle element. The radius r and the density p have been evaluated 

as 0.26m and 1.706xl0 2 (£g/m 3 ), respectively, using physical examination results. The mass and 

inertia moments are calculated according to these values. Four human elements are put on each floor, 

and the impact accelerations experienced by these people are computed. The human elements on the 

first floor are numbered as 1, 2, 3, and 4 from the left, and those on the second floor are numbered as 

5,6, 7, and 8 from the left. 

Spring constant and Damping 

Coefficient 

Spring constant and damping coefficient (K H ,K^C a ,C^ 

of wooden houses are assumed to be

320 

proportional to the element mass and those of human bodies are obtained from a simple experiment. 

The values are shown in Table 4.1. 

Time Interval 

In DEM, a time interval of computation has a strong influence on the stability of results. If the time 

interval is too large, the result will diverge. Cundall (1974) recommends the following time interval. 

(4.1) 

From the above equation, the limitation is Ar /v> 

1UUJ 

*ytn 

AA 

AAA 

B 

A 

AA 

Q 45 60 75 

Fig.5.1 Classification of damage levels 

C 

B 

Chest-G 

0 5 10 15 20 25 30 35 40 

TIME (sec) 

Fig.6.1 Input ground motion 

ASSESSMENT OF DAMAGE TO HUMAN BODY 

In the field of the car engineering, the values of HIC (Head Injury Criteria) and Chest-G are often used 

to assess the damage to human body. We use the same criteria as HIC and Chest-G. The allowance 

level of chest is considered to be 60G, and HIC to be 1000. Using these values, the damage level of the 

people in the house can be estimated. As is shown in Fig.5.1, the damage level is classified into 6 

levels (AAA, AA, A, B, C, D). AAA is the safest level and D is the most dangerous level. 

RESULT 

We used the acceleration record of Kobe Marine Meteorological Agency during the 1995 Hyogo-ken 

Nanbu earthquake to the modeled structures (Fig.6.1). The result of each case is as follows: 

Case A 

The structure behavior in Case A is shown in Fig.6.2. The structure started tilting at 7.0 second and 

completely collapsed at 11.4 second. The computed values of Chest-G, HIC, and the human damage 

level are shown in Table 6.1 (a). In this case, joint elements are not considered. Therefore the structure 

easily collapsed and crushed the human bodies. All people are on the worst damage level 'D' and have

321 

the highest probability to get seriously injured. Compared with the people on the second floor, people 

on the first floor were undergone severer damage because they were crushed by the second floor and 

the roof. Compared with the people on the right side, people on the left side got severer damage 

because the structure collapsed to the left side as it can be seen in FIG.6.1. 

CaseB 

The structure behavior in Case B is shown in Fig.6.3. The structure started tilling at 7.6 second and 

completely collapsed at 12.4 second. The computed values of Chest-G, HOC, and the human damage 

level are shown in Table 6.1 (b). In this case, joint elements are considered, but the joint strength of the 

first floor is half of the second floor. Therefore, only the first floor collapsed and the second floor stood 

still. Because the second floor remained without collapse, the people on the second floor did not get 

injured and the damage levels are 'AAA 1 or 'AA'. Both Chest-G and HEC is lower than its allowance 

level. However, the people on the first floor get injured because they are affected by the crush of the 

second floor. Especially, No.3 and No.4 are on the worst damage level 'D' and are expected to get 

seriously injured because the structure collapsed to the right side. 

CaseC 

The structure behavior in Case C is shown in Fig6.4. Because the joints were strong enough to sustain 

the input ground motion, the structure swayed but did not collapsed. The computed values of Chest-G, 

HIC, and the human damage level are shown in Table 6.1 (c). In this case, joints have the sufficient 

strength. Therefore, the structure did not collapse and nobody got injured. All people are on the safest 

level 'AAA 1 . Compared to the people on the first floor, the criteria of the people on the second floor are 

a little bit larger. This is because the structure was swaying during the earthquake and the second floor 

was swayed more than the first floor. 

CONCLUSION 

Collapsing process of wooden house has been simulated by DEM, and the damage to the human body 

also has been estimated. The collapsing pattern and the extent of the damage vary depends on the joint 

strength. Joint elements are introduced and the influence of the joint strength has been discussed as 

well. As the collapsing process of the structure is strongly affected by the spring constant and damping 

coefficients in DEM, it is very important to evaluate the actual restoring force and damping force of 

the wooden house, and to try to simulate a more realistic failure behavior of the structure using 

parametric studies. In this study, human body has been modeled as a circle element. We have to make a 

more hurnan-like model if we want to obtain detailed damage to each body part. 

REFERENCES 

Cundall, P. A (1974). Rational Design of Tunnel Supports - A Computer Model for Rock Mass 

Behavior Using Interactive Graphics for the Input and Output of Geometrical Data. Technical Report 

MRD-2-74, Missouri River Division, U.S. Army Corps of Engineers 

National Organization for Automotive Safety & Victims' Aid (2001). New Car Assessment Japan

322 

y - -i " " \ 

(a) 10.0 (sec) (b)10.2 (sec) (c) 10.4 (sec) (d) 10.6 (sec) 

(e)10.8 (sec) (f)11.0 (sec) (g) 11.2 (sec) (h) 11.4 (sec) 

Fig.6.2 Collapsing process in Case A 

/..../ 

(a) 11.0 (sec) (b)11.2 (sec) (c) 11.4 (sec) (d) 11.6 (sec) 

(e)l 1.8 (sec) (f)12.0 (sec) (g) 12.2 (sec) (h) 12.4 (sec) 

Fig.6.3 Collapsing process in Case B 

(a)8.0 (sec) (b)9.0 (sec) (c) 10.0 (sec) (d) 11.0 (sec) 

(e)12.0 (sec) (f)13.0 (sec) (g) 14.0 (sec) (h) 15.0 (sec) 

TL 

Fig.6.4 Collapsing process in Case C


Advances and New Challenges in Earthquake 323 


THE DESIGN OF STRUCTURAL CONCRETE REGIONS FOR 

SEISMIC ACTIONS BY THE STRUT-AND-TIE METHOD 

Daniel A, Kuchma 1 and Tjen N. Tjhin 1 

'Department of Civil and Environmental Engineering, University of Illinois at Urbana-Champaign, 

Urbana,IL61801,USA 

ABSTRACT 

The Strut-and-Tie Method (STM) is gaining acceptance as a consistent design methodology for 

structural concrete regions that are subject to complex variation in strains, such as beam-column 

joints, squat walls, and pier caps. Provisions for designing by the STM were recently incorporated 

into the primary US bridge code, US building code, as well as other international codes. Although 

the STM provides a conceptually simple design methodology, the application of this method is 

usually complicated by the need to perform iterative and time-consuming calculations and graphical 

procedures. In addition, the current STM provisions are primarily for the design of structures 

subjected to non-reversed loadings and thus seismic design issues have not be properly addressed. 

This paper provides an overview of the STM and codes provisions for its use. It also discusses how 

the STM can be extended for seismic design and evaluation in the framework of performance-based 

seismic engineering. An STM design tool called CAST (Computer-Aided Strut-and-Tie), which is 

being developed to overcome the complexity of STM design process, is also presented. CAST 

functionality is demonstrated by the design example of a squat wall with openings. 

THE STRUT-AND-TIE METHOD 

B- and D-Regions 

For the purpose of selecting the appropriate 

design approach, it is useful to divide the 

structure into either B-(Beam) regions or D- 

( Discontinuity) regions. B-regions are 

where it is reasonable to assume that there is | t 

a linear variation of strain over the flexural Fig. 1 Example of B-Regions and D-Regions in a 

depth of the cross-section. D-regions are Common Structure 

where there is a complex, variation in strain 

due to abrupt changes in geometry (geometrical discontinuities) or concentrated forces (statical 

discontinuities). D-regions can be considered to extend a longitudinal distance equal to the largest 

of the depth or width of the member from the discontinuity. The distinction between B-regions and 

D-regions is illustrated in Fig. 1.

324 

Development Background 

Although both B- and D-regions are of similar importance, their design methodologies are not 

equail/well-founded. Strong rational bases have long been established for the design of B-regions, 

but empirical rules or common detailing practices are still currently adopted for D-regions. In B- 

regions, for examples, the Bernoulli beam theory is the basis for the flexural design and the truss 

analogy method is used for the shear design. For common types of D-regions, such as beam-column 

joints, "corbels, and deep beams, approaches based solely on test results have still been used. For 

other D-regions, there is little to no guidance found in literature. 

Motivated to unify the design methodology for B- and D-regions, Marti (1985) and Schlaich et al. 

(1987) promoted the STM. This 

method basically extends the use of | ^| 

the truss model employed in the 

truss analogy method. 

Design Methodology 

The STM is based on the lowerbound 

theory of limit analysis. In 

the STM, an internal trass is 

envisioned to carry the applied 

loading through the D-region to its 

supports or boundaries. The truss is 

Beam-Column Joint 

Coupling Beam in 

Coupled Shear Wai 

Non-Slender Wall 

Fig. 2 Examples of Strut-and-Tie Models 

termed strut-and-tie model. Like the steel truss, the strut-and-tie model consists of struts, ties, and 

joint (or nodes). Struts are the compression members while the ties are tension members. Nodes are 

the meeting points of struts and ties. Examples of strut-and-tie models for a few typical D-regions 

are illustrated in Fig. 2. 

The strut-and-tie model has to be in equilibrium externally with the applied loading and reactions 

and internally at each node. Reinforcing or prestressing steel is selected to serve as the ties, and the 

dimensions for struts and nodes are selected so that they have sufficient strength and do not violate 

the D-region boundary. 

Steps in Design by the STM 

The design process using STM consists mainly of five steps described below. 

1. Define the boundaries of the D-region and then evaluate the concentrated, distributed, and 

sectional forces that act on the boundaries of this region. 

2. Sketch a strut-and-tie model and solve for the truss member forces. 

3. Select the reinforcing or prestressing steel that is necessary to provide the required tie 

capacity and ensure that this reinforcement is properly anchored in the nodal zones (joints of 

the truss). 

4. Evaluate the dimensions of the struts and nodes such that the capacity of these components 

is sufficient to carry the design force values. 

5. Provide distributed reinforcement to increase the ductility of the D-region.

325 

DESIGN SPECIFICATIONS 

Overview 

The STM has been incorporated in various major structural concrete codes. These includes, among 

others, the Canadian Building Code (1984), CEB-FIP Model Code (1993), and AASHTO LRFD 

Bridge Design Specifications (1994). In 1998, a more detailed set of guidelines was published in a 

special report from the FIP (Federation Internationale de la Precontrainte) "Practical Design of 

Structural Concrete" (1998). Most recently, provisions have been developed to form Appendix A 

"Strut-and-Tie Models" to the ACI318-02 Building Code Requirements for Structural Concrete 

(2002). 

The STM design provisions typically specify rules for arranging the strut-and-tie models, rules for 

defining the dimensions and ultimate stress limits of struts and nodes, requirements for the 

distribution and anchorage of ties, and requirements for crack control and ductility detailing 

(MacGregor 2002). However, the rules vary from one code to another because of uncertainties 

associated with defining the characteristics of an idealized truss within a continuum of structural 

concrete. 

ACI Appendix A 

As described in the previous section, the latest edition of ACI318-02 Building Code (2002) includes 

provisions for the STM as a design procedure for all forms of D-regions. Table 1 shows the 

summary of stress limits and strength reduction factors used in Appendix A. Fig. 3 illustrates the 

various struts defined in the code. More detailed information about ACI Appendix A as well as the 

comparison of the code with AASHTO LRFD (1994) is provided elsewhere (Kuchma and Tjhin, 

2001). 

Struts: 

/,„ = 0.85P,/,' 

Table 1 Stress Limits and Strength Reduction Factors 

Stress Limits 

where: P, =1.00 for prismatic struts in uncracked compression zones (Fig. 3: (d) and (f)) 

(3^ = 0.40 for struts in tension members 

P, = 0.75 when struts may be bottle shaped and crack control reinforcement is included (Fig. 3: (a)) 

P s = 0.60 when struts may be bottle shaped and crack control reinforcement is not included (Fig. 3: (b)) 

P t = 0.60 for all other cases (Fig. 3: (c) and (e)) 

Notes: The angle between the struts and adjoining ties has to be greater than 25° 

Crack control reinforcement requirement is ^p w sin7, > 0.003, where p w is the steel ratio of the i-th 

layer of reinforcement crossing that strut, and y, is the angle between the axis of a strut and the bars. 

Nodes: f cu = 0.85(3 J L 

where: p, = 1.00 when nodes are bounded by struts and/or bearing areas 

P, = 0.80 when nodes anchor only one de 

P 5 = 0.60 when nodes anchor more than one tie 

326 

(d) 

fH| (e) (c) „,, (a) _, (b) 

4-i 

- 

^ 

— 

\Jv 

v 

^>. 

^~^~ 

x < 

\ 

N - \/ •^ 

s 

x 

\ 

Fig. 3 Types of struts in a typical D-region: (a) prismatic in uncracked field (b) prismatic in 

cracked field where struts are parallel to cracks (c) prismatic in cracked field where 

struts are not parallel to cracks (d) bottle-shaped with crack control reinforcement (e) 

bottle-shaped without crack control reinforcement (f) confined strut 

USE OF STRUT-AND-TIE MODELS IN SEISMIC DESIGN AND EVALUATION 

A strut-and-tie model is an idealization of the actual flow of forces in a D-region because it 

approximates the principal stress flow in the structure and is in equilibrium with the boundary 

forces. The required reinforcement in D-Regions can be calculated from the tie forces determined 

from a strut-and-tie model. This approach has been applied in seismic design and detailing. An 

example of this use is in the design of joints of cap beam and column in multiple column bridge 

bents (Sritharan et al. 2001). 

For the past decade, the performance-based design and evaluation has been an extensively explored 

topic in the earthquake engineering community. Conceptual guidance (e.g., Vision 2000 (1995)) as 

well as implementation (e.g., Priestley (2000)) of the methodology has been well documented. 

Additional considerations must be made when the STM is used in the framework of this 

performance-based methodology. A few of them are now described. 

Selection of Strut-and~Tie Models 

Strut-and-tie models are to be selected according to the selected design performance objectives. 

Structures are usually expected to perform elastically for performance objectives with no damage 

performance levels. Consequently, strut-and-tie models that follow elastic stress trajectories should 

be selected to avoid excessive stress redistribution. For performance objectives with damage state 

performance levels, it is often necessary to consider strut-and-tie models that deviate significantly 

from those suggested by elastic stress distributions. Due to limited ductility in structural concrete, 

however, the selected strut-and-tie models should be within workable limits. 

Multiple Load Conditions and Combinations 

Because the loading induced by earthquake is cyclic in nature, multiple load conditions, one for 

each direction as well as those from gravity loads, must be considered. The design load 

combinations are defined by superposition of relevant load conditions. Although superposition of 

strut-and-tie models associated with each load condition is allowed by the STM, strut-and-tie 

models should be established considering the design load combinations directly as a result of strain 

compatibility requirements.

327 

Strength Degradation 

The strength of struts is typically taken as a fraction of uniaxial concrete compressive strength 

obtained from compressive cylinder tests. This reduction in the usable concrete strength depends on 

many conditions; some of which were described in Fig. 3. In design considering reversed cyclic 

loading, two strut-and-tie models associated with each loading direction may share the same region. 

In addition, struts may change from struts to ties or visa-versa. This condition causes strength 

degradation in the struts and must be considered. 

Ductile Detailing 

Since earthquake loading is still poorly understood, capacity design concepts should be employed to 

ensure ductile behavior in all design performance objectives. In this regard, the capacity of the strutand-tie 

models should be governed by the capacity of ties. Also, the struts and nodes must still have 

sufficient strength under the expected strength of ties. 

Load-Deformation Prediction 

When load-deformation prediction needs to be made, basic requirements in structural mechanics 

must be satisfied in the strut-and-tie model for each loading step, namely equilibrium, strain 

compatibility relationships (kinematics), and stress-strain relationships. In addition, the geometry of 

the strut-and-tie model should be adjusted so that the internal energy of the system is minimum. 

Strain compatibility relationships should include components from struts, ties, as well as the nodes 

(Stojadinovic 1999). Stiffness degradation due to load reversal (Kinugasa and Nomura 2000) should 

also be considered. 

THE COMPUTER AIDED STRUT-AND-TIE (CAST) DESIGN TOOL 

Overview 

The simplicity and versatility of the STM can be hampered by the need to perform numerous 

calculations and geometric manipulations in order to complete a design. To overcome this 

encumbrance, various computer-based design tools have been developed to bring efficiency and 

transparency to the STM design process. A summary of the capabilities of a few computer-based 

STM design tools is discussed by Tjhin and Kuchma (2002). 

The authors have been developing the CAST (Computer Aided Strut-and-Tie) design tool that 

provides a single graphical environment (Fig. 4) in which the designer can sketch the D-region, 

draw the strut-and-tie model, solve for truss members forces, select tie reinforcement, define the 

dimensions of struts and nodes, adjust all design variables, as well as create a printout that 

summarizes the design. This design tool enables users to make on screen adjustments of all screen 

variables, to investigate design possibilities, and to optimize the design. In addition, the CAST tool 

enables multiple load cases and combinations to be considered. 

CAST has been under development since Fall 1998 and runs on the Windows 32-bit family 

operating systems. The current version of this program is freely available from

http-//www cee.uiuc edu/kuchma/strut&tie. There have been approximately 650 downloads made 

by people from outside of the University of Illinois, with feedback from users being used to make 

valuable changes and additions to this program. 

328 

Design Example Using CAST 

The three-story squat wall with openings described in the book by Paulay and Priestley (1992) is 

used to show how CAST can be used for seismic design. In this design, ACI code (ACI 318-02 

2002) is employed The design only considers the seismic effects in term of equivalent static lateral 

forces; the gravity load condition is not performed. Because the structure is not symmetric, two load 

conditions, one associated with the left lateral forces and the other associated with the right lateral 

forces, need to be considered. 

HCAST- 5«nt Watt with Openinis 

F8e £d* View Select £ombu* Define Assign Anal^n* £tsplaj> Qptiorw Window fcjelP 

(a*, a, -° ' Ji -* F — > ES

I 

329 

file £dit Vww Select Construct Define 6s»Qn Analasis Djsplay flptions Window /indow Help 

SI Unit -Dl^S^Jl 0_-° M 

: i «• u* t 

«; 

FJi 

> E 

^ El 

n 

lie 

General Info 

I ID i Bl ' 

Function | Strut and-Tie 

connectivity 

Start STM Node [ 

End STM Node \~ 

8 

1 

JLJXJ 

- Properties ——— 

Length | 315Q.Q mm Pfopertylype 

Orientation | 3SQQOQ deg Non Ptestressed flanforcement Tie J^j 

ReL Stiffness) 1 

Property Type Name 

TwForce j 2400 kN 3rd Flow Steel _^| 1 ShowDetaJ. |t 

, 

nTTiPKM—i ... . ,.. ,„., TrTi.r., NUTT r, , 

yp u t 

iy(>fe Propciiic* 

•T"^ - • ^strength —— - - 

& r 300 00 

Pa 

. . .." , Number of Bar Layc s j 4 ^j 

"Defined Types- 

-• Bat Layer Oat* 

| 

IIV 

NODE 

ifffibr. 

rn 4fi9\ 

ExS[RJt)S e teel ' " , Bar Oes^jnation 

Interior Steel dpi' | NumfagrofBars 

I — — — .j Nominal Diameter -111 mm 

813 

-^ Aiw-284mrrf 

I 300 mm 

Mote. Oak the 'Apply button to the tight after 

ng a bar layer 

Note. This tie type cannot be deleted ~ 

because tt has been assigned to STM g -~~ 

Element* 

summaiy 

. Total Steel Area | 2272.0 

Dos* Section 

Strength Reduction 

•**• j Q75 

~- » : Yield Force | " 5112 

kH 

| TensionZone Extension | 50 mm 

^_j' 

_ 

mrrf 

Use Default 

Controd 

lie j iusl M«tTieV/idth | 7191 

mm 

j 

• » Tip- In addition to using "Layer Number updown contol above to select 

a bar layer you can ako do it by clicking a bar laiyer from the image 

! tothelertorbyc£ckmgtheimageandthenpres: wig up or down 

u F __i ^•Tljrl - 

arrowkey 

JU |1 rr,r| [ OK | Caned 

Fig. 5 Results of truss analysis, dimensioning of struts and nodes, and selection of 

reinforcement for ties. 

The truss analysis for each load condition was then conducted. Afterwards, the reinforcement for 

each tie was selected, and the dimensioning of the struts and nodes were performed. Fig. 5 shows 

the results after the design was finalized. The top left display window partially shows the member 

forces and stress ratios associated with the left lateral load condition. The bottom left display 

window partially shows the member forces and stress ratios (i.e., the member stress intensities 

normalized to the corresponding stress limits) associated with the right lateral load condition. Note 

that colorized members were used to describe the level of stress ratios. The top right window is the 

"Show Element Info" dialog box displaying the summary of tie B-L This dialogue box was 

displayed by selecting tie B-I using the right mouse button. From that dialog box, the "Show 

Detail" button was pressed to reveal the "Define Non-Prestressed Reinforcement Tie Types" 

dialogue box (bottom right window) displaying reinforcing bar detail for Tie B-L 

SUMMARY 

Provisions for the design of D-(Discontinuity) Regions in structural concrete by the Strut-and-Tie 

Method (STM) have recently been incorporated into several international design specifications. 

The next step in the evolution of these provisions is to incorporate seismic design rules that address 

the strength degradation of struts and nodes, ductile detailing, and how to design structures for 

cyclic loading combinations. To aid designers in the use of the STM, computer-based design tools 

are under development, including the CAST design tool (www.cee.uiuc.edu/kuchma/strut&tie). 

ACKNOWLEDGMENTS

The development of CAST has been supported by a grant from the University of Illinois, a 

fellowship from Portland Cement Association, and a CAREER award from the National Science 

Foundation. 

REFERENCES 

American Association of State Highway and Transportation Officials (1994). AASHTO LRFD 

bridge specification. 1st ed., American Association of State Highway and Transportation Officials 

Washington, DC. 

American Concrete Institute Committee 318 (ACI 318) (2002). Building code requirements for 

structural concrete (ACI 318-02) and commentary (ACI 318R-02), American Concrete Institute, 

Farmington Hills, MI. 

Comite Euro-International du Beton (1993). CEB-FIP model code 1990, Thomas Telford Services, 

Ltd., London. 

CSA Committee A23.3 (1984). Design of concrete structures for buildings (CAN3-A23.3-M84), 

Canadian Standards Association, Rexdale, Canada. 

FIP Commission 3 (1998). FIP recommendation 1996, practical design of structural concrete, 

Federation Internationale de la Precontrainte, Lausanne, Austria. 

Kinugasa, H., and Nomura, S. (1996). Failure mechanism under reversed cyclic loading after 

flexural yielding (Paper No. 177). Proceeding of the llth World Conference on Earthquake 

Engineering. 

Kuchma, D. A., and Tjhin, T. N. (2002). Design of discontinuity regions in structural concrete using 

a computer-based strut-and-tie methodology," Proceedings of the 81st TRB Annual Meeting. 

Marti, P. (1985). Basic tools of reinforced concrete beam design. ACI Journal, Proceedings 82:1, 

45-56. 

MacGregor, 1 G. (2002). Derivation of strut-and-tie models for the 2002 ACI code. To be 

published in ACI Special Publication entitled "Examples for the design of structural concrete with 

strut-and-tie models." 

Paulay, T., and Priestley, M. J. N. (1992). Seismic design of reinforced concrete and masonry 

buildings, John Wiley & Sons, New York. 

Priestley, M. J. N. (2000). Performance-based seismic design (Paper No. 2831). Proceeding of the 

12th World Conference on Earthquake Engineering. 

Schlaich, J., Schafer, K., and Jennewein, M. (1987). Toward a consistent design of structural 

concrete. Journal of the Prestressed Concrete Institute 32:3, 74-150. 

Sritharan, S., Priestley, M. J. N., and Seible, F. (2001). Seismic design and experimental 

verification of concrete multiple column bridge bents. ACI Structural Journal 98:3, 335-346. 

Stojadinovic, B. (1999). Strut-and-tie models for cyclic loading. Informal group discussion of Sub- 

Committee 445-2. 

Tjhin, T. N., and Kuchma, D. A. (2002). Computer-based tools for design by the strut-and-tie 

method: advances and challenges. To be published in Sept.-Oct. issue of ACI Structural Journal 

Vision 2000 (1995). Performance-based seismic engineering of buildings. Vision 2000 Committee, 

Structural Engineers Association of California, Sacramento, CA. 

330




SEISMIC RESPONSE OF ARCH DAMS INCLUDING 

STRAIN-RATE EFFECTS 

Gao Lin and Shi-yun Xiao 

Department of Civil Engineering, Dalian University of Technology, 

Dalian 116024, China 

ABSTRACT 

Most of the experimental studies on concrete have shown that it is sensitive to the rate of loading. 

The material strength, stiffness and ductility are rate-dependent. When a dam is subjected to the 

earthquake excitation, the stress-strain relationship in different parts of the structure at different 

instant will be different due to different rate of strain it experiences. Based on the concept of 

viscoplastic consistency model, a modified three-parameter William-Warnke viscoplastic model is 

developed to take account of the rate-dependent behaviour of concrete. The seismic response of a 

27 8 m high arch dam is analyzed as an example. It is shown that the strain rate affects the 

displacement and stress distribution of the dam response, it plays an important role in the safety 

assessment of arch dams located in earthquake active area. 

INTRODUCTION 

Many high arch dams have been built and will be built in high seismic area in China. The highest of 

them reaches 300 m . The safety of these structures against earthquake shocks is of great concern. 

During the past two or three decades, our ability to analyze mathematical models of dam structures 

subjected to earthquake ground motions has improved dramatically. Sophisticated computer 

programs have been developed and used for numerical analysis of the arch dams. Nevertheless, the 

current design practice in the seismic design of arch dams is still based on the linear elastic 

assumption. The key property which determines the capacity of arch dams to withstand earthquakes 

is the tensile strength of concrete. However, the design criteria for tensile stress is still a problem at 

issue. A widely accepted standard is not available. As a consequence of this, there has not been a 

corresponding improvement in the earthquake resistant design. Therefore, improvement of the

332 

mathematical modeling of real structures has great significance. 

Concrete is a structural material sensitive to the rate of loading. Numerous tests have been carried 

out to investigate its response to rapid loading. Experimental results reveal that the material strength, 

stiffness and ductility (or brittleness) are rate-dependent. Hence, when a dam is subjected to the 

earthquake excitation, the stress-strain relationship in different parts of it at different instant will be 

different due to different strain rate it experiences. However, the conventional design practice 

accounts for rate sensitivity by means of drastic simplifying assumptions. That is, in all cases, the 

allowable stresses of an arch dam under earthquake load is increased by, say, 30% of the value 

specified for static case. Similarly, the dynamic modulus of elasticity is assigned 30% higher than 

its static value. The effect of dynamic behaviour of the dam and the effect of the waveform of the 

earthquake excitation have been disregarded. As a result, the true response of the dam may be 

altered to some extent. This paper aims to investigate the effect of strain-dependency on the seismic 

response of arch darn. In the literature several researchers (Lee et al 1998, Cervera et al 1996) 

studied the seismic response of a two dimensional concrete gravity dam by employing 

rate-dependent damage model. In the plastic-damage model of Lee et al (1998), the rate-dependent 

regularization is used to guarantee a unique converged solution for softening regions. No effect for 

rate-dependency on the stress distribution is involved. In the rate-dependent isotropic damage 

model, Cervera et al (1996) used a stress-strain curve that does not agree with the experimental 

results of majority of investigators (Bischoff et al 1991, Bazant et al 1982). In this paper, 

experiments on dynamic behaviour of concrete at high stain rates have been carried out and a 

consistency viscoplastic three-parameter William-Wamke model is developed to study the strain 

rate effects on the seismic response of arch dams. 

STRESS-STRAIN RELATIONSHIP AT HIGH STRAIN RATE 

Dynamic tensile and compressive tests of concrete at high strain rates have been carried out in our 

laboratory. The stress-strain curves obtained are shown in Fig.l and Fig.2. 

Strain (1Cf 8 ) 

Fig.l Typical stress-strain response of concrete 

in tension 

Strain (t(T 6 ) 

Fig.2 Typical stress-strain response of concrete 

in compression 

The strain rate enhancement in tension and in compression are given by

333 

(1) 

(2) 

Where f t and _/£ are the dynamic tensile and compressive strengths respectively; f ts and f cs are the 

static tensile and compressive strengths respectively; £ r and £ c are the tensile and compressive strain 

rates respectively;ând£ cs are the static strain rates in tension and in compression respectively 

(f tt = €„ = 10" 5 j' 1 ). Test results indicate that the dynamic modulus of elasticity increases with the 

strain rate. A slightly decrease of critical strain (strain at peak stress) in compression at high stain 

rates has been observed, but no visible change of critical stain in tension with the strain rate can be 

found. Some researchers indicated that the Poisson's ratio of concrete increases with the strain rate 

when subjected to tension and decreases with the strain rate when subjected to compression. In our 

experiments test results dispersed to some extent, it is difficult to correlate the dynamic Poisson's 

ratio with the stain rate. 

RATE-DEPENDENT CONSTITUTIVE MODEL OF CONCRETE 

Wang (1997) proposed a viscoplastic consistency model for analyzing metal, which can be seen as 

an extension of the classical elasto-plastic approach to account for rate dependency. The model can 

relatively, easily be implemented in place of classical rate-independent plasticity models (Winnicki 

et al 2001). In this model during viscoplastic flow, the actual stress state must remain on the yield 

surface. That is, for visvcoplastic loading, the yielding surface is expressed as 

/(cr y ,/r,K-) = 0 for A>0 (3) 

Where o;, is the component of stress tensor; K is the internal variable and A is the viscoplastic 

multiplier. The viscoplastic strain rate is defined as €"f - /bi y , and for the associated flow rule 

m tj = d//3cr y . Thus, the yield surface is rate dependent and can change its size and shape 

according to the value of the viscoplastic strain rate. 

The consistency condition is formulated as 

Assume that the rate of internal variable K is a linear function of viscoplastic multiplier of the form

334 

ft = Ag(0 tJ ), the consistency condition becomes 

Where 

m y

335 

JL dalt (10) 

Where 

Solution of the problem is accomplished by an implicit backward Euler integration scheme, where 

the Newton-Raphson iterative procedure is employed. The calculated stress-strain relationships for 

uniaxial compression and tension by this model are shown in Fig.3-6. The material properties used 

wereyc0=22.0MPa,y;0=1.45MPa, v=0.17 and E=1.9xl0 4 MPa. Comparison with experimental data is 

made and a fairly good agreement is achieved. 

-SOO 0 SCO 1000 1500 2000 2500 3000 3500 

Strain (10 8 ) 

Fig.3 Stress-strain relationship of the model for 

uniaxial tension 


uniaxial compression 

Experimental 

Numerical 

1 strain rate: 10 s /s 

2 strain rate. 10 3 /s 

1 strain rate 10 s /s 

2 strain rate. 10 a /s 

3 strain rate 10 T /s 

60 SO 100 12Q 140 160 

Strain (lO*) 

5QO 100Q 1500 2000 2500 3000 3500 

Strain (10"*) 


uniaxial tension with different stain rate 


uniaxial compression with different stain rate

336 

SEISMIC RESPONSE OF AN ARCH DAM 

In order to illustrate the effect of the rate dependency on the dynamic structural response, a 278 m 

high arch dam in China subjected to earthquake excitation is analyzed by the proposed model. The 

material properties for the study are as follows: for the dam body E=2.4xl0 4 MPa, v=0.17, 

p=2.4xl0 3 kg/m 3 , static compressive strength/ c =30MPa, and static tensile strength/=3MPa; for the 

foundation rock E^1.6xl0 4 MPa, v=0.25, p=2.0xl0 3 kg/m 3 . The dam and the foundation are 

discretized into 450 and 1040 isoparametric elements respectively. Fig.7 shows the discretized 

dam-foundation system. The five lowest vibration frequencies of the dam in care of full reservoir 

are://=0.997 Hz ,/r=1.0Q4 ^ /3=1.450 Hz ,f 4 =lA91 ^ and/5=1.542 Hz . An assumption of massless 

foundation is introduced to simplifying the dam-foundation interaction analysis though more 

rigorous interaction effects can be included. The design earthquake acceleration is 0.32 Ig. 

Fig.7 Geometry and mesh of arch dam. 

Fig. 8 Time history of Earthquake input 

Three-dimensional earthquake waves are used as the input. Fig.8 shows the typical artificial 

acceleragram that meets the requirement of (Chinese Specifications for Seismic Design of 

Hydraulic Structures }> . 

Three analyses were performed: an elastic analysis, a rate-independent plastic analysis and a rate 

dependent viscoplastic analysis. The maximum values of the first and the third principle stresses in 

the dam are shown in the Tab.l. It is seen that in all three cases, the maximum compressive stress 

are the same, the material remains working in the elastic range. However, for the maximum tensile 

stress there is marked difference between the calculated results of rate dependent model and that of 

rate independent model. 

Table.l 

MAXIMUM VALUES OF PRINCIPLE STRESSES 

Model 

Elastic 

Rate independent Plastic 

rate dependent Viscoplastic 

Compression(MPa) 

-12.5 

-12.5 

-12.5 

Tension(MPa) 

5.17 

2.87 

3.22

337 

Fig.9 Distribution of first principle stresses of an arch darn (upstream and downstream face) 

- rate independent plastic model 

'u;r 

Itf l' ' k ,-.-, ,' 1 V 

*£i£xV —-^.- .»-• 

,.-A'. I .,• ^j',V,'.« 

t .' 'i» -•' ,-;J 

-J'r.';-'-" J / ^, %S^-,'-Xc'./ 

'•"•-- ••• - —'--.„"$' &/ 

338 

rate-independent and rate dependent model, Fig. 12 shows that distribution of equivalent plastic 

strain rate calculated by the rate dependent model. 

CONCLUSIONS 

A modified three-parameter William-Wamke consistency viscoplastic model is proposed to 

investigate the strain-rate effects on the seismic response of arch dams. It is shown that the rate 

dependent analyses alter the stress distribution over the dam body and lead to an increased 

maximum tensile stress. These effects should be taken into consideration in the safety assessment of 

arch dams located in high seismic area. 

A CKNO WLEDGEMENT 

This research was founded by the National Science Foundation of China under grant No. 50139010 and No. 59739180. 

REFERENCES 

Bazant, Z.P. and Oh, B.H.(19982). Strain-rate effect in rapid triaxial loading of concrete, Journal of 

the Engineering Mechanics Division, ASCE, 108: EMS, 764-782 

Bischoff, P.H. and Perrry, S.H. (1991). Compressive behaviour of concrete at high strain rates, 

Materials and Structures, 24,425-450 

Cervera, M, Oliver, J. and Manzoli, 0. (1996). A rate-dependent isotropic damage model for the 

seismic analysis of concrete dams, Earthquake Engineering and Structural Dynamics, 25, 987-1010 

Chinese standard (2001). Specifications for Seismic Design of Hydraulic Structures, Chinese 

Electric Power Press, 17 

Harris, D.W., Mohorovic, C.E. and Dolen, T.P. (2000). Dynamic Properties of mass concrete 

obtained from dam cores, ACI MaterialJournal, 97, 290-296 

Japan Society of Civil Engineers (1996). Earthquake Resistant Design for Civil Engineering 

Structures in Japan, 16-18 

Lee, J. and Fenves, GL. (1998). A plastic-damage concrete model for earthquake analysis of dams, 

Earthquake Engineering and Structural Dynamics, 27, 937-956 

Wang W. (1997). Stationary and propagative instabilities in metals - a computational point of view, 

phD, TU Delft, Netherlands. 

Winnicld, A., Pearce, C.J. and Bicanie, N. (2001). Viscoplastic Hoffman consistency model for 

concrete, Computers and Structures, 79,7-19 

Xiao Shi-yun (2002). Rate-dependent constitutive model of concrete and its application to dynamic 

response of arch dams, ph.D dissertation, Dalian University of Technology, China.




NONLINEAR STATIC ANALYSIS METHOD BASED ON 

DISPLACEMENT 

Tu Wen-ge and Zou Yin-sheng 

Department of Civil Engineering, Hunan University, Changsha, China, 410082 

ABSTRACT 

In this paper, nonlinear static method based on displacement is presented. The earthquake-resistant 

performance of a building structure is represented as a sequence of functions of the maximum 

displacements that occur during strong seismic ground motions. The acceleration of ground motion 

being monotonously increased step by step, maximum displacements can be estimated using the 

concept of elastic displacement (or acceleration) spectra and be distributed using SRSS. These spectra, 

together with recently proposed natural periods for frame structures being into inelastic range, are then 

incorporated into a nonlinear static method based on displacement. Tow examples illustrate the 

proposed method. 

INTRODUCTION 

Traditional earthquake-resistant design methods emphasize that adequate strength, instead of 

displacement, is considered as the main design parameter. In the design procedure, the structural 

response of maximum base shear based on an elastic spectrum analysis is placed computation of and 

considered to be adequate; and then the member sizes are calculated via the distribution of this 

maximum base shear, based on elastic strength demand, along the height of the structure. Traditional 

methods do not explicitly consider the effects of ground motions' duration and the effects of hysteretic 

behavior of structural members, which will affect the overall force and deformation patterns on the 

structures. The above effects are implicitly considered by means of checking displacement limits (even 

extending to inelastic range) after sizing the structural members from the calculated elastic forces. 

In the case of large earthquakes, when yielding occurs in a structure and the structure is into inelastic 

range, the structural inter-storey drift increase more quickly than inter-storey shear, and structural 

members' deformation increase more quickly than members' force. Besides, it has been proved in

340 

many research works that adequate strength does not have a decisive influence on expected structural 

drift. Limit state can best be represented by deformation rather than by strength as damage that can be 

directly correlated with displacement. Therefore, maximum displacements, rather than maximum 

stresses, represent the proper design criteria. This differs from current force-based design that is based 

on acceleration spectra, code performance factors that correlate poorly with damage potential, and 

displacement checks to ensure that structural displacement limitations are not exceeded. 

In this paper, nonlinear static method based on displacement is presented. The earthquake-resistant 

performance of building structures is represented as a sequence of functions of the maximum 

displacements that occur during strong seismic ground motions. The acceleration of ground motion 

being monotonously increased by degree, maximum displacements can be estimated using the concept 

of elastic displacement (or acceleration) spectra and be distributed using SRSS. 

NONLINEAAR STATIC PROCEDURE 

There are four levels of structural analysis procedures appropriate for the evaluation of existing 

buildings and new design ones, as followed: 

(1) The Linear Static Procedure (LSP), which is a basic procedure, and mainly applicable to buildings 

whose respond is primarily in the elastic range; 

(2) The Linear Dynamic Procedure (LDP), which has such characteristics as LSP; 

(3) The Nonlinear Static Procedure (NSP), which can evaluate buildings loaded beyond the elastic 

range, but does not fully reflect dynamic response of structures, specially the effects of higher 

mode; 

(4) The Nonlinear Dynamic Procedure (NDP), which is the most complete form of analysis, 

modeling both dynamic effects and inelastic response, but which is sensitive to modeling and 

ground motion assumptions. 

The more advanced the form of analysis provides a more accurate model of the actual performance of 

a building subjected to earthquake loads, but greater the time and memory space are required. So, The 

Nonlinear Static Procedure (NSP) is a very practical analysis method, whose basic steps of are: 

(1) Assume the nonlinear force-displacement relationship of individual elements of structure 

(including yield strength, post yield stifmess and stiffness degradation, etc); 

(2) Calculate the target displacement of structure; 

(3) Select a reasonable lateral load pattern, and pushing the structure under this load pattern which is 

monotonously increasing step by step, when a structural member yields, then its stiffness is 

modified, until the roof displacement of structure is up to the target displacement or the structure 

collapses. 

At this time, the evaluation of seismic performance of structure is obtained. It is clear that the 

Nonlinear Static Procedure (Pushover) is a force-based analysis method which adds checking 

computation of displacement and deformation of structures by means of either base shearing force 

versus peak displacement or seismic demand spectrum versus structural capacity spectrum.

341 

DISPLACMENT-BASED NSF 

In this paper, a nonlinear static procedure based on displacement is presented. The method is a 

displacement-based analysis method that adds checking computation of displacement and deformation 

of structures. It maintains the simplicity of pushover analysis, and extends pushover analysis to 

investigate structural dynamic characteristic, and to account for development of displacement, and to 

describe ductility capacity of members set plastic hinge on. 

In this method, the acceleration of ground motion is monotonously increased step by step based on 

different seismic hazard levels. The lateral displacement, instead of base shearing force, is calculated. 

Moreover, the distribution of lateral displacement is selected by the mode shapes, which are 

considered based on structural dynamic modification considering the residual displacement of inelastic 

members, and post yield stiffness etc. 

In this method, modal analysis is adopted, because the displacement or force distribution can be 

expressed with the modal shapes, and the maximum displacement of each joint or floor of a structure 

subjected to the ground motion acceleration can also be expressed with the maximum coordinate using 

the displacement response spectrum. Furthermore, the higher modals are taken into consideration via 

the SRSS/CQC method. When the displacement is increased to some extend, the structure enters 

inelastic range, whose overall stif&iess is modified and whose dynamic characteristics is changed. 

Then, the structure is anew done modal analysis in order to obtain dynamic characteristics of the 

structure in inelastic range. 

Modal Analysis 

The structural motion differential equation is expressed as 

(1) 

Where M is the mass matrixes, C is the damping matrixes, K is the stiffness matrixes, X is the 

displacement vector, x w is the ground acceleration history. 

Calculate the eigenvalues and eigenvectors (natural frequencies and modal shapes) of the structure 

under undamping free vibration. Assume that X satisfies 

where O is the shape matrixes , q is the modal coordinate. 

Substitute this expression for X in equation (1): 

X=< (2)

342 

= -MIx g (3) 

Pre-multiplied by ® r , equation (3) becomes 

mq + cq + kq = -£, (4) 

Where 

r=$> J MI (5) 

(6) 

(7) 

(8) 

If it is a proportional or classical damping system, according to Duhamel integration, the modal 

coordinates can be calculated on the basis of equation (4). Then, the displacement can be calculated on 

the basis of equation (2). Hence, the structural members' force and deformations can also be obtained. 

Maximum displacement 

At the initial stage of structural design, a structural engineer should consider reasonable deformations 

under various earthquake intensities. Nevertheless, the prediction of important characteristic of input 

ground motion and the basic structural parameters are uncertain. Based on displacement spectrum, the 

maximum value of displacement is given. Because the inelastic displacement spectrum is not presented 

by Code for seismic design of buildings [2001], the displacement spectrum is transform from the 

acceleration spectrum, as followed: 

Oi^ (9) 

So, the maximum value of thej-f/z modal coordinates can be expressed by: 

W=aK>f>« 0°) 

When the periods and modes of a structure at previous step is known, according to modal analysis, 

both the relative-displacement of the structural members and the inter-storey drifts can be calculated 

using SRSS (Square Root of the Sum of the Square). And then force corresponding to them can also be 

obtained.

343 

Define the displacement of the i-th joint X,, using SRSS of N modes: 

(11) 

where 4> g is the z-f/z joint corresponding to j-th mode shape, &>j is the natural frequency 

corresponding to j-th mode, £j is the damping ratio corresponding to j-th mode, YJ is the mode 

participant parameter corresponding loj-th mode. 

When the maximum of acceleration of ground motion is increased step-by-step, formation of a desired 

earthquake-resistant mechanism comes into effect, and various modal parameters are modified. 

Therefore, the modal parameters and X,- are calculated anew. The dynamic performance of the 

structure in inelastic range is expressed via modal parameters and X, modified. 

Modify 

stiffness 

When formation of a desired earthquake-resistant mechanism comes into effect, one or some structural 

members enter inelastic range, whose stiffness is or are substituted post yield stiffness for. In this paper, 

the hysteretic model is bilinear one, no stiffness degradation. Furthermore, both residual displacement 

and P-Delta effect is taken into account as geometric nonlinearity. The i-th inter-storey drift, A I|9 is 

equal to sum both inter-storey lateral drift at yield, A y, and inter-storey lateral drift after yield, A p. 

ApA^ + A, (13) 

Determine the inter-storey yield 

For each structural member, its nonlinear stiffness model is a bilinear stiffness model. For each storey, 

its nonlinear storey shearing stiffness is also simplified to a bilinear stifmess model, that is, the 

nonlinear storey lateral force versus inter-storey lateral drift response of the prototype structure is 

represented by a bilinear approximation. The bilinear approximation has three defining characteristic: 

(a) The initial stiffness, Kj , (b) the maximum storey force and inter-storey drift, F u and A a , 

respectively and (c) the second (post yield) stifmess, £• 

The initial elastic stiffness is calculated from a linear-elastic analysis where the linear stiffness of the 

members. The maximum load and deformation at the limit state, F u and A u , respectively, in the 

bilinear approximation are equal to those of the nonlinear curve. The second stiffness, £2, is calculated 

by equating the strain energy in the nonlinear curve to that in the bilinear approximation, where the 

strain energy can be calculated by means of integrating the piece-wise-linear inter-storey force-drift 

curve. 

Meanwhile, The storey lateral force, F y , and niter-storey lateral drift, A 7 , at yield are defined by the

344 

intersection of the two stiffaess in the bilinear approximation of the pushover response curve of each 

storey. The relationship between them is shown by equation (12). 

Limit state 

The yield and limit-state deformations are used to defined the limit-state displacement ductility, // A : 

=-~ and //

345 

EXAMPLE 

The nonlinear static analysis method based on displacement is illustrated with two examples: Ex. A is a 

storey shear model. Ex. B is a structure that collapses from 7 th storey to 13 th storey in Tang Shan 

earthquake. 

EX.A 

Its structural parameters are shown in Table 1. The structure is in region of VII. The characteristic 

period value is 0.30 second. The nonlinear static analysis method based on displacement is carried out 

with 5 steps. Data at each step are shown in Table 2. The inter-storey drift at each step is shown in Fig 

1, and displacement of top floor versus the acceleration of ground motion in Fig 2. 

^\ 

1 st floor 

2 na floor 

3 ra floor 

4 m floor 

Mass 

(xlO 6^) 

7.8 

7.0 

6.8 

5.0 

Stiffiiess 

(xW*N/m) 

3.86314 

4.03524 

3.03524 

2.24514 

Table 1: structural parameters 

Post stiffness 

(xlQ 7 #/w) 

7.72628 

8.07048 

6.07048 

4.49028 

Inter-storey 

yield drift (m) 

0.0101 

0.0080 

0.0080 

0.0080 

Inter-storey 

limited drift (m) 

0.0910 

0.0720 

0.0720 

0.0720 

\^ 

1 st step 

2 nd step 

3 rd step 

4 th step 

5 m step 

Floor 

9 nd 

i st 

3 ra 

4 m 

^nd 

Table 2: Frequency and damage states at each step 

State 

Yield 

Yield 

Yield 

Yield 

Utmost 

Acceleration of 

ground motion( mis 2 } 

1.0065 

1.0889 

1.1266 

1.7323 

1.7311 

1 st 

2.6809 

1.7751 

1.3350 

1.2268 

1.1989 

Frequency (rad/sec) 

ond 

3 rd 

6.7741 10.1263 

6.7537 7.9907 

4.4852 7.1603 

3.4175 5.6160 

3.0295 4.5286 

4th 

13.0079 

11.1671 

11.1592 

9.0872 

5.8173 

EX.B 

The height of every floor of Ex. B is 4m except 4.5m at both 11 th and 12 th floor, and 3m at 13 th floor. 

The height from the base of structure is 52.0m. The span between columns is 6.0m. The cross-sections 

of columns decrease from 600 x 800mm(at 1 st floor) to 300 x 300mm (at 13 th floor). The cross-section 

of beams is 300 x 700mm. The grade of concrete is C20. The structure is in region of VIIL The 

characteristic period value is 0.40 second. The elevation of the frame structure is shown by Fig 3. The 

Giberson model is adopted in analysis, so yields of members occur only at the ends of elements. Fig 4 

shows the damage formation pattern of the structure. It is clear that members* deformation limitation

346 

occur at 7 th storey, which is consistent with the structure's collapse from 7 th storey to 13 th storey in 

Tang Shan earthquake. 

Fig l(Left): inter-storey drift at each step 

Fig 2(Riglit): curve of displacement of top floor versus the acceleration of ground motion 

Fig 3(Left): Elevation of frame structure 

Fig 4(Riglit): Distribution of damage (* represent plastic hinges, a represent limited state) 

SUMMARY 

Nonlinear static analysis method based on displacement is presented in this paper. Nonlinear action of 

structure is divided into some segments, in which nonlinear action is equivalent to linear action. For 

each segment, nonlinear respond of structure is made computation with effective elastic-linear 

parameters. Maximum displacements are calculated using displacement/ acceleration spectra and are 

distributed via SRSS at each increment of acceleration of ground motioa For two examples, the 

method is an available tool, and can account for structural nonlinear action.

347 

REFERENCE 

Code for seismic design of buildings, (GB 50011-2001), National standard of the people republic of 

China, Architecture Building Press of China 

T. Paulay and M. J. N, Priestley (1992). Seismic design of reinforced concrete and masonry buildings, 

John Wiley & Sons. Inc. 

Peter Fajfar (1996), "Simple Push-Over Analysis of Buildings Structures", 11 th World Conference on 

Earthquake Engineering, CD-ROM, paper No. 1011, Mexico. 

Mehdl Salidl (1981). "Simple Nonlinear Seismic Analysis of R/C Structures", Journal of Structures, 

ASCE, Vol.107, No.ST5,pp937-953. 

Ekwueme, M S (1999). "Determination of displacement limits for seismic rehabilitation of concrete 

buildings", The Structural Design of Tall Buildings, 8:79-115. 

Bommer, J J and Elnashai, A S (1999). "Displacement spectra for seismic design", Journal of 

Earthquake Engineering, 3(1): 1-32.




EDUCATIONAL AND CODE REQUIREMENTS 

FOR SEISMIC DESIGN IN HONG KONG 

Gordon Williams 

WSP Hong Kong Ltd., 

Hong Kong 

ABSTRACT 

In a number of infra-plate seismic regions of the world, authorities are considering the need to 

introduce seismic design into their structural codes where previously there was none. One of the major 

challenges for educators will be to ensure that the local practising structural engineers are well 

prepared for the more challenging conceptual and detailed design requirements of seismic design. 

This paper provides the author's thoughts on the educational requirements of designers who will need 

to be introduced to seismic design and also ideas on the type of code needed by designers in low to 

moderate risk seismic regions. 

The complexities of earthquake design for inexperienced designers in key areas such as, risk, loads, 

site factors, peak ground acceleration, response spectra, structural form and reinforcement detailing are 

discussed and some recommendations made. 

In the section on response spectra, a comparison of seismic loads using the latest spectrum response 

values from China, USA, and Europe indicate the significant variability with both previous versions of 

codes and each other. 

INTRODUCTION 

Designer's needs 

Structural designers are frequently required to cope with new or changing codes during their 

professional life. Although they are well able to understand the changes, the demands of time and 

economics usually mean that new codes take a long time to be accepted. Acceptance is improved when 

modified or new codes are logically presented, clearly explained and do not involve too many different 

factors or load cases to be considered for routine calculations. The requirement for clear unambiguous 

codes is even more necessary now than in the past since clients are increasingly demanding reduced 

design periods. 

Brief History of Seismic design in intra-plate areas 

Before the 1970's it was uncommon for building structures in intra-plate areas to be designed for 

earthquakes, however seismic design is already required in many urban areas and a number of cities

350 

including Hong Kong are considering introducing seismic design where previously there was none. 

Table 1 provides a brief summary of key dates when seismic design was officially introduced for some 

areas of low to moderate seismic risk. 

TABLE 1 

DATES SEISMIC DESIGN INTRODUCED 

Location 

Boston USA i 

Australia " 

Canada J 

China 4 

New York USA ' 

Date 

1975 

1979 

1985 

1990 

1995 

Comments 

After 7 magnitude earthquake in Meckering 1968 

Seismic provisions since 1953 

Official code 1974 but seismic design not required for intensity less than 7 

Classified zone 2 on map in 1982 

1. Nordensen(1989) 2.Bubb(1998) 3. Heidebrecht(1998) 4.Hu(1993) 

It is probable that seismic design was carried out for some government buildings prior to legislation 

being introduced. For example the author earned out the design for a new replacement school in South 

Africa at Tulbach near Cape Town incorporating seismic provisions in 1972 following an earthquake 

estimated to be 6.3 on the Richter scale on 29 th September 1969. Similarly hi Hong Kong there are 

some buildings which have incorporated seismic design and details even though there is no statutory or 

code requirement to do so. From the above it can be seen that earthquake design in intra-plate areas is 

relatively recent 

Existing seismic knowledge in Hong Kong 

In HK at present there are no requirements for buildings to be designed for seismic loads. Civil 

engineering structures, however, including bridges and railway stations have been routinely designed 

for seismic forces since at least 1983, (Civil Engineering Manual Chapter 4 1983). The Structures 

Design Manual for Highways and Railways (HKSDM) was published in 1993 (Highways Department 

of HK. 1993) and became more accessible to structural engineers designing buildings from May 1994 

when it was referenced to in PNAP 115 (Buildings Department HKSAR 1994). The nominal seismic 

load in the HKSDM is .05 times the mass, to which is applied as a static force with a dead load factor 

of 1.4 giving an equivalent to an ultimate acceleration of 0.07g. No consideration of structural form or 

soil conditions is required. 

In HK seismic design knowledge and awareness of structural engineers has increased through a 

number of initiatives by Government, Universities and others since tie early 1990's. This knowledge 

has also increased because there have been many designs carried out by HK structural engineers for 

developments in China where seismic design is required. In some ways this may not be an advantage 

since conceptually there are differences in the Chinese approach when compared to some of the 

seismic codes in other countries. Further Chinese design institutes frequently cany out the detailed 

design of a building so some of the detailed provisions are not so well known in HK. In 2002 the 

Buildings Department invited tenders to study the seismic effects on buildings in HK and so it is 

possible that a seismic code will be developed within the foreseeable future. 

The major seismic regions of the world such as California, Japan, China, Taiwan and New Zealand 

have long had seismic codes that are well established. Many of the HK local practicing structural 

engineers who have had their education in these countries understand seismic codes but have not

351 

necessarily had significant working experience using them. 

Structural engineers that have had seismic design experience in high seismic risk areas are likely to 

have less difficulty adapting to a new code in HK than those who have had no seismic experience. 

They may not, however, easily adapt their knowledge of design in regions of high seismicity to a 

region of low/moderate seismicity during the schematic design stage of a project This is because it is 

not necessary or appropriate to design structures in these regions for high levels of ductility. A further 

complication is that whereas HK, China and Europe use separate material factors for steel and concrete 

the US, Australia and New Zealand adopt strength reduction factors. This makes it more difficult to 

compare the codes. 

HK has some unique types of construction, especially when considering the dominance of residential 

high-rise construction. Different materials, types of construction and heights will mean that experience 

gained in other countries may not be readily transferred to HK, Concrete frames are predominately 

used for commercial buildings in New Zealand and Australia but their residential buildings are mainly 

low rise or separate family dwellings. In the US, although there are numerous concrete buildings, steel 

framing is mainly used for their commercial high-rise buildings. 

EDUCATIONAL REQUIREMENTS 

Educational Challenges for introduction of seismic codes 

One of the key challenges for educators in cities considering introducing seismic provisions in their 

codes is to ensure that the local practicing structural engineers are well prepared for the more 

challenging conceptual and detailed design requirements of seismic design. This is required so as to 

ensure that the economic, functional and aesthetic qualities of the built environment are not 

excessively compromised. 

It is at the schematic design stage that it is necessary to provide sound structural advice to the architect 

Inappropriate structural forms will result in additional cost or time implications to the project and 

could produce buildings that might experience significant structural failure during a severe seismic 

event. 

In order to determine how best to put across the key seismic design issues it is useful to consider the 

structural engineer's experience. Generally in HK there is a difference between the working 

experiences gained by structural engineers who design buildings as compared to civil engineers who 

design structures such as bridges, port facilities, dams, pumping facilities etc. Civil and geotechnical 

engineers in HK tend to be more familiar with design using more variable materials and more 

uncertain loads such as vehicle loads, impact, temperature, varying water tables etc. On the other hand 

structural engineers for buildings in HK mainly design for dead, live, wind and soil loads in 

accordance with rigid codes and even wind loads are still fairly basic and generally follow a simplified 

form of the method used in the outdated British code of practice CP3. 

Geotechnical engineers are more familiar with designing for consequence of failure in slope design. 

Post university training of structural engineers seldom require consideration of such topics. Structural 

engineers generally do not have to make decisions based on subjective criteria and often consider that 

their computer analysis and design models reflect an accuracy that hi reality may not necessarily exist 

This higher level of uncertainty of seismic design as compared to design without seismic 

considerations may adversely affect the ready comprehension of the requirements by practicing

352 

structural engineers. Although deflection calculations in reinforced concrete design only predict actual 

deflections within 25 to 30% this is usually not of concern to structural engineers since they only have 

the need to use simplified code methods of checking using BSS110 Part 1. In BS8110 Part 2 the more 

detailed methods of calculating deflection are not frequently referred to and are generally only used in 

automated structural computer programs. 

It is likely that the uncertainty, variability and perceived requirement to have a better than normal 

understanding of geotechnical engineering will be one of the hurdles preventing structural engineers to 

easily make the changes required to incorporate seismic loads into their structural design. 

Unfortunately there are a number of terms and different ways of presenting seismic risk that can be 

confusing for the non-specialist. It is important therefore when discussing the subject that all parties 

have a common understanding of the definitions adopted. 

The best way to train structural designers may be to play down the load variability and all of the 

geotechnical complexities and concentrate on the importance of detailing for ductility, providing well 

defined load paths and economical yet robust design. As Bubb (1998) said in a keynote address it 

should be realised that in building structures that "if there is a weakness, the earthquake will seek it out 

and even cause collapse and therefore loss of life. The earthquake will also find the weakness that the 

ignorance builds in." As found in Kobe, Taiwan, and Northridge, failures commonly occur at changes 

in strength or stiffness, because non structural walls affected the structural response or because the 

joints have not performed as well as expected. Bubb (1998) follows on by saying "It is the earthquake 

and the principles of mechanics that determines which are structural and which are non structural 

elements. Common sense and 'normal' practice would wrongly assume that the architect, the owner or 

the builder could determine which are non-structural elements". 

Seismic loads 

The main difference in seismic horizontal loads to wind loads is that for wind the load at right angles 

to the longer side is usually the critical load case. However for long narrow buildings seismic loads are 

dominant in the direction normal to the short side i.e. at right angles to the maximum wind load 

direction because the load is dependent on mass, which is the same in any direction. This means that 

the required relative seismic capacity of medium height and tall buildings in the long direction will 

need to be higher than that required for wind. However this may not control the design since concrete 

core buildings frequently have much greater capacity than required in the long direction because the 

reinforcement provided for the short direction controls the design. 

Static Equivalent Seismic Design Equation 

For concept design the lateral loads should be calculated from the static equivalent design equation 

Eqn. 1. but can be visualised most easily using a spectrum response curve see Fig. 4 and 5. 

7TC 

^~W (1) 

where V - Base shear, 2 = Zone Factor, I = Importance factor (Building use), 

C = Dynamic factor, R = Structural Characteristic, W = Weight of the building. 

Although the equation is simple it is somewhat difficult to understand in detail because the variability 

of the factors is greater than usually encountered by structural engineers.

353 

1) Zone factor / Seismology 

Generally structural engineers do not need much understanding of seismology other than when a time 

history analysis is required. Even topics such as attenuation, frequency, shear wave velocity, peak 

ground acceleration etc are concepts practitioners need to know little about provided they are given an 

easily defined structural spectrum from which they can carry out the design. 

The zone factor Z is the equivalent peak ground acceleration (pga) on rock for the location under 

consideration. In most codes the benchmark is the 10% probability that an earthquake of this size will 

occur in 50 years (equivalent to a return period of 475 years). The generally accepted range for the pga 

on rock in Hong Kong varies between O.OTg and 0.12g. This value tends towards a low level of 

earthquake risk rather than a medium level. 

Recently the Chinese map shows Kowloon and New Territories as 0.1 g and Hong Kong Island as 

0.15g. see Fig. 1 (Gao Mengtan 2002). A finer zoning would be more appropriate in order to reduce 

the difference between the two areas see Fig. 2 (Pun et al.2002). However "Drawing lines on maps, in 

social conduct, in religion, in law itself, has not proved one of humankind's greatest skills. Misery, 

expense and death are the usual outcomes," (Bill Mantle Canberra Times 15/8/98). It may be more 

pragmatic to adopt a single value for Hong Kong. In the authors opinion because of the many 

multiplying factors of this basic parameter it would be best to adopt a low value of say 0.8 for HK. 

Fig 1 (Gao Mengtan 2002) 

Illustration of seismic hazard in Hong Kong 

Fig 2 (Pun et al.2002) 

10% in 50-year pga bedrock contours 

2) Importance factor 

The Importance factor I can vary between 0.8 and 1.5. There may be a case to use this factor to 

provide a higher level of protection to properties on Hong Kong Island as compared to the New 

Territories for economic reasons rather than to use variable pga or Z factors in HK. 

3) Dynamic factor 

The Dynamic factor C is a function of the structural period and the frequency characteristics of the 

earthquake. The latter are highly dependent on the soils at the site; soft sites tend to amplify longer 

period motions, while on rock sites most of the energy is at short periods. C is therefore a function of 

the soil characteristics at the site, as well as the period of the structure being considered. C varies 

similar to the structural response of the building discussed under the building period & structural 

response spectra section.

354 

4) Structural characteristics 

R is a factor to account for ductility of the structure and varies between 4 and 8.5 for high ductility 

buildings and as low as 1 for low ductility buildings. 

5; Weight of Building 

W is the seismic weight of the structure. Engineers use force rather than mass units and so when 

accelerations are used care needs to be taken with the units to ensure the results are not incorrect by a 

factor of g (9.81 m/sec 2 ). Codes use different percentages of live load to be included for the building 

weight, which leads to some confusion. Structural engineers would prefer clear guidelines as to the 

appropriate factor. 

Base Shear 

In Hong Kong for earthquake design when the above variability is considered the value of V could be 

anything between .0014W and .225 W (a factor of 160). Fortunately very few buildings would be 

designed for the low value since they would tend to be over 100 metres in height where the wind loads 

are likely to be the dominant lateral load. In the Australian / New Zealand draft seismic loading code it 

is recommended that all buildings should be designed for a minimum horizontal load of .01W 

Building period & structural response spectra 

It is important that structural engineers understand the importance of the building period and the 

spectrum response when carrying out the seismic design of building structures. 

One concern is the accuracy of calculating natural frequencies. A plot of fundamental frequency versus 

height for 163 rectangular plan buildings is shown in Fig 3 (Ellis, B.R. 1980). If the actual measured 

frequency is checked against the theoretically calculated frequency the spread of frequency is even less 

accurate (Ellis B. R. 1995). 

E. 

Fig 3. Plot of fundamental frequency V, height for 163 

rectangular plan buildings. ISO 4866 'Evaluation and 

measurement of vibration in buildings' Annex A - 

'Predicting natural frequencies and damping of 

buildings.'(Ellis 1995) 

7= 

» 

I 

T2.5- 

O.Z5 0.5 r 0 

Frequency 

2.0 

Most structures are analysed assumed the frame to be skeletal made up of beams, columns and shear 

walls. In measurements of real buildings so called non-structural elements can contribute significantly 

to the stiffness of the structure resulting in a wide scatter of results. 

A comparison of acceleration spectrum responses for different codes normalised to O.OSg for rock and 

soil is shown in Figs, 4 and 6. From fig. 3 the rule of thumb calculation for a building height of say 50

355 

metres is about 1 hertz yet measurements on real buildings varied between 0.5 hertz and 2 hertz. If this 

variability is used to determine the base shear from the acceleration spectrum response curves in fig. 6 

the results can vary between 0.05g and 0.3g. This order of variability in loads exceeds by far that 

normally encountered by structural engineers. Fig. 6 shows that there is considerable variability in the 

different spectrum response curves in terms of acceleration up to one second. When the displacement 

curves are considered in Fig. 7 it is the 3 to 6 second period range that appear to be more variable. 

HONG KONG SEISMIC CODE REQUIREMENTS 

Analysis, P-delta, time history etc. 

Techniques such as push over analysis and non-linear analysis help to identify hinge locations but as 

yet these tools rely heavily on the accuracy of the structural modeling and the correct choice of many 

variables. Generally the requirement in codes to use time history is not particularly well understood by 

practicing engineers and appears to be more of academic than practical value. A significant amount of 

judgment is needed and codified guidance is seldom provided to enable practitioners to apply time 

history analysis confidently. It may be better to avoid these complex analysis methods for routine 

projects in HK. 

P-delta, accidental eccentricities, allowable overall and inter-storey drifts and building separation 

required to avoid pounding of adjacent buildings will need to be clearly explained in the HK code to 

avoid ambiguity or incorrect application by practitioners. 

Detailing rules and Ductility 

In all the seismic codes confinement reinforcement of potential hinge points is required and one of the 

most significant decisions to be made during the schematic design stage is the choice of which specific 

detailing rules should be adopted. If options are permitted in the HK code for high, moderate or no 

special detailing then it is likely, for the majority of buildings, to be more economical to adopt the least 

stringent detailing rules. However it may be better to allow only low levels of ductility detailing for 

routine designs. 

Large quantities of reinforcement could be saved if nominal ductility reinforcement could be 

minimized by not requiring it at all beam column junctions. For instance is it necessary to have 

ductility detailing for columns where small beams frame into large columns High-rise structures may 

only require ductility detailing of columns for the lowest three stories or so (much like shear walls) and 

when the column size reduces or the reinforcement changes. Further high-rise buildings will be 

controlled by stiffness due to wind loads rather than by seismic considerations and so a blanket 

ductility-detailing requirement would be wasteful of steel reinforcement. 

Particular care needs to be taken for items such as transfer floors, soft stories and other locations where 

high ductility demand can be expected. Providing additional capacity or higher ductility levels at these 

locations may be more economical than increasing the design loads or ductility of the building as a 

whole. 

Redistribution of moments is allowed by codes. It is probable that where redistribution of moments has 

been adopted the clarity of the design will be compromised and the standard of confinement 

reinforcement needs to be increased due to an expected higher ductility demand. Redistribution of 

moments should be maintained hi codes for elements such as slabs and secondary beams in order to

356 

maintain economy of design but may need to be prohibited for frames designed to resist seismic loads. 

Ductility of heavily reinforced sections 

It is not unusual in Hong Kong for columns to have nearly 6% reinforcement and even to use staggered 

laps and couplers to comply with the maximum 10 % allowable by BS 8110. If plastic hinges are not 

expected it may be possible to continue to allow heavily reinforced concrete members. 

Avoidance of shear failure of vertical members 

The capacity design requirement to provide sufficient shear reinforcement at ends of beams to ensure 

that the longitudinal reinforcement yields before shear failure occurs is a straightforward concept for 

beams. It is not as straightforward for compression members or where the gravity loads determine the 

design of the beams. Recent experience from the 1999 Athens earthquake indicates that even 

seismically designed columns failed, in shear, at mid height rather than at the top or bottom of the 

column. This will need further consideration when defining the nominal detailing rules especially since 

in HK column cross sections frequently are large compared to their length.. 

Development of HK Seismic Code 

The New Zealand (NZ) and Australian concrete codes closely follow the US code that uses strength 

reduction factors to reduce the ultimate design capacity, rather than material factors which are used in 

the Chinese, British and European standards. Since the US and NZ codes have been developed mainly 

for regions of high seismicity they may not be the most suited for development of a HK seismic code. 

One of the main difficulties in Hong Kong will be to extend the material design codes currently used, 

to allow seismic detailing. Neither the concrete code - BS 8110, nor the steel code - BS5950 has any 

guidance in this regard. Most countries separate the material design codes from the loading code. It 

will therefore, be a major decision as to what to do in Hong Kong where the material design codes 

have been adapted from British versions a number of years after their introduction in Britain. Even 

today HK uses the 1985 Version of the concrete design code even though the latest British version is 

1997. The British codes never had seismic design and in any case are being superseded by the Eurocodes 

so Hong Kong could follow the European seismic design code. However the Eurocode has 

tended to follow the rather complicated, yet pioneering seismic design codes developed in NZ, 

It is thus probable that using the Euro-code as a model to particularise for HK conditions will be the 

quickest way to produce a Hong Kong seismic code, however serious consideration should be given to 

adopting the Chinese code approach to seismic design. After all, HK is part of China and so it would 

be shortsighted to adopt a different system when first introducing a seismic code in HK. Whatever 

method is used, the main aim of the Hong Kong Seismic Code should be to keep it simple and make it 

easy to understand and use. 

CONCLUSION 

Before introducing a seismic code in HK, one of the challenges for educators will be to ensure that 

designers are confident in the determination of base shear and are familiar with the highly variable 

values derived. The key seismic code requirements for Hong Kong will be to provide clear guidance 

on the ductility factors and analysis methods to be adopted and that nominal detailing rules do not 

require high percentages of ductility reinforcement at all column beam joints.

357 

REFERENCES 

Booth E. D. A. J. Kappos & Park R. (1998) A Critical review of international practice on seismic 

design of reinforced concrete buildings. The Structural Engineer Vol. 76/No 11, June 1998 

Bubb Charles (1998) Earthquake Engineering in Australia - Before Mekering and after Newcastle, 

Australian Earthquake Engineering Society Conference, Perth 1998 Bulletin of the New Zealand 

Society for Earthquake Engineering Vol. 32, No 1, March 1999 

Buildings Department, Hong Kong Special Administrative Region (1994). Legislation and 

Publications Affecting the Building Industry PNAP 115 Practice Notes for Authorised Persons and 

Registered Structural Engineers 

Ellis, B. R. An assessment of the accuracy of predicting the fundamental natural frequencies of 

buildings and the implications concerning dynamic analysis of structures Proc. ICE, Pt2, Sept. 1980 

Ellis, B. RA feedback mechanism for improving analysis The Structural Engineer Vol. 73/No 21, 

November 1995 

Gao Mengtan New national seismic zoning map and seismic hazard in Hong Kong Region. One-day 

seminar on recent developments in Earthquake engineering. The Hong Kong Institution of Engineers 

and the Institution of Structural Engineers Joint Division May 2002. 

Heidebrecht Arthur C. (1998) Comments on Seismic Design for Regions of Moderate Seismicity & 

Case History on Effect of Seismic Hazard Level on Performance of Six Storey Moment Resisting Steel 

Frame Structures. International Workshop on Earthquake Engineering for Regions of Moderate 

Seismicity pp 51-77 

Hu Shipping (1993) Seismic Design of Buildings in China. Earthquake spectra Vol. 9 No 4 pp 703-737 

Nordensen Guy J.P. (1995). Built Value and Earthquake Risk. National Centre for Earthquake 

Engineering Research, 167-174 

Pun W K. Pang P L R, Wong A C W Recent Developments on the Understanding of the Seismic 

Hazard of Hong Kong. One-day seminar on recent developments in Earthquake engineering. The 

Hong Kong Institution of Engineers and the Institution of Structural Engineers Joint Division May 

2002.

358 

R 

C3&OC11 L8C34 USC3 ••• 

Fig. 4 - Elastic RSAS Using Different Codes, PGA=0.08g, Rock Site 

2 3 4 

Natural Panod of Structures. (s.| 

Fig. 5 - Elastic RSDs Using Different Codes, PGA=0.08g, Rock Site

359 

d Ptatod af Structure (t\ 

UBC37 • SurocadeS - 

Fig. 6 - Elastic RSAS Using Different Codes, PGA=0.08g, Soil Site 

Natural Penao of Structures fs) 

-GB-50011 

- IWC-SSC 3DCO 

Fig. 7 - Elastic RSDs Using Different Codes, PGA=0.08g, Soil Site




PERFORMANCE-BASED DESIGN OF 

GRAVITY RETAINING WALL IN SEISMIC AREAS 

X. Zeng 

Department of Civil Engineering, Case Western Reserve University 

Cleveland, Ohio, USA 

ABSTRACT 

Failure of gravity retaining walls occurred many times in past earthquakes. Typical failure mechanism 

was either an overturn or a large lateral deformation beyond repair. Traditional design approaches based 

on conventional limit equilibrium-based methods would emphasize on adopting a factor of safety against 

such failure. The goal was to design a structure that can resist forces generated by a seismic event. 

However, the cost of satisfying such requirement can be very high, especially for high intensity 

earthquake of rare occurrence. 

Performance-based design is a new methodology that emerged in the 1990s. The objective of this method 

is to avoid the limitations of conventional seismic design method. In a performance-based design method, 

a gravity retaining wall would be allowed to have an acceptable level of deformation that depends on the 

function of the structure. The critical element in this design method is to calculate the displacement of a 

gravity wall under a design earthquake. This paper reviews the methods that have been developed for 

calculating lateral, rotational, and coupled displacement of gravity retaining walls. It also discusses the 

role of comprehensive numerical codes in a performance-based design when simple analytical methods 

cannot produce satisfactory results. An example is used to illustrate the application of a performancebased 

design method, which in general leads to a more economical design. However, in order to apply this 

approach in the field with confidence, the time history of the design earthquake needs to be known. 

INTRODUCTION 

Many failures of gravity type of retaining walls have been reported in past earthquakes, for example, 

Tsuchida (1991) and Inagaki et al (1996). Most of the failures were caused by a large displacement of the 

structure in forms of lateral displacement, settlement, or rotation. For example, during the Kobe 

earthquake in 1995, caisson type quay walls at Port Island and Rokko Island had lateral displacement up 

to more than 5 meters, settlement up to more than 2 meters, and rotation of several degrees, Inagaki et al 

(1996). This kind of displacement was far above the acceptable level of damage allowed for this type of 

structure and was very expensive to repair.

362 

Traditionally, seismic resistant design of gravity retaining walls is based on a limit equilibrium-based 

approach, in which the dynamic forces induced by an earthquake are calculated. A factor of safety was 

then introduced to ensure that the retaining wall has sufficient resistance against the tendencies of sliding 

and rotating. However, in recent years this design philosophy went through a lot of debating and 

discussion because of its limitations. The main points of the discussion, as summarized by the 

International Navigation Association (2001), are: 1) in a typical design, deformations in ground and 

foundation soils and the corresponding structural deformation and stress states are key design parameters; 

2) conventional limit equilibrium-based methods are not well suited to evaluate these parameters; and 3) 

some residual deformation may be acceptable for the operation of most structures. 

To overcome these limitations, the concept of performance-based design was introduced (lai and Ichii, 

1998, Steedman, 1998). The main argument for this new design approach are: 1) if we demand that limit 

equilibrium not be exceeded in conventional design for the relatively high intensity ground motions 

associated with very rare seismic event, the construction/retrofitting cost will most likely be too high; and 

2) if force-balance design is based on a more frequent seismic event, then it is difficult to estimate the 

seismic performance of the structure when subjected to ground motions that are greater than those used in 

the design. 

Two of the most important steps in a performance-based design are to select the acceptable level of 

damage for the structure designed and to calculate the deformation of the structure under the design 

earthquake so as to check whether the deformation is acceptable. Some guidelines on how to choose the 

acceptable level of damage were proposed by the International Navigation Association (2001). Table 1.1 

is a summary of its recommendations. 

TABLE 1.1 

ACCEPTABLE LEVEL OF DAMAGE IN PERFORMANCE-BASED DESIGN 

Structural 

Minor or no damage 

Controlled damage 

Extensive damage in near 

collapse 

Complete loss of structure 

Level of damage 

Degree I: Serviceable 

Degree II: Repairable 

Degree El: Near collapse 

Degree IV: Collapse 

Operational 

Little or no loss of serviceability 

Short-term loss of serviceability 

Long-term or complete loss of 

serviceability 

Complete loss of serviceability 

Of course, for most structures, the purposes of performance-based design is to limit the level of damage to 

degree I or maybe degree U in some exceptional cases. The acceptable level of damage depends on the 

importance and the function of the structure. Once that is determined, the next step is to choose the design 

earthquake. Then, the critical part of the design is to analyze the performance of the structure being 

designed during the chosen design earthquake. For a gravity retaining wall, the performance is measured 

by the displacement of the wall under the design seismic loading.

363 

DISPLACEMENT CALCULATION FOR GRAVITY RETAINING WALL UNDER 

EARTHQUAKE LOADING 

Sliding Displacement 

Traditionally, sliding displacement of a gravity retaining wall is calculated using Newmark's (1965) 

sliding block method. Using the concept of a yield or threshold acceleration, relative displacement 

between a rigid wall and the ground beneath will accumulate from the instant the ground acceleration 

exceeds the threshold until the block and the ground have the same velocity again. For design purpose, 

Franklin and Chang (1977) analyzed a number of historic earthquake records to develop "standardized" 

sliding displacements as a function of the ratio of threshold to peak acceleration. 

The sliding block approach was modified and extended by Richards and Elms (1979) for the design of 

gravity retaining walls. A design envelope based on Franklin and Chang's standardized displacements 

was proposed to estimate the displacement of gravity walls, defined by 

where 

V 2 

d = 0.087—(N/A)" 4 (2.1) 

A 

d - displacement of a gravity wall in inches 

V = peak velocity of the design earthquake in inch/sec 

N = threshold acceleration for sliding in in/sec 2 , and 

A = peak lateral ground acceleration in in/sec 2 

Rotational Displacement 

For a gravity retaining wall, the displacement can also be caused by the rotation of the wall. A rotating 

block method was introduced by Zeng and Steedman (2000) to calculate such displacement. Similar to 

sliding block method, when the threshold acceleration for rotating is exceeded, a wall will start to rotate. 

Rotational displacement will accumulate until the angular velocity of rotation is dropped to zero. The 

important influential parameters on the angle of rotation are: peak acceleration, number of significant 

cycles, and the frequency of base motion. This approach was developed further to calculate coupled 

rotating and sliding displacement. 

Gravity Retaining Wall with Saturated Backfill 

Both the sliding and rotating block methods are used mainly for retaining walls with dry backfill. If the 

soil around the wall is saturated such as in the case of a quay wall, the calculation become too complex 

for these analyses to be used. Firstly, there are hydrodynamic pressures on both sides of the wall. The 

calculation of hydrodynamic pressure on the seaside is quite straightforward for a rigid wall by using the 

Westergaard's (1933) theory. However, the calculation of hydrodynamic pressure on the backfill side is

364 

complex as it depends on the soil-water-wall interaction. Secondly, excess pore pressure will build up m 

saturated soil. It is difficult to estimate excess pore pressure using simple calculation and hence it is not 

possible to use simple design calculation to determine the threshold acceleration or the resulting 

displacement. Therefore, small-scale model simulation or comprehensive numerical codes need to be 

used. 

Physical Modeling 

Since earthquakes in the field are unpredictable, it is difficult to record seismic response of gravity walls 

during earthquakes in the field. In most cases, what geotechnical engineers have are the pre-earthquake 

information and data collected during post-earthquake investigations. The most crucial information i.e. 

the response of the structure during earthquakes such as the amplification of vibration, excess pore 

pressure, and dynamic soil-structure interaction is generally not available. In order to calibrate or verify 

analytical procedures or numerical simulation, realistic physical data are needed. 

In the absence of field data recorded in earthquakes, data generated by physical model tests such as shake 

table tests and centrifuge tests become very valuable. In recent years, significant progress has been made 

in physical modeling technique in geotechnical earthquake engineering as reported in the 2001 NSF 

International Workshop on Earthquake Simulation in Geotechnical Engineering (Zeng, 2002). Shake table 

tests conduct at Ig have been used to study seismic response of gravity quay walls by Inagaki et al (1996) 

after the Kobe earthquake. While the interpretation of data from Ig shake table test is complicated by the 

scaling law issues, centrifuge modeling has the ability to create the same stress and strain condition as 

expected in the field. Since most mechanical properties of soils are stress dependent, centrifuge modeling 

has the advantage of satisfying most of the scaling laws. In the 1990s, a group of centrifuge tests of 

gravity retaining walls were conducted by Zeng (1998) for the VELACS project. The result of one model 

test is shown in Fig. 2.1. 

-• 

water 

' 

I 

! 

" -i* 

l°s L 44 

:, 

5 

1 

40 \ 

* :~" 

i 

! 

i\ • :i 

1- 96 ^| 

water saturated backfill 

Us 

~ initial boundary 

" boundary after EQl 

Fig, 2.1 Cross-sectional view of a gravity wall model after earthquake test (unit:m)

365 

As shown in the figure, a model test can simulate the type of deformation observed in the field such as 

sliding, rotation, and settlement. In addition, during a centrifuge test, the transducers used (such as 

accelerometers, pore pressure transducers, and LVDTs) can record in detail the dynamic response of the 

soil and the wall, excess pore pressure generation in soil, and deformation process of the structure. Such 

information is very helpful in the study of mechanism of response, in validating design calculations, and 

in calibrating or verifying numerical analysis. 

Numerical Modeling 

For complex design situations such as a gravity wall with saturated backfill, comprehensive numerical 

codes need to be used. In recent years, considerable progress has been made and several effective-stressbased-fully-coupled 

programs have been developed. One of the challenges numerical simulation is facing 

is how to verify the results of a numerical simulation. In the regards, physical modeling especially 

centrifuge modeling has an important role to play. For example, as part of the VELACS project, the 

response of the gravity wall shown in Fig. 2.1 during an earthquake was simulated using a computer 

program SWANDYNE (Chang, 1989). The results are shown in Fig. 2.2 (Madabhushi and Zeng, 1998) 

Fig, 2.2 Deformed finite element mesh after earthquake (unit: m) 

The deformation simulated by the finite element program is similar to that recorded in the centrifuge test. 

Thus, the code can be considered to have the ability to analyze complicated deformation of a gravity wall 

with saturated backfill under earthquake loading.

366 

AN EXAMPLE 

With the ability of calculating the displacement of a gravity retaining wall developed, the concept of 

performance-based design can be applied straightforwardly in the field. As an example, suppose we need 

to design a gravity retaining wall shown in Fig. 3.1 a) with the following parameters: friction angle of 

backfill

367 

history of a design earthquake is a modified version of a previous earthquake at the same site or a record 

of earthquake from a place with similar geological conditions. As an earthquake in the field never repeat 

itself even at the same site, wise engineenng judgment is needed in choosing such an design earthquake. 

CONCLUSIONS 

The recently emerged design approach of performance-based design can be applied to seismic resistant 

design of gravity type retaining walls. From the discussions and the example presented in this paper, the 

following conclusions can be made: 

1) Performance-based design emphasizes on limiting deformation of the structure rather than 

avoiding displacement completely as in the traditional limit-equilibrium type of design 

approach. 

2) The most crucial step in the performance-based design of gravity retaining wall in seismic area 

is to develop reliable procedures that can calculate sliding, rotating or coupled sliding and 

rotating displacements. Traditional sliding block method, rotating block method, or 

comprehensive numerical procedures can be used. Physical modeling techniques also have an 

important role. 

3) From the example presented in the paper, it seems that a performance-based design can lead to 

a more economical solution. However, the accurate application of such design concept requires 

the complete time history of the design earthquake be known, which can be difficult to obtain 

and at the same time bring in some uncertainties. 

References 

Chang, A.H.C. (1989). User manual for DIANA SWANDYNE-H, Dept. of Civil Engineering, University 

of Glasgow, Glasgow, U.K. 

Franklin, A.G. and Chang, F.K. (1977). Permanent displacements of earth embankments by Newmark 

sliding block analysis, Report 5, Miscellaneous Paper S-71-27, U.S. Army Corps of Engineers Waterway 

Experiment Station, Vicksburg, Mississippi. 

lai, S. and Ichii, K. (1998). Performance based design for port structures, Proceedings of UJNR 30 th Joint 

Meeting of United States-Japan Panel on Wind and Seismic Effects, Gaithersburg, NIST (3-5), 1-13. 

Inagaki, H., lai, S., Sugano, T., Yamazaki, H. and Inatomi, T. (1996). Performance of caisson type quay 

walls at Kobe port. Soils and Foundations, January, 119-136. 

International Navigation Association (2001). Seismic design guidelines for port structures, A.A. Balkema 

Publishers, Netherlands. 

Madabhushi, S.P.G. and Zeng, X. (1998). Seismic response of gravity quay walls. II: numerical modeling, 

Journal of Geotechnical and Geoenvironmental Engineering, ASCE, 124:5,418-427.

368 

Newmark, K.M. (1965). Effect of earthquakes on dams and embankments, Geotechnique, London, U.K. 

15:2,139-160. 

Richards, Jr., R. and Elms, D.G. (1979). Seismic behavior of gravity retaining walls, Journal of 

Geotechnical Engineering, ASCE, 105:4, 449-469. 

Steedman, R.S. (1998). Seismic design of retaining walls, Geotechnical Engineering, Proceedings of 

Institution of Civil Engineering, 131, 12-22. 

Tsuchida, H. (1991). Disasters caused by earthquakes in costal areas and countermeasures. Proceedings of 

the International Conference on Geotechnical Engineering for Costal Development, Volume 2, 918-926, 

Yokohama, Japan. 

Westergaard, H.M. (1933), Water pressure on dams during earthquakes, Transactions of ASCE 95, 418- 

472. 

Zeng, X. (1998). Seismic response of gravity quay walls. I: centrifuge modeling, Journal of Geotechnical 

and Geoenvironmental Engineering, ASCE, 124:5, 406-417. 

Zeng, X. and Steedman, R.S. (2000). Rotating block method for seismic displacement of gravity walls, 

Journal of Geotechnical and Geoenvironmental Engineering, ASCE, 126:8, 709-717. 

Zeng, X, (editor) (2002). Proceedings of NSF International Workshop on Earthquake Simulation in 

Geotechnical Engineering, Dept. of Civil Engineenng, Case Western Reserve University, Cleveland, 

Ohio, USA, http://ecivwww.cwru.edu/civil/xxzl6/proceeding/index.htm.




AN OPTIMIZATION PERFORMANCE-BASED 

EARTHQUAKE-RESISTANT DESIGN 

Zou Yin-sheng and Tu Wen-ge 

Department of Civil Engineering, Hunan University, Changsha, China, 410082 

ABSTRACT 

In this paper, an optimization procedure of design based on performance is presented. The 

performance-based design should consider two types of target performance levels: "generality" and 

"individuality". The latter is put sufficiently emphases upon to satisfy clients' demands for 

earthquake-resistant performances of structures in large earthquake. 

In order to satisfy "individual" performance levels, the investigation includes steps as followed: 1) 

determine target performance levels based on clients 1 various and particular demands, by means of 

Fuzzy Set Theory, 2) develop detailed criteria for the performance-based design of specific structure 

types and performance levels. The framework is demonstrated with a simple example involving a 

six-story RC frame. 

INTRODUCTION 

In recent years, the performance-based design of earthquake-resistant structures is emphasized and 

popularized. Based on the code for seismic design, the structural performance in inelastic range is paid 

greater attention to satisfied economic and social, and even clients' demands for structures in the large 

earthquake. Performance-based design results from and extends the traditional earthquake-resistant 

design, which inherits characteristics of traditional design and has some characteristics that the 

traditional design have not. 

To emphasize the structural performance in inelastic range is to satisfy clients* particular demands for a 

structure. It is important that the performance design criteria, which engineers determine, must be hi 

agreement with economic and social, as well as clients' demands for a structure subjected to large

370 

earthquake loads. In the first stage of performance-based design, it is a key step to clarify and simplify 

the target performance objectives from the views of protecting "human life", "properties contents" and 

"structure function", etc. These demands are generally described by language, need to be transformed 

into target performance objectives which structural engineers are familiar with. In the transformation 

step, Fuzzy Set Theory is an effective and available tool. 

In this paper, the characteristics of performance-based design are described. Because of clients' various 

and particular demands, "individual" performance levels is determined by means of Fuzzy Set Theory. 

Then, detailed criteria for the performance-based design of specific structure types and performance 

levels are developed. 

GENERALITY AKD INDIVIDUALITY 

Performance-based earthquake-resistant design, which is the modern approach, makes an attempt to 

predict structures with desired earthquake-resistant performances that are not only in agreement with 

economic and social demands, but also satisfy particular ones from "building users" and "building 

owners". Therefore, there should be two types of target performance levels, which reflect the 

"generality" and "individuality" of the design, respectively. 

"Generality" roots in the codes for earthquake-resistant design, and is described by means of a series 

of design parameters, empirical code formulations, limit states, and some disciplines required. Its 

objectives such as life safety and collapse prevention, both of which are hi agreement with the 

economic and social demands, are used to define the state of every structure subjected to earthquake 

load. Therefore, it is a target performance level that every earthquake-resistant structure must be 

equipped with. In a word, "generality" is the compulsory performance level that every 

earthquake-resistant structure must satisfy. 

For "generality", to protect human life safety is the most important, and primary target performance 

objective, which promises to produce every structure with desired mere earthquake-resistant 

performance that is described by designating the maximum allowable damage state for an identified 

seismic hazard. Hence, "generality" is the minimum performance level that every earthquake-resistant 

structure must satisfy. 

"Individuality" is a unique target performance level of an actual earthquake-resistant structure, which 

promise a structure to satisfy some particular demands from "building users" and "building owners", 

except for "generality". To emphasize "individuality" in earthquake-resistant design is to consider 

particular demands for a structure. Therefore, a performance objective may be chosen in according 

with damage states for several levels of seismic hazard, to assure not only "safety of human life", 

"prevention of structure collapse', but also "preservation of property" and "maintenance of structure 

function". Then, a dual or multiple-level performance objective would be adopted. It is clear that 

"individuality" extend target performance level based on "generality", and is enhanced performance. 

For an earthquake-resistant structure designed with performance-based design method, "individuality"

371 

reflects the structure's usefulness, and particular demands for "building users"' and "building owners"; 

"generality" reflects universal and compulsory demands for the society. It is clear that if "generality" is 

not satisfied, "individuality" will be not accepted. 

FUZZY PERFORMANCE CRITERIA 

In performance-based design method, "individuality" is paid more attention to. "Individuality" results 

from various aspects of earthquake-resistant performances demands from "building users" and 

"building owners". In accordance with clients' demands, engineers must determine design 

earthquake-resistant performances of each particular structure, and then establish technical design 

criteria and the design values corresponding to each performance level, as well as required structural 

strength levels to satisfy the technical design targets. It is necessary to clarify and identify what level 

of performance is required to meet each objective in the design stage. 

Classify clients' demands 

The ground motion uncertainty can be described by seismic hazard analysis. Consequently, to assure 

structural target performance levels, each design criteria must be determined in according with clients' 

demands for performances either in small earthquake (frequent occurrence), or in intermediate 

earthquake, or in large earthquake (infrequent occurrence). Then, to classify clients' demands will help 

us to make assessment to these demands. 

For each identified probability of earthquake occurrence, the following aspects need to be known from 

"building users" and "building owners": 

(1) Severity of hazard to human life (safety of human life) 

Select one of the following levels of injury: 1) no injury, 2) slight injury, 3) injury of fracture, 4) 

serious injury, and 5) death; which should be the minimum performance that the structure should 

provide. 

(2) Extent of loss of property value 

Select one of the following levels of loss: 1) no loss, 2) small loss, 3) middle loss, 4) large loss, and 5) 

almost total loss can not be avoided; which is the state in which the property value in structure is 

deemed to be preserved. 

(3) Required function of structure 

Select one of the following levels of function: 1) keep overall function, 2) keep main function and 

nonstructural damage can not be avoided, 3) keep essential function and partial repair is necessary, 4) 

keep limited function and can be used after repair, and 5) guarantee for life safety and it is necessary to

372 

rebuild; which is the state in which the structure is deemed as being preserved. 

As for "building users", generally speaking, both the "severity for hazard to human life" and the "loss 

in value of property" are given more attention than the other items, in which "no injure" and "slight 

injure", "no loss" and "small loss" are deemed as acceptable performances. 

As for "building owners", both the "severity of hazard to human life" and the "loss in value of 

property" are emphasized, as well as the "required function of the structure", in which "keep limited 

function and can be used after repair" and items under it is taken of as acceptable performances, is also 

paid greater attention to than "building users" 

Structural damage levels 

Structural engineers are familiar with structure damage levels, which are distinctly defined in codes for 

earthquake-resistant design in lots of countries. The structure damage levels are defined as followed: 

Select one of the following levels of damage: 1) no damage 2) slight damage, 3) small scale damage, 4) 

middle scale damage, and 5) serious damage and no guarantee for life safety, which is the damage 

state. 

In the level of serious damage, serious damages in structural members are not able to avoided, and 

partial collapse is probable to occur. Consequently, the hazardous damage to human life must be 

avoided. 

As for structural engineers, clients' demands about structural performance in complex range have to be 

clarified and simplified into key definitions and procedures of performance-based design, according to 

the principles in design Code, and then detailed criteria for the performance-based design of specific 

structural types and performance levels need to be defined. Clarifying and simplifying clients' 

demands, and then identifying design criteria are carried out by means of Fuzzy Set Theory. 

Fuzzy Set Theory 

The Fuzzy Set Theory, which was first introduced by Lotfi A. Zadeh in 1965, is considered as a means 

for quantifying the ambiguity and including it in an understandable logic. The most innovating aspect 

is the fuzzy reasoning hi which both variables and the relations between them are considered as fuzzy. 

The Fuzzy Set Theory can be considered as a complement to the theory of probabilities. There are, in 

fact two sources of uncertainties: randomness and fuzziness. The latter describes the vagueness, 

imprecision or ambiguity of the information. They are fundamentally different: a random or 

probabilistic phenomenon becomes certain after an event occur, whereas the ambiguity in the 

definition of a concept persists and is toe-insensitive, hi general, the imprecise or fuzzy information 

is replaced for which the membership is a binary notion (Le. yes/no or 0/1), by fuzzy sets, or for which 

the transition from membership to non-membership is smooth and not abrupt (i.e. sets whose elements

373 

have a membership degree between 0 and 1). 

A fuzzy set A of a universe X is characterized by a membership function ju 4 (x) which affects to 

each element x of Xa real number of the interval [0, I],ju 4 : X -> [0,1] It is usually represented as: 

A = & A (x)/x\x€X} (1) 

Where ju±(x) is the membership degree of* to A: the closest is this value to unity, the highest is the 

membership of x to A. The symbol "/" is used as a separator between the element x and its membership 

degree 

JJ. A (x) .If A is a. crisp set, this function can take only one of the two values 0 or 1 whether x 

belongs to A or not. The construction of the membership function of a parameter requires statistical 

data or the opinion of an expert who gives an estimation of the interval where the parameter can vary 

as well as its most likely value. 

The survey in both "building users" and "building owners" is carried out to know their demands. In the 

survey, informants are requested to answer to "hazard to human life", "loss of property value" and 

"required function of structure" according to various probabilities of earthquake occurrence. Its results 

are shown with the form of membership functions of fuzzy theory for damage level expressed by 

language. The derivation of the membership functions uses the number of the relevant option selected 

divided by the total number (ratio of each option selected). The evaluation points are evaluated as this 

membership value. The fuzzy set H is used to expressed severity degree of hazard to human life, 

whose membership function is defined as: 

(2) 

Based on the same method, the fuzzy set, L and F, about "loss of property value" and "required 

function of structure" can also be obtained. 

"Hazard to human life", "loss of property value" and "required function of structure" due to structural 

damage should not be independently and individually evaluated, but should be properly correlated. 

Hence, the damage levels corresponding to items studied about clients' demands are expressed by 

means of fuzzy reasoning method. 

In view of engineering, economic and social evaluation for earthquake resistant safety of structures, 

stressed degrees of three aspects about clients' demands are different. In this paper, the Cfc weight" of 

each aspect is used to measure its stressed degree. The sum of "weight" values is equal to 1. 

The studies about all information collected are carried out for using fuzzy reasoning method. These 

results are input into the three fuzzy relationships, the membership functions of the structural damage,

374 

D, can be obtained by evaluating each aspect using the following equation. 

D = (3) 

Where w(#), w(Z), and w(F) is the "weight" of fuzzy set "H", "I", and "P, respectively. 

Hence, The membership functions for comprehensive structural damage can be obtained by using the 

fuzzy reasoning method based on the membership functions of fuzzy sets about three aspects. Based 

on the membership functions, result can be simplified by applying the fuzzy maximizing decision 

[Bellman and Zadeh, 1970]. The structural damage levels in the ground motion hazard level, whose 

earthquake probability of occupancy 20% in 50 years and whose mean return period is 225 years, is 

demonstrated with Table 1, whose result is "slight injure". 

Table 1: Clients' demands transformation into damage leveis (performance levels) by Fuzzy Set Theory 

—^—-^ 

Hazard to human 

life 

Loss of property 

value 

Required function 

Damage levels 

Weight 

0.4 

0.3 

0.3 

No injure 

0.634 

No loss 

0.334 

Keep overall 

function 

0.301 

No damage 

0.4441 

Fuzzy membership 

Slight injure 

Injure of 

fracture 

Serious injure 

0.366 

0 

0 

Small loss 

0.613 

Keep main 

function 

0.585 

Slight 

damage 

0.5058 

Middle loss 

0.053 

Keep 

essential 

function 

0.114 

Small scale 

damage 

0.0501 

Large loss 

0 

Keep limited 

function 

0 

Middle scale 

damage 

0 

Death 

0 

Almost total 

loss 

0 

Guarantee for 

life safety 

0 

Serious 

damage 

0 

DETAILED DESIGN 

Clients lay stress on the performance of a structure in large earthquake. Moreover, structural engineers 

the capacity of a structure in inelastic range. In order to assure structural earthquake-resistant 

performance, the distribution of strength through a building is more important than the absolute value 

of the design base shear. A frame building would perform better under seismic attack, if it could be 

assured weak beam / strong column mechanism, that is, plastic hinges would occur in beams rather 

than in columns, or if the shear strength of members exceeded the shear corresponding to flexural 

strength. This can be identified as structural performance levels based on multiple design criteria, 

where the overall performance of the building is satisfied. Nevertheless, there are some factors that 

have a hold on the performance, for example the uncertainties of dynamic environment, uncertainties

375 

of initialization of structure, modeling uncertainties and so on. 

Based on different performance levels according to probability model of earthquake occurrence, the 

position of plastic hinges and the sequence of hinge formation are determined. One or some structural 

members come to be inelastic to dissipate seismic energy, which should be provided with sufficient 

ductility and not occur to spalling damage to resist large infrequent earthquakes. In the other hand, 

performance-protected members and others that are inappropriate for dissipation of energy should be 

always maintained in elastic response range. Consequently, the action on the columns and beams after 

yielding should be paid enough attention to. 

EXMAPLE 

Model building 

Function: Teaching building 

Construction site: Design characteristic period of ground motion: 0.40 second. 

Building type: Reinforced concrete frame (See Fig 1) 

Number of stories: 4 and 6 stories Story height: 3.0m (all stories) 

Number of spans: 3 spans Length between spans: 6.0m 

The cross-section of columns: 

450 x 500mm (side column at l~3 rd floor) 

500 x 500mm (middle column at l~3 rd floor) 

500 x 550mm (side column at 4-6* floor) 

550x 550mm (middle column at 4~6th floor) 

The grade of concrete: C35 

The cross-section of beams: 

Fig 1: Structure mode! 

I l l 

Fig 2: Plastic hinges formation 

Ex. A: 250x 600mm, 

Ex. B: 300x 500mm, 

Ex. C: 250x600mm(side beams) and 250x500mm(middle beams) 

Ex. D: 250x600mm(side beams) and 250x550mm(middle beams) 

The grade of concrete: C25. 

Performance objectives: Slight damage (earthquake probability of occupancy for the 20% in 50 years) 

Middle scale damage (the 10% in 50 years) 

Performance requirement: 

Beam / Column moment: flexural cracks to remain small and no significant spalling should occur 

Beam / Column shear: shear failure dose not occur and shear cracks are minor 

Mechanism: weak beams / strong column mechanism 

Analysis method: the static nonlinear analysis procedure using secant stiffness 

Analysis result 

Ex. A, B, C, and D is calculated and analyzed, whose position of plastic hinges is shown by Figure.2.

376 

Ab are shown by Figure 3-6, their capacity spectra are given, respectively In Table 2, their 

performance points and maximum value of base shear force and top floor displacement are given It is 

clear that Ex D is an optimum design, comparing with Ex A, B, and C 

SpcctraE DispPacemenf Spectral Displacement *" 

Fig 3 Capacity spectrum of Ex. A 

"~ "* SpeciraE Displacement 

Fig 4 Capacity spectrum of Ex B 

'Speclral Displacement " " 

"* w ~' 

Fig 5 Capacity spectrum of Ex. C 

Fig 6 Capacity spectrum of Ex. D 

Table 2 Performance points of examples 

~~ — "—--—, 

(V,D) 

(kN m) 

(Sa, Sd) 

(Vmax, Dmax) 

(kN m) 

Ex A 

(2177415 0283) 

(0170 0218) 

(2668 8237, 0 3779) 

Ex B 

(1935827,0322) 

(0155 0245) 

(22502539,04061) 

Ex C 

(2073727,0301) 

(0164 0230) 

(2696 4783, 0 4423) 

Ex D 

(2131635,0293) 

(0 168 0224) 

(2872 9229,0 4522) 

SUMMARY 

In this paper, the Ghent's performance demands for an earthquake-resistant structure in earthquake by 

language are clarified and simplified into damage levels (i e performance levels) by Fuzzy Set Theory 

Based on these damage levels (or performance levels), the structural quality in large earthquake load is 

inspected and modiiy if it is not satisfied, which assure that both clients' demands and economic and 

social ones for a structure in large earthquake load is satisfied

377 

REFERENCE 

Zadeh, L A (1965), "Fuzzy Sets", Information and Control, 8, pp 338-353 

Bellman, R E and Zadeh, L A (1970), "Decision-Making in a Fuzzy Environment" Management 

Sciences, Vol 17, No 4, B-141-B-164 

T Pauiay and M J N Priestley (1992) Seismic design of reinforced concrete and masonry buildings, 

John Wiley & Sons Inc 

Mehdl Salidl (1981) "Simple Nonlinear Seismic Analysis of R/C Structures 

ASCE Vcl 107, No ST5, pp937-953 

Journal of Structures

STRUCTURAL CONTROL




ADAPTIVE COLONY SYSTEM FOR STABILITY ANALYSIS 

OF SLOPES IN EARTHQUAKE ZONE 

Chang-Fu Chen 1 ' 2 and Xiao-Nan Gong 2 

1 Institute of Geotechnical, The Hunan University, 

Changsha, China 

2 Institute of Geotechnical, The Zhejiang University, 

Hongzhou, China 

ABSTRACT 

Earthquake is an important factor that induces the movements and failure of slopes. The stability of 

slopes in earthquake zone has received wide attention due to its practical importance. In this paper, the 

structure of ant colony system and the transfer probability of ants are modified to analyze the stability 

problem of complicated slopes. A new adaptive ant colony system (AACS) is proposed to determine 

critical slip surface and it is used to assess the influence of earthquake on the stability of slopes in 

earthquake zone. Two examples are presented to verify the proposed approach. 

INTRODUCTION 

Movements and failure of cuts and natural slopes constitute an important geotechnical problem 

(Leroueil, 2001). According to Schuster(1996), the economic losses associated with slope movements 

reach about US$4.5 billion per year in Japan, US$2.6 billion per year in Italy, on the order of US$2 

billion in the United States, and US$1.5 billion in India. China is probably the country that suffers the 

most from fatalities due to landslides. The number of landslide-related fatalities exceeds 100 per year 

(Li, 1989). Earthquake is an important factor that induces movements and failure of slopes. For 

example, the Haiyuan landslides in 1920 in China killed 100000 (possibly 200000) people. Therefore, 

the stability of slopes in earthquake zone has received wide attention due to its practical importance. 

Generally, limit equilibrium techniques (bishop, 1955; Morgenstern and Price, 1965; Spencer, 1967; 

Janbu, 1973; Sarma, 1979; Hoek, 1987) are commonly used to assess the stability of the slopes, as 

complex geological subsoil profiles, seepage, and external loads can be easily dealt with. The key 

problem in stability analysis of slopes is how to accurately determine the critical slip surface giving the 

minimum factor of safety F±. Many approaches have been developed to automate the search for the 

critical slip surface. The traditional mathematical optimization methods that have used include 

dynamic programming (Baker, 1980), conjugate-gradient (Arai and Togyo, 1985), random search 

(Siegel et al., 1981; Chen, 1992; Greco, 1996), and simplex optimization (Nguyen, 1985). The main

382 

shortcoming of these optimization techniques is the uncertainty as to the robustness of the algorithms 

to locate the global minimum factor of safety rather than the local minimum factor of safety for 

complicated and nonhomogeneous geological subsoil conditions (Goh, 1999). For this reason, Goh 

(1999) has presented a genetic-based evolution technique to search for the critical slip surface of slope. 

In this paper, an adaptive ant colony system is proposed to determine critical slip surface. It is used to 

assess the influence of earthquake on the stability of slopes in earthquake zone. Ant colony system, 

which is developed by imitating the colony behavior of ant colony, is mostly used to solve 

combination optimization problems at present. For solving the complicated slope-stability problem, its 

structure and the transfer probability of ant are ingeniously modified. Two examples are presented to 

verify the effectiveness of the proposed approach. 

ADAPTIVE ANT COLONY SYSTEM FOR SLOPE STABILITY ANALYSIS 

Recently, ant colony system (ACS; Dorigo et al, 1996) is often applied to solve the combination 

optimization problems, especially the traveling salesman problem (TSP; Dorigo and Gambardella, 

1997). For ACS being fit for solving the problem of slope stability, the ant colony system's structure 

and the computation method of the transfer probability of ant must be modified. For this purpose, a 

slope is discrete by the separating lines and nodes shown in Figure 1 (Chen and Xie, 2002). An ant 

starts from the point START, and then passes the nodes on the separating line one by one up to the 

point END. Thus, a circulation is finished and a slip surface formed. 

START , 

i 

//' 

V 

A 

\' 

\ 

ṛ 

'•> 

\ 

-+1 

!w;\ 

(r+ 

^ 

N&^ 

^- exit 

END 

* A 

"""/I 

/ /r- 

Km 

M tt JLfl 

r 

Figure 1 Search technique of critical slip surface of slope 

It is noticed that the dotted lines in Figure 1 are only used to guide the ants to search for slip surfaces 

and they are not considered in the computation of the slope safety factor. 

Suppose that (r, /) represents the fth node on the rth separating line; (r+1, y) represents theyth node on 

the (rfl)th separating line; [(r, i), (r+l,y)] represents the route from (r, i) to (r+1,;) and the number of 

ants in ant colony is m. During the movement of ants, ant k (fc=l, 2, —, m) decides its transfer direction 

according 

to the pheromone quantity on the routes. At moment r, the probability P^r0(r+l;) (0 of ant k 

transferring from (r, i) to (r+1,7) can be expressed as

.0,(r+l,,)] (0 + A (3) 

383 

-l.y)] ( f ) J 

(1) 

7=1,2, • 

where t [(rtl)t(r + ljn (t) is the intensity of the remained pheromone on the route [(r, /), (r-f l,y)] at moment 

r. At initial moment, the intensity of the pheromone on each route is equal, that is r [ (r , /), (r+i,;)] (0)=C (C 

is a constant). 7] [(/v)((rl . lt;)] (0 is the visibility of the pheromone on the route [(r, z), (r+lj)] at moment t. 

Here, we have 

(2) 

•l-s) 

where M r+ i is the number of nodes on the separating line (r+1). a and / respectively represent the 

index of the relative importance level of the intensity and the visibility of pheromone on the route of 

the ant choosing (a>0, /3>0). 

With time going on, the intensity of the remained pheromone on the routes ants have passed will 

gradually abate. Parameter p (0

384 

(Janbu, 1973 ) 

4 1 Y 1 \ C L cosa ' +^ ~ U J> cosc U tan( P ( 1 

4 

~ TX tana - 1 cos2a ,( 1 + tan( P, tana/F/) J 

where W, is the weight of the zth slice; a, is the angle of inclination at base of slice; /, is the length of 

the zth slice's base; M, is the pore water pressure; c £ ,

385 

The first example is come from Arai and Tagyo's paper (1985). A simple slope, shown in Fig. 2, consists 

of a homogeneous soil with zero pore pressure. The height and grade of slope are respectively 20m and 

1:1.5. The computation parameters of soil are cohesion c=41.65kPa, angle of internal friction 0=15°, unit 

weight Y = 18.82kN/m 3 . For the convenience of analysis, the slope is divided into 11 slices, each of which 

has the width 6m as illustrated in Fig.2, and the spacing of nodes on each slice is 0.5m. 

Figure 2 The computing results of Example 1 

Based on the simplified Janbu's analysis method of slope stability and the present search technique of slip 

surface, the location of the most dangerous slip surface is illustrated as the dash double-dot line in Fig.2, 

and the factor of safety is 1.268. For explaining the accuracy of the present method, the Arai and Tagyo's 

result obtained by the use of conjugate gradient technique is depicted as the solid line in Fig.2. The factor 

of safety is 1.265. The Fig.2 shows that the results obtained by the present AACS are very consistent with 

those given by the Arai and Tagyo's conjugate gradient search technique. This demonstrates that the 

AACS is effective and reliable for stability analysis of slopes. 

Example 2 

An open-pit mine is located in gobi in Xinjiang Autonomous Region. In this mine zone, the mean annual 

rainfall is 33.4mm, but the quantity of evaporation is as high as 3222.2mm. The difference of atmospheric 

temperature between the day and night reaches up to 40°C, and the most difference during all year long is 

about 76°C. The design height of slope at the northern part of the mine is 98m- 120m, and the inclination 

angle of slopes is 44°C. The open-pit slope consists of crystalline lime-rock, marble, andesite, and iron ore. 

The physical and mechanical parameters of rock mass, shown in table 1, are obtained by the laboratory test 

and in-situ investigation. The influence factor of earthquake is 0.08. 

TABLE 1 

PHYSICAL AND MECHANICAL PARAMETERS OF ROCK MASS 

Material 

Crystalline lime-rock 

Marble 

Unit weight, 

y 

28.0kPa 

28.1 kPa 

Internal 

friction, 

9 

36.6° 

43.8° 

Cohesion, 

c 

162.0 kPa 

213.0 kPa 

Influence Factor of 

Earthquake, K c 

0.08 

0.08

386 

Andesite 

Iron ore 

27.4 kPa 

31.0 kPa 

39.2° 

46.2° 

187.0 kPa 

318.0kPa 

0.08 

0.08 

By the use of the proposed AACS, the five slopes in above open-pit mine are analyzed by assuming m 

=100, a=1.0, /J =1.0, p =0.6, p Q =Q.%, ^=200 and the computation results are shown in Table 2. 

TABLE2 

THE COMPUTATION RESULTS OF EXAMPLE 2 

Slope number 

N7 

Nil 

N15 

N19 

N23 

Height of 

slope, H 

120.0m 

118.0m 

115.0m 

117.0m 

116.0m 

Dip angle of 

slope, a 

44.0° 

44.0° 

44.0° 

44.0° 

44.0° 

Safety factor with 

no earthquake, F s 

1.4910 

1,5098 

1.5405 

1.5104 

1.5217 

Safety factor with 

earthquake, F^ 

1.2846 

1.2932 

1.3143 

1.2979 

1.3122 

Relative 

deviation 

13.84% 

14.35% 

14.68% 

14.07% 

13.77% 

Table 2 shows that the safety factors of slopes in earthquake zone decrease by 13.77 to 14.68 percent as 

compared to the safety factors of slopes in no earthquake condition. The average decrease of safety factor 

is 14.14%. 

CONCLUSIONS 

The stability analysis of complicated slope is actually a nonlinear programming problem with multiple 

extreme values and the objective function usually is not explicit. The traditional optimization methods are 

difficulty to find out the most dangerous slip surface. For the ant colony algorithm being fit for evaluating 

the influence of earthquake on the stability of slopes in earthquake zone, the structure of basic ant colony 

system (ACS) is modified. A heuristic operator is introduced to determine the transfer probability of ant 

and an adaptive operator is presented to avoid the phenomenon of stagnation in evolution process and then 

a new adaptive ant colony system (AACS) is proposed. 

Two examples are given to verify the adaptive ant colony system. Example 1 is used to verify the accuracy 

of the proposed method The computation results show that this method is feasible and reliable and it can 

also be used to evaluate various slopes with complicated soil layers and complicated profiles. Example 2 is 

used to assess the influence of earthquake on the stability of slopes in earthquake zone. The results show 

that the average decrease of safety factor of slopes under earthquake is 14.14% in the above open-pit. 


This work is partly supported by the Fund of Science and Technology from Hunan province (Grant No 

X99P51059).

387 

REFERENCES 

Leroueil, S. (2001). Natural slopes and cuts: movement and failure mechanisms. Geotechniqu 51:3, 

197-243. 

Schuster, R. L. (1996). Socioeconomic significance of landslides. In Landslides: Investigation and 

Mitigation, Special Report 247, Washington: Transportation Research Board, pp. 12-35. 

Li, T. (1989). Landslides: extent and economic significance in China. Proc. 28th Int. Geol Cong.: 

Symp. Landslides, Washington, 271-287, 

Bishop, A. W. (1955). The use of the slip circle in the stability analysis of slopes. Geotechniqu 5:1, 7- 

17, 

Morgenstern, N. R. and Price, V. E. (1965). The analysis of stability of general slip surface. 

Geotechniqu 15:1, 79-93. 

Spencer, E. (1967). A method of analysis of the stability of embankments assuming parallel interslice 

forces. Geotechniqu 17:1, 11-26. 

Janbu, N. (1973). Slope stability computations. In Embankment dam engineering. Edited by R.C. 

Hirschfield and J. Poulos. John Wiley and Sons, New York, pp. 47-86. 

Sarma, S. K. (1979). Stability analysis of embankments and slopes. Journal of the Geotechnical 

Engineering Division, ASCE 105:12, 1511-1524. 

Hoek, E, (1987). General two-dimensional slope stability analysis. In Analytical and computational 

methods in engineering rock mechanics. Edited by E.T. brown. Allen and Unwin, London, pp. 95-128. 

Baker, R. (1980). Determination of the critical slip surface in slope stability computations. 

International Journal of Numerical and Analytical Methods in Geomechanics 4, 333-359. 

Arai K, and Tagyo K. (1985). Determination of noncircular slip surface giving the minimum factor of 

safety in slope stability analysis. Soils and Foundations 25:1, 43-51. 

Siegel, R. A., Kovacs, W. D., and Lovell, C. W. (1981). Random surface generation instability 

analysis. Journal of the Geotechnical Engineering Division, ASCE 107:7, 996-1002. 

Chen, Z.-Y. (1992). Random trials used in determining global minimum factors of safety of slopes. 

Canadian Geotechnical Journal 20, 225-233. 

Greco, V. R. (1996). Efficient Monte Carlo technique for locating critical slip surface. Journal of the 

Geotechnical Engineering, ASCE 122:7, 517-525. 

Nguyen, V. U. (1985). Determination of critical slope failure surfaces. Journal of the Geotechnical 

Engineering, ASCE 111:2, 238-250. 

Goh, A. T. C. (1999). Genetic algorithm search for critical slip surface in multiple-wedge stability 

analysis, Canadian Geotechnical Journal 36, 382-391.

388 

Dongo, M. et al. (1996). Ant system: optimization by a colony of cooperation agents. IEEE Trans, on 

SMC-Part £26:1,29-41. 

Don go, M. and Gambardella, L. M. (1997). Ant colony system: a cooperative learning approach to the 

travelling salesman problem. IEEE Trans, on Evolutionary Computation 1:1, 53-56. 

Chen, C.-F. and Xie, X.-B. (2002). Ant colony algorithm used to search for critical slip surface of 

open-pit slopes. Journal ofXiangtan Mining Institute 17:1, 62-64.




DYNAMIC ANALYSIS OF FRICTION-DAMPED STRUCTURES 

L. L. Chung 1 , L Y. Wu 2 , and Y. P. Wang 3 

1 National Center for Research on Earthquake Engineering, Taiwan 

Department of Civil Engineering, National Taiwan University, Taiwan 

Department of Civil Engineering, National Chiao-Tung University, Taiwan 

ABSTRACT 

In this paper, the feasibility of friction damper as a passive control devices is verified through 

theoretical and numerical investigation. The stiffness and damping may be enhanced simultaneously 

by adding friction dampers to the conventional structures. Significant reduction in dynamic response 

can be achieved. After the friction dampers are added to the structure, friction mechanism makes the 

structural system highly non-linear and the motion of the friction damper may be shifted from the nonsliding 

mode to the sliding mode, and vise versa. No matter in the non-sliding or sliding mode, either 

the kinematic or kinetic condition is known. A simple and efficient numerical method is developed for 

the dynamic analysis of friction-damped structures. Only a single equation of motion is used to 

describe the dynamic system for both non-sliding and sliding modes. The stability and accuracy are 

guaranteed even though the time step of integration remains constant during the process of the 

solution. 

INTRODUCTION 

Friction damper is one of the promising passive control devices for civil engineering structures. The 

stiffness and damping may be enhanced simultaneously by adding friction dampers to the conventional 

structures. With appropriate allocation of friction dampers, vibration energy will be dissipated 

efficiently when the structure is subjected to environment loads. The research on slotted bolted 

connection for friction damper can be traced back to 1976 (Venuti 1976). The friction damper consists 

of two outer plates and one inner plate. Circular holes are drilled on the outer plates and slotted holes 

are drilled on the inner plate for the bolted connection so that relative sliding between the inner and 

outer plates is allowed. Energy is dissipated through friction mechanism (Pall and Marsh 1979; 

Fitzgerald et al 1989; Roik et al 1988). In 1982, X-shaped bracing system was used to mount the 

friction damper (Pall and Marsh 1982). The stiffness is induced from the tension of the bracing. At 

the connection of the X-shaped bracing, energy is dissipated by the friction mechanism of the link and 

braking pad (Filatrault and Cherry 1987; Pekau and Guimond 1991). Such kind of friction damper has 

been put into real application (Pall et al. 1993).

390 

After the friction dampers are added to the structure, friction mechanism makes the structural system 

hishly non-linear. When the structure is driven with external loads, the motion of the friction damper 

may be shifted from the non-sliding mode to the sliding mode, and vise versa. In the non-sliding 

mode, the relative displacement between the interfaces of the damper remains unchanged but the shear 

force between the interfaces is unknown. In the sliding mode, the relative displacement is unknown 

but the shear force is the maximum friction force. No matter in the non-sliding or sliding mode, either 

the kinematic or kinetic condition is known. In this paper, a simple and efficient numerical method is 

developed for the dynamic analysis of friction-damped structures. Only a single equation of motion is 

used to describe the dynamic system for both non-sliding and sliding modes. The stability and 

accuracy are guaranteed even though the time step of integration remains constant during the process 

of the solution. Finally, the effectiveness of the friction dampers is studied. 

MATHEMATICAL DERIVATION 

When a discrete-parameter structural system with n degrees of freedom is subjected to environmental 

loads w(r) and counteracted by a friction force w(r), its governing equation can be taken as: 

Mx(r) + Cx(r) + Kx(r) = bu(0 + Ew(0 (1) 

where x(f) is the displacement vector; M, C and K are the mass matrix, damping matrix and 

stiffness matrix, respectively; b is the location vector of the friction damper; E is the location matrix 

of the environmental loads. Represented in state-space form, the second-order differential equation (1) 

is changed to the first-order differential equation as: 

where z(r) = ., I, A =, . ,, Ur =, 

I x(r) -M- 1 K -M- l C f c M' l b 

and E c = | ,.__!„ I. Since the recorded 

load functions are commonly discretized and the friction force are piece-wise linear in nature, it is 

logical to assume linear variations of these loading functions between two consecutive sampling 

instants. Therefore, With the sampling period &* , the solution of the state equation (2) becomes a 

difference equation as: 

z[k + 1] = Az[*] + b Q it[k] + bû[k + 1] + E 0 w[fc] + E! w[& + 1] (3) 

where A = e AA

391 

maximum friction force and there is no change in the relative displacement between the interfaces. 

The non-sliding conditions are: 

(4a) 

(4b) 

where /j. is the faction coefficient; N is the normal force applied to the interfaces; s[k -f 1] and s[k] 

are the relative displacement between interfaces at the current and the next time steps. In the nonsliding 

phase, the shear force is unknown but the relative displacement remains unchanged. During the 

sliding phase, the shear force is large enough to make the damper slide and there is change in the 

relative displacement. The sliding conditions are: 

(5a) 

(5b) 

In the sliding phase, the shear force is known while the relative displacement is unknown. Therefore, 

no matter the damper slides or not, either the shear force or the relative displacement between the 

interfaces is known. With this additional condition, the shear force at the next time step can be 

determined, and the response of the system can in turn be solved. 

First, the friction damper is assumed in the non-sliding phase, so that the relative displacement of the 

interfaces remains unchanged: 

s[k + l] = y b [k+l]-y(k + l]~y b [k]-y[k] = S[k] (6) 

where y b [k] is the relative displacement of the bracing system by which the friction damper is 

mounted; y[k] is the relative displacement between the two stories where the friction damper is 

located. The story drift y[k] is a linear combination of the state vector z[&] as: 

y[*] = Dz[*] (7) 

where D = [-& T 0 T J is the output row vector; T denotes the transpose of vector or matrix. The shear 

force of the damper and the relative displacement of the bracing system are related as: 

u[k] = k b y,[k] (8) 

where & b is the stiffness of the bracing system. After pre-multiplying the stiffness & b to equation (6), 

it gives: 

Utilizing equations (7) and (8), the estimated shear force at the next time step is expressed as: 

u[k + 1] = Jt b Dz[* + 1] + u[k] - * b Dz[*] (10)

392 

After substituting equation (3) into the above equation, it gives: 

u[k +1] = Jfc b D( Az[&] + b Q u[k] + b&k +1] + E 0 w[£] + Ej w[fc +1]) + w[&] - & b Dz[&] (11) 

After rearrangement, the shear force at the next time step can be estimated as: 

u[k -f 1] = Pz[k] + Qu[&] + R 0 w[fc] + R t w[£ +1] (12) 

where P = (l-* b Db i r l fc b D(A-I); Q = (l~£ b Db i r l (£ b Db 0 +1); R 0 =(l-*J>bO -I U>E 0 and 

There are two possible cases for the estimated value of the shear force at the next time step u[k +1]: 

When the estimated shear force is less than the maximum friction force, the interfaces stick together 

and the assumption stated in equation (6) is satisfied. The shear force at the next time step is the same 

as the estimated one: 

(2) \u[k 4-1]| >it [Wi ~/tN 

When the estimated force is greater than the maximum friction force, the friction between the 

interfaces is overcome and the interfaces slide relative each other. The assumption stated in equation 

(6) is violated. Therefore the magnitude of the shear force is the same as the maximum friction force 

and the direction of the shear force is the same as the estimated one: 

«[* +1] = «m« sgn02[* +1]) = /.iNsgn(u(k +1]) (14) 

where sgn(») = •/[«! is the signum function. The procedures for the dynamic analysis of the frictiondamped 

structure are shown in the flow chart (Figure 1). 

NUMERICAL VERIFICATION 

In order to illustrate the effectiveness of friction damper, a single-degree-of-freedom structure, with 

mass m = 100 ton, stiffness k = 4000 kN/m, and damping coefficient c = 40 kN/(m/sec), is investigated. 

Thus, the natural period and damping ratio of the structure are T = 0.993 sec and £=3.16 %, 

respectively. The top of the friction damper is mounted underneath the beam while the bottom of the 

damper is connected to the inverted V-shaped bracing through the T-shaped beam. Under earthquake 

excitation, shear force between the interfaces of the friction damper is induced. If the friction is 

overcome with the shear force, the damper slides. During the excitation, the motion of the damper 

switches from the non-sliding phase to the sliding phase, and vise versa. The location vector of the 

damper is b = -1 and the output vector for the story drift is D = [-b 7 oj= [l o]. The ratio between

393 

the stiffness of the bracing and that of the structure is assigned to be two so that the stiffness of the 

bracing is k b = 8000 kN/m. 

Since the non-linear behavior of the friction damper is utilized to dissipate the structural energy 

induced by external excitation, both the effective stiffness and damping may be increased after 

implementation of friction damper. When the maximum friction force of the damper is very low, that 

is ("nuxM) -» 0, it is very easy for the damper to slide. It is equivalent to the case of bared frame 

where friction damper does not exist. When the maximum friction force is very high, that is, 

( w mx/ m £) —»°°, the damper does not slide. It is equivalent to the case of braced frame. If the 

excitation is known a priori, the maximum friction can be designed to maximize the damping of the 

friction-damped structure. Under El Centra earthquake which is normalized to 1 g, the optimal ratio 

between the maximum friction and the weight of the structure is 2. After friction damper is added to 

the structure, the relative displacement, relative velocity and absolute acceleration are all reduced 

dramatically (Figures 2 and 3). Comparing to the bared frame, the reduction in relative displacement, 

relative velocity and absolute acceleration of the friction-damped frame are 64.4% * 54.6% and 48.5%, 

respectively. The hysteresis loop is in rectangular shape which is consistent with the assumption of 

Coulomb friction. 

CONCLUSIONS 

After theoretical derivation and numerical simulation, it is verified that friction damper is an efficient 

device for energy dissipation. If the characteristics of the excitation are known a priori, optimal ratio 

of the maximum friction force to the weight of the structure can be designed in order to achieve the 

best damping effectiveness. In this paper, an innovative and systematic numerical procedure by which 

a unified motion equation can be adapted for both the non-sliding and sliding phases of the system. 

ACKNOWLEGEMENT 

This research was supported in part by the National Science Council. This support is greatly 

appreciated. 

REFERENCE 

Venuti, W. J. (1976). Energy absorption of high strength bolted connections, Test Report, Structural 

Steel Education Council, California, USA. 

Pall, A. S., and Marsh, C. (1979). 'Energy dissipation in panelized buildings using limited slip bolted 

joints', Proceedings, AICAP-CED conference, 3, Rome, Italy. 

Fitzgerald, T. F., Anagnos, T., Goodson, M., and Zsutti, T. (1989). 'Slotted bolted connections in 

aseismic design of concentrically braced connections', Earthquake Spectra, 5. 

Roik, K., Dorka, U., and Dechent, P. (1988). 'Vibration control of structures under earthquake loading 

by three stage friction grip elements', Earthquake Engineering and Structural Dynamics, 16, 501-521.

394 

Pall, A S , and Marsh, C (1982) 'Response of friction damped braced frames', Journal of Structural 

Division, ASCE, 108,1313-1323 

Filatrault A , and Cherry, S (1987) 'Performance evaluation of faction damped braced steel frame 

under simulated earthquake loads', Earthquake Spectra, 3, 57-87 

Pekau, 0 A, and Guimond, R (1991) Controlling seismic response of eccentric structures by 

faction dampers', Earthquake Engineering and Structural Dynamics, 20, 505-521 

Pall, A , Vezina, S , Proulx, P , and Pall, R (1993) 'Friction dampers for seismic control of Canadian 

space agency headquarters' Earthquake Spectra, 9, 547-557 

Assign initial value 

jfc = 0 

z[o] = 0 

Assume non-sliding mode 

Estimate u[k +1] by equation (12) 

Assumption is contradicted 

Compute z[k +1] by equation (3) 

Figure 1 Flow chart for numerical analysis of faction-damped structures

395 

splacemen 

(m) 

Bared 

—Braced/Damped 

>» .-* 

±! O 

o 

o

Acceleration 

(m/sec 2 ) 

Velocity 

(m/sec) 

Displacement 

(m)




BEHAVIOR OF SEISMIC PROTECTIVE DEVICES: A 

COMPUTATIONAL MECHANICS APPROACH 

G. F Dargush, H. Cho and R Radhaknshnan 

Department of Civil, Structural and Environmental Engineering 

State University of New York at Buffalo 

135 Ketter Hall, Buffalo, New York 14260, USA 

ABSTRACT 

Structural control technology continues to advance rapidly on many fronts through the introduction of 

new devices and systems. At the same time, there is a need to continually improve our understanding 

Of existing protective technologies via carefully controlled physical experiments and mathematical 

modeling. In this paper, we present work related to computational continuum mechanics analysis of 

several passive devices. In particular, we consider two different metallic dampers and a solid 

viscoelastic damper. For each case, an overview of the constitutive modeling is provided, along with 

some details from the finite element analyses relating to damper behavior. Finally, we argue that such 

an approach is advisable in general, as we attempt to move toward the goal of developing disasterresilient 

communities. 

INTRODUCTION 

A wide variety of seismic protective systems have been developed over the past several decades and, 

as the concept becomes more well known, an increasing number of structures employing these systems 

appear each year Whether one considers base isolation systems (Skinner et al., 1993), passive energy 

dissipation systems (Soong and Dargush, 1997; Constantinou et al., 1998) or active/semi-active control 

systems (Soong, 1990), there are key elements that will to a large extent determine the performance of 

the entire structure during significant earthquakes In many cases, these key elements are the seismic 

protective devices. In this paper, we advocate a philosophy that treats seismic protective devices as 

critical components, necessitating a high level of reliability This, in turn, requires a rigorous program 

of engineering analysis, testing and design. Here we focus on aspects associated with engineering 

analysis, and present the application of computational continuum mechanics to study the behavior of 

several passive devices. 

First, we examine metallic dampers and develop a two-surface cyclic plasticity model to characterize 

their behavior. Model parameters are established to capture essential features of structural steel under 

constant amplitude cyclic loading The resulting model is then utilized in finite element analyses to 

study the behavior of an E-damper for seismic isolation systems and a triangular plate energy absorber 

(TPEA) for passive energy dissipation. In both cases, the numerical solutions are compared with the 

results of physical experiments. Additionally, several significant aspects of the response can be 

elucidated with a computational mechanics approach.

398 

As a further illustration of this approach, we consider the behavior of viscoelastic dampers. 

Constitutive models based upon fractional derivative (Makris and Constantinou, 1991; Makris et al., 

1993) and generalized Maxwell representations have been proposed in the literature. Here we employ 

the latter representation to study device response. A finite element method is used to solve the 

associated coupled thermomechanical problem for a two-dimensional model of solid viscoelastic 

dampers. Again, numerical results are compared with data from physical experiments and some 

interesting aspects of the cyclic response are revealed. 

Finally, based upon the results of these case studies, arguments are given in favor of an increasingly 

important role for computational continuum mechanics in the development and design of seismic 

protective devices. 

METALLIC DAMPERS 

Constitutive Model 

Practical two-surface phenomenological models, based primarily on the work of Krieg (1975), have 

found wide application in the computational mechanics literature, and are adopted here for the cyclic 

analysis of metallic dampers. The specific formulation selected is a modified version of the model 

developed in Banerjee et al. (1987) and Chopra and Dargush (1994). The stress space behavior of the 

model is depicted in Fig. la, which shows two distinct, but nested, cylindrical yield surfaces. The inner 

or loading surface separates the elastic and inelastic response regimes. It is characterized by its center 

and radius represented by the back stress and inner yield strength, respectively. On the other hand, the 

outer or bounding surface, which completely contains the smaller inner surface, is always centered at 

the origin of stress space with radius equal to variable outer yield strength. Translation of the inner 

surface corresponds to kinematic hardening, while expansion of the outer surface produces isotropic 

hardening. This separation of kinematic and isotropic hardening mechanisms proves to be quite useful 

in representing the stabilized cyclic response of structural steel. The yield criteria, flow rules, and 

hardening rules are established to ensure that the state of stress always lies on or within both surfaces, 

that all transitions during loading are smooth, and that infinitesimal strain cycles do not cause 

anomalous behavior. The present model requires the determination of five inelastic material 

parameters, including the yield strength of the loading surface, the initial yield strength of the 

bounding surface and three hardening parameters. (Details of the model are recorded in Sant, 2002). 

The model has been implemented as a user-defined material model within the general-purpose finite 

element program ABAQUS (1998). Three-dimensional, plane stress and uniaxial versions of the model 

have been developed. Both explicit and implicit integration schemes are available. 

This two-surface model was used to represent ASTM A36 structural steel. Material parameters were 

established as follows. The elastic modulus and Poisson ratio were equated with the usual handbook 

values, while the remaining five parameters were established from the stabilized cyclic data presented 

by Cofie and Krawinkler (1985). Parameter values, obtained from the Marquardt (1963) algorithm for 

nonlinear least squares curve fitting, are presented Dargush et al. (2002). A comparison of the stressstrain 

response obtained from the two-surface model and the experimental data is displayed in Fig. Ib.

399 

Figure 1: Two-surface Model Definition (a) Stress Space, (b) A36 Steel Cyclic Response 

E-shaped Metallic Damper Analysis 

The Italian company ALGA has developed a seismic isolation system consisting of lubricated sliding 

bearings to resist gravitational load and energy dissipating devices shaped like the capital letter "E" to 

dissipate horizontally transmitted destructive earthquake energy. A pair of E-shaped devices is 

responsible for the dissipation of energy in one direction. Therefore, two pairs are used in one bearing 

to dissipate the horizontally transmitted energy. The central leg and external legs are hinge-connected 

to the top and bottom plates of the sliding bearings. 

The two-surface model discussed in the previous section is now applied in a finite element study of the 

cyclic response of the E-damper. A deformed shape plot of the damper obtained with plane stress 

elements is shown for the maximum displaced position in Fig. 2a, while load-displacement curves 

from the numerical model and experimental testing (Marioni, 1991) are compared in Fig. 2b. The 

overall agreement is quite good, particularly since the constitutive model parameters were determined 

independently from the damper experiments. Although some minor differences exist, relating 

primarily to the material stabilization processes, the detailed plastic strain distributions could be used 

for a preliminary assessment of durability. 

KT7 

• 

Marioni (1991) 

30 . Two Surface Model 

20 

~ 10 

« / r 

-150 -100 -50 0 50 100 150 

Displacement (mm) 

Figure 2: E-Damper (a) Deformed Shape, (b) Cyclic Response

400 

Triangular Plate Metallic Damper Analysis 

Next the two-surface plasticity model is applied to study the cyclic response of metallic triangular 

plate dampers. For the finite element analysis, a solid element model for one-half of one plate is used. 

The particular triangular-shaped damper plate studied here has an overall length of 0.304m, a base 

width of 0.1333m and a 0.0361m thickness. This damper is designated 2B2 by Tsai et al. (1993). The 

wide end of the plate is clamped, symmetry is enforced about the centerline and a constant amplitude 

displacement-controlled loading history is applied normal to the plate at the tip. The plate is subjected 

to five complete loading cycles at low frequency. Quasistatic response is assumed. 

In the finite element analysis, the amplitude of the enforced displacement at the tip of the plate is set at 

0.0912m. This corresponds to a nominal angle x 0 =0.30 as defined in Tsai et al. (1993) and Dargush 

and Soong (1995). The longitudinal plastic strains are presented in Fig. 3a at maximum displacement 

with the deformation shown at actual scale. Meanwhile, the damper force-displacement response, 

using the present two-surface model is displayed in Fig. 3b. Not only does the response stabilize, the 

shape is quite similar to that obtained hi experiments (Tsai et al., 1993). Furthermore, the stiffening 

that occurs at large strain amplitudes can be attributed to the effects of large deformation. One often 

must account for these effects in passive devices due to the intentional high concentration of energy 

dissipation. Although these trends are well predicted, the magnitude of the forces obtained from the 

present analysis is approximately 20% higher than those measured in the experiments. Further 

investigation is needed to properly account for this difference. 

Figure 3: Triangular Plate Damper (a) Longitudinal Plastic Strains, (b) Cyclic Response 

VISCOELASTIC DAMPERS 

Constitutive Model 

A number of models for viscoelastic solids and fluids have appeared in the literature. For example, 

Shen and Soong (1995) used a Williams four-parameter model, while Makris et al. (1993, 1995) and 

Kasai et al. (1993) employed a fractional derivative model. Both models have attractive features that 

permit material characterization over a rather broad frequency range with a limited number of 

parameters. Under certain circumstances, one can also establish a theoretical basis for these models 

from the underlying micromechamcs. However, neither of these models is particularly effective for 

transient nonlinear analysis, which is necessary for proper description of response under significant 

seismic excitation. The difficulty pertains to the fact that the time-dependent relaxation moduli 

associated with these models are not separable. This means that to compute the stress at any instant of

401 

time, one must integrate over the entire strain history beginning from time zero. Thus, stress 

calculations become more and more expensive as the solution progresses. On the other hand, 

generalized Maxwell (or Prony series) models possess a separable relaxation modulus that greatly 

reduces the computational burden. Additionally, the generalized Maxwell model can provide a very 

reasonable representation of the frequency dependent response of viscoelastic solids, and consequently 

is adopted in the present work. Although details cannot be provided here, Fig. 4 illustrates the 

behavior of the model in the frequency range of primary interest. Furthermore, we assume that the 

material is thermorheologically simple, leading to the definition of an intrinsic time scale that may 

vary throughout the damper, depending on current temperature. 

VE CONSTITUTIVE MODEL 

Generalised Maxwell Model 

Comple< Moduli 

Figure 4: Generalized Maxwell Model - Frequency Domain Response 

VE Damper Analysis 

The thermally sensitive generalized Maxwell constitutive model described in the previous section is 

readily available in the commercial finite element code ABAQUS (1998) within the context of a 

quasistatic coupled thermomechanical analysis. In order to incorporate dynamic effects within the 

damper, user-defined subroutines must be developed (Radhakrishnan, 2000). However, results 

reported here utilize only the quasistatic formulation. 

As an illustrative example, consider a constant amplitude cyclic strain-controlled analysis of a single 

layer VE damper. For this analysis, the width of the damper is 3.5in, while the thickness of the 

viscoelastic layer is 0.5in. The adjoining 0.5in thick steel plates are also included in this plane strain 

thermomechanical analysis. The damper is assumed to begin at rest in an unstressed state at a uniform 

temperature of 21.7°C. Then constant shear strain range sinusoidal cycling begins with 50% 

amplitude at an excitation frequency of 1 Hz for a duration of 10s. 

Some results are shown in Figs. 5-7. Figure 5 presents the temperature response as a function of time 

for points in the center of the viscoelastic layer and at the VE-steei interface. The most significant rise 

occurs at the center, where an increase of approximately 6°C during the ten seconds of excitation. On 

the other hand, very little temperature increase occurs at the interface because heat is conducted into 

the steel plates. The overall wavy nature of these curves is a consequence of the sinusoidal excitation.

402 

This trend has also been reported in experiments (Kasai et al., 1993). Meanwhile, the overall forcedisplacement 

response is displayed in Fig. 6. Viscous dissipation leads to the development of the 

increased temperatures that, in turn, cause thermal softening of the VE layer. This produces the 

degradation of the hysteresis loops shown in the figure. Again, this is evident in experiments and must 

be properly accounted for in VE damper design. Detailed contour plots of temperature and shear strain 

throughout the VE layer at maximum lateral displacement are provided in Fig. 7. Of particular 

concern are the shear strain concentrations that appear on the VE-steel interface at the free edges. 

These can lead to debonding failures. 

Figure 5: VE Damper - Temperature Response 

Figure 6: VE Damper - Force-Displacement Response 

f .««*-««*,.. *,.•*., f 

Figure 7: VE Damper - Temperature and Shear Strain Distributions

403 

CONCLUSIONS 

In the previous sections, we have provided an overview of some computational mechanics studies of 

seismic protective devices. Both metallic dampers and viscoelastic dampers are considered. 

For the metallic dampers, a two-surface cyclic plasticity model is developed that has the capability to 

reproduce stabilized hysteresis loops under constant amplitude cycling, while retaining an inherent 

self-consistency under more general loading paths. Model parameters are established from 

experimental data on structural steel that is independent of the damper response tests. For both the E- 

damper and the triangular plate damper, the subsequent finite element analysis captures the main 

features of the experimental damper results, while providing additional insight into the behavior. In 

the case of the triangular plate damper, for example, the analysis identifies the important role of large 

deformation effects and potential locking. The analysis also indicates that the triangular shape may not 

be ideal for response under severe lateral deformations. 

Meanwhile, the viscoelastic dampers are modeled as a thermorheologically simple generalized 

Maxwell material. This provides a reasonable representation of both the frequency and temperature 

dependence, while retaining computational efficiency. The finite element analysis captures the overall 

damper behavior, including thermal softening. The continuum model also provides detailed 

information that could be used to better understand damper durability, particularly in terms of 

interfacial debonding. 

In ail three cases, the computational continuum mechanics approach provides additional insight into 

the behavior of the control device. This complements the results of physical experiments and the 

initial back-of-the-envelope calculations that are more routinely performed. During a significant 

earthquake, the performance of the entire structural system will often ultimately depend upon the 

reliability of these critical control devices. Consequently, we recommend detailed computational 

continuum mechanics evaluations of all seismic protective devices. 


Support for the work described in this paper was provided in part by the Multidisciplinary Center for 

Earthquake Engineering Research under a cooperative agreement from the U.S. National Science 

Foundation (Grant EEC-9701471). The authors gratefully acknowledge this support. 

REFERENCES 

ABAQUS (1998), Theory Manual, Version 5.8, Pawtucket, RI. 

Banerjee, P.K., Wilson, R.B. and Raveendra, S.T. (1987), Advanced Applications of BEM to Threedimensional 

Problems of Monotonic and Cyclic Plasticity, Int. J. Mech. ScL s 29(9), 637-653. 

Chopra, M.B. and Dargush, G.F. (1994), Development of BEM for Thermoplasticity, Int. J. Solids 

Struct., 31, 1635-1656, 

Cofie, N.G. and Krawinkler, H. (1985), Uniaxial Cyclic Stress-Strain Behavior of Structural Steel, J. 

Engrg Mech., ASCE, 111(9), 1105-1120.

404 

Constantinou, M.C., Soong, T.T. and Dargush, G.F. (1998), Passive Energy Dissipation Systems for 

Structural Design and Retrofit, Monograph No. 1, Multidisciplinary Center for Earthquake 

Engineering Research, Buffalo, NY. 

Dargush, G.F., Cho, H. and Sant, R.S. (2002), Cyclic Elastoplastic Analysis of Metallic Dampers for 

Seismic Energy Dissipation, Seventh U.S. Nat. Conf. Earthquake Engrg., EERI, Boston, MA. 

Dargush, G.F. and Soong, T.T. (1995), Behavior of Metallic Plate Dampers in Seismic Passive Energy 

Dissipation Systems, Earthquake Spectra, 11, 545-568. 

Kasai, K., Munshi, J.A., Lai, M.L. and Maison, B.F. (1993), Viscoelastic Damper Hysteretic Model: 

Theory, Experiment and Application, Proc. ATC 17-1 on Seismic Isolation, Energy Dissipation, and 

Active Control, San Francisco, CA, 2, 521-532. 

Krieg, R.D. (1975), A Practical Two Surface Plasticity Theory, J. Appl Mech., ASME, E42, 641-646. 

Makris, N. and Constantinou, M.C. (1991), Fractional-Derivative Maxwell Model for Viscous 

Dampers, J. Struct Engrg., ASCE, 117(9), 2708-2724. 

Makris, N., Constantinou, M.C. and Dargush, G.F. (1993), Analytical Model for Viscoelastic Fluid 

Dampers, J. Struct. Engrg., ASCE, 119(11), 3310-3325. 

Makris, N., Dargush, G.F. and Constantinou, M.C. (1995), Dynamic Analysis of Viscoelastic Fluid 

Dampers, J. Engrg. Mech., ASCE, 121(10), 1114-1121. 

Marioni, A. (1991), Antiseismic Bearing Devices on the Mortaiolo Viaduct, Third World Congress on 

Joint Sealing and Bearing Systems for Concrete Structures, Toronto, Ontario, Canada, 1263-1280. 

Marquardt, D.W. (1963), An Algorithm for Least-Squares Estimation of Nonlinear Parameters, 

J. Soc. Ind Appl. Math., 11(2), 431-441. 

Radhakrishnan, R. (2000), Coupled Thermomechanical Analysis of Viscoelastic Dampers, M.S. 

Thesis, University at Buffalo, NY. 

Sant, R.S. (2002), Evolutionary Structural Optimization for Aseismic Design, Ph.D. Dissertation, 

University at Buffalo, NY. 

Shen, K.L. and Soong, T.T. (1995), Modeling of Viscoelastic Dampers for Structural Applications, J. 

Engrg. Mech., 121(6), 694-701. 

Skinner, R.I., Robinson, W.H., and McVerry, G.H. (1993), An Introduction to Seismic Isolation, 

Wiley, Chichester and New York. 

Soong, T.T. (1990), Active Structural Control: Theory and Practice, Longman London and Wilev 

y 

New York. 

Soong, T.T. and Dargush, G.F. (1997), Passive Energy Dissipation Systems in Structural Engineering, 

Wiley, London and New York. 

Tsai, K.C., Chen, H.W., Hong, C.P. and Su, Y.F. (1993), Design of Steel Triangular Plate Energy 

Absorbers for Seismic-Resistant Construction, Earthquake Spectra, 9(3), 505-528.




LARGE-SCALE TESTS ON SMART STRUCTURES 

AND SEMI-ACTIVE CONTROL BY MR DAMPER 

H. Fujitani 1 , T. Azuhata 1 , K. Morita 2 , T. Hiwatashi 3 , Y. Shiozaki 3 , K. Hata 4 , K. Sunakoda 5 , 

and C. Minowa 6 

1 Department of Structural Engineering, Building Research Institute, 

Tsukuba, Japan 

2 National Institute for Land and Infrastructure Management, 

Ministry of Land, Infrastructure and Transport, 


3 Visiting Researcher, Department of Structural Engineering, Building Research Institute, 


4 Director, Central Research & Development Division, Bando Chemical Industries, Ltd. 

Kobe, Japan 

5 Senior General Manager, Engineering Division, Sanwa TekM Corporation, 

Tokyo, Japan 

6 National Research Institute for Earth Science and Disaster Prevention, 


ABSTRACT 

The objectives of these large-scale tests are to confirm the effectiveness of several proposed smart 

materials, members and structural systems. The tests were conducted on the shaking table of the test 

facility of the National Research Institute for Earth Science and Disaster Prevention (NIED), Tsukuba, 

Japan. It has a maximum displacement of 22 cm, a maximum velocity of 75 cm/s, and a maximum 

weight of 500 ton. The following test series was conducted: 1) semi-active control by MR dampers 

(base-isolation system and seismic control system), 2) damage detection system (identification of 

damaged members and evaluation of new sensors), 3) rocking energy dissipation system (lift-up 

system). 

In this paper, the test methods of semi-active control by an MR damper are discussed as the main 

subject. An MR damper changes its damping force by changing the magnetic field acting on the MR 

fluid according to an electric current. Semi-active control using an MR damper stabilizes building 

responses in an earthquake better than the conventional passive control. The MR damper's basic 

characteristics have been clarified. 

This paper outlines the objectives of the large-scale tests, the design of the test frame for each 

objective and test methods of semi-active control by an MR damper.

406 

INTRODUCTION 

The Building Research Institute (BRI) of Japan and the U.S. National Science Foundation (NSF) 

initiated the U.S.-Japan Cooperative Research Program on Auto-adaptive Media (Smart Structural 

Systems) in 1998 (Otani, 2000), under the aegis of the U.S.-Japan Panel on Wind and Seismic Effects 

of the U.S.-Japan Cooperative Program in Natural Resources. At the Joint Technical Coordinating 

Committee (JTCC) meeting, research items and plans were discussed in detail for three research 

thrusts: (1) structural systems, (2) sensing and monitoring technology, and (3) effecter technology. 

BRI planned a series of large-scale tests to verify some smart systems developed m this project. The 

effectiveness of "Semi-active control by MR dampers", "Damage detection system" and "Rocbng 

energy dissipation system" will also be confirmed. 

This paper outlines the objectives of the large-scale tests, the design of the test frame for each 

objective, and static and dynamic characteristics of the test frame and one of the test results of semiactive 

control of base isolation by MR damper. 

OBJECTIVES OF LARGE-SCALE TESTS 

The basic test frame used for this large-scale test is a 3-story steel frame. It is 3m (1 span) x 4m (2m 

x 2 span) in plan and 6m high, weighs a total of 20 tons (including the weight of the 1st floor), and has 

a natural period of approximately 0.5 sec. Some devices for the response control system with semiactive 

MR dampers and for damage detection are installed in the middle frame in the shaking direction. 

Devices for the rocking energy dissipation system are installed in each column foot of the 1 st story. 

NIED's shaking table (Photo 1) was used. Figure 1 shows the performance of the shaking table. 

The shaking table has a maximum displacement of 22 cm, a maximum velocity of 75 cm/s, and a 

maximum weight of 500 ton. 

LIMIT PERFORMANCE 

Photo 1: Shaking Table of NIED 


Figure 1: Performance of NIED's Shaking Table 

The following subjects are discussed with regard to the large-scale shaking table test.

407 

Response Control by Magneto-Rheological (MR) Fluid 

Magneto-rheological (MR) dampers have been expected to control the response of civil and building 

structures in recent years because of their large force capacity and variable force characteristics. In 

general, passive control has a limitation of damping effects within a certain range of frequency and 

input levels. Semi-active control reduces both response displacements and accelerations. The MR 

damper generates a damping force, which does not depend on the piston speed (Fujitani 2000). The 

target of this subject is to improve the safety, functionality and habitability by controlling the response 

displacements and accelerations by using MR dampers. For this purpose, MR dampers and a control 

algorithm have been developed, and their validity is discussed by an analytical study and shaking table 

tests. 

MR damper 

(a) Control of base isolation system 

Damage Detection by Sensors 

(b) Two types of story drift control 

Figure 2: Test frame for response control by MR 

dampers 

Monitoring the structural soundness of a building effectively reduces its life cycle cost. A series of 

experimental tests for a monitoring system has been conducted by measuring the process in which a 

concrete device is broken. Devices are inserted in the central structural plane in each frame layer 

(Figure 3). The same device is installed in each layer, and the purpose of the test is to detect damage. 

The input earthquake wave is El Centro 1940 NS. The table was shaken three times with maximum 

velocities of lOcm/sec, 30cm/sec, and 40cm/sec. White noise excitation and micro tremor 

measurement were carried out before and after shaking. 

Two identification methods were carried out: one using the data of before and after shaking, and the 

other using the data during shaking. For the first, we used the flexibility method, the layers stiffness 

method and the identification method using multiple natural frequency changes, etc. For the second, 

we used the ARX model and parallel processing identification method (Morita 2001). 

Several types of "smart sensors" (such as line-saving systems, RDIF tags, maximum value memory 

sensors, smart temperature sensors, and AE sensors) are examined. 

Rocking Energy Dissipation System 

Rocking systems that cause rocking vibration under appropriate control during earthquakes are now 

under development (Midorikawa et. al. 2002). Some researchers have pointed out that the effects of 

rocking vibration can reduce seismic damage to buildings subjected to strong earthquake ground

408 

motions. One of these systems has weak base plates at the bottom of each steel first-story column. 

When the weak base plates yield during a strong earthquake, rocking vibration is caused in the 

building, as illustrated in Figure 4. 

Device 

Transducer 

Crack Sensor 

Control PC 

-r 

E2Z3 

Figure 3: Central frame of span direction for damage detection and smart sensors 

A base plate yields 

Figure 4: Outline of rocking system with base plate yielding 

CHARACTERISTICS OF TEST FRAME 

Figure 5 shows the Fourier spectrum of response acceleration of each story of the superstructure 

excited by white noise. The roofs primary predominant (1 st mode) frequency is 1.83Hz. This 

means that the first natural period 0.546 sec. However, the 2 nd spectrum value is smaller than the 3 rd 

value on the 3 floor, and me 2nd spectrum value is larger than the 3 rd value on the 2nd floor. This 

means that the secondary mode of vibration is being generated.

409 

s 

e a ooo 

j| soo 

y 

0 

c 

l83Hz 

,530Hz 

V.. Jv JL 7WHI 

Acceleration 

i 1 i 

Roof 

Frequency (Hz) 2nd Floor Frequency (Hz) 

I 

1500 

1 i 

s s. 

^ * 500 

5 

J 

u 

3rd Floor 

.1 X 1 

3 

" 80 

« a 

II 60 

II 40 

S 20 

5 r~*iw^^ 

2 4 6 8 10 0 2 4 6 3 1 

FreqaencvlHz) Shaking T-ible Frequency (H.) 

Figure 5: Founer spectra of acceleration response of test frame 

Table 1 shows the characteristics of test frame. 

TABLE 1.1 TABLE 1.2 

MASSES CHARACTERISTICS OF FRAME 

TABLE 1.3 

CHARACTERISTIS 

OF BASE ISOLATION 

Roof 

SrdFbor 

2nd F bor 

IstFbor 

Total 

467 

478 

478 

6 DO 

2023 

3F 

2F 

1F 

Stiffness(kN/cni) 

276 

284 

354 

EfestcOBpfetcementfcni ) 

174 

3 DO 

227 

Stiffness(kN/cm) 

Frctbn Force (kN) 

DamphgCoeffbBntWs/cra) 

DamphgRatb 

NaturalPerbdfeec) 

101 

0692 

OD333 

OD37 

284 

DEVELOPMENT OF MR FLUID 

Trial product of MR fluid "#230", which was made on an experimental basis by fourth author of 

Bando Chemical Co. LTD., was used for the MR damper. The "#230" is an oil-based fluid and has 

good stability. Properties of "#230" are shown in TABLE 2. 

TABLE 2 

PROPERTIES OF MR FLUID "#230" 

Properties 

Base fluid 

Density 

Stability (%)=V(t) / V(0) *100 

V(t):sedimental volume after t min :V(0): original sedimental volume 

#230 

Oil 

3.3*10 3 kg/m 3 

Approximately 98% 

(by volume) after 2*10 4 min.

410 

DEVELOPMENT OF MR DAMPER 

The design of the bypass-flow-type MR damper, which was developed by fifth author for semi-active 

control of base isolation, is shown in Figure 6 and TABLE 3. It is composed of three parts: bypassflow, 

pressure chambers and a reservoir. The bypass flow portion has an orifice for effectively 

magnetizing the fluid. The uniform magnetic field is applied perpendicularly to the MR fluid flow at 

the annular orifice. The thermal expansion due to the temperature rise of the MR fluid is absorbed by 

the reservoir. 

Cyclic loading tests were conducted to clarify the performance of the MR damper. Figure 7 shows 

the force-displacement relationship. The electric current to the electromagnet was set at a constant 

value of OA~ 3 A in the 0.3A interval. The hysteretic loop shows rigid-plastic like characteristics 

caused by friction force of the MR fluid with magnetic field. The force increases almost 

proportionally to the increase in the electric current. Figure 8 shows the force-velocity relationship. 

The velocity is the maximum velocity of the piston in the case of sinusoidal loading. This figure also 

shows that the force increases almost proportionally to the increase in the electnc current, and the 

force does not depend so much on the piston velocity. 

Bypass Fbw 

.Reservoir TABLE 3 

DESIGN SPECIFICATION 

Coil 

Figure 6: Structure of MR damper 

M ax Force 

Stroke 

Cy Inder Bore 

Secton 

Orifce 

Length 

Inductance 

E lectrom agnet Res stance 

M ax.Curreni 

MR fliri 

40kN 

± 295m in 

..»,-». —-iir'-v » j~,^ 

•"' *'* * ••'££ 

^rrr^^ —^-^ 

7»^^» • > -4j^ .- .^ ^" r. •'•*•« 

>— ^T-np /-T -w-^ 

1 "^ 

T 

OA 

-15 -10 -5 0 5 10 15 

Displacement (m) 

(a) O.lHz, 10cm, 6.28cm/sec 

40 

^20 

z; 

A* 

o 0 

CJ 

O 

i, 

-20 

-40 

T 

3.0A 

T 

OA 

-30 -20 -10 0 10 20 30 

Displacement

411 

*fU 

O A 

20 

7 HA I 

, 2 4A 

.'.-%-.', -21A 

* ..*•'. +• 1 SA 

• * x x V • 1 iA 

ê V* * x | * 1 M 

10 

0 

"" _^ 0 6A 

" H. - 0 3A 

s -^- — *~ a ~~~ f 7 f *^ * 0 OA 

0 10 20 30 

Velocity

412 

-OA -«— 03A - 

-09A X -Control 

-06A 

R 

t 3 

100 200 300 

Accelerator! fcm/s/s) 

(a) Absolute acceleration 

10 20 

D isplacem ent fcm ) 

(b) Relative displacement 

Figure 10: Average values of response of each story of positive side and negative side 

(El Centra 1940 NS, Maximum velocity is 50 cm/sec) 

30 

CONCLUSIONS 

This paper has described the characteristics of a large-scale test frame. One test frame was used for 

four series of experimental tests to verify each developing technology, such as semi-active control for 

base-isolation and seismic response control, damage detection and rocking energy dissipation. This 

large-scale experimental test verified the validity of application of those systems to smart structures, 

and useful experimental results were obtained for the settlement of future research. As one of the test 

results, authors verified that an MR damper can reduce the response displacements while reducing the 

response acceleration. 

ACKNOWLEGEMENT 

This work has been carried out under the US-Japan cooperative structural research project on Smart 

Structure Systems (Chairperson of Japanese side: Prof. S. Otani, University of Tokyo, Chairperson of 

U.S. side: Prof. M. A. Sozen, Purdue University). We would like to acknowledge the hard work and 

contribution of all members of the project for their useful advice and suggestions. 

REFERENCES 

Fujitam H., Sodeyama H., Haia K., Iv/ata N., Komatsu Y., Sunakoda K. and Soda S. (2000) 

Dynamic Performance Evaluation of Magneto-Rheological Damper, Proceedings of Advances in Structural Dynamics 2000, 

Vol.1, pp 319-326. 

Kamopp, D., Crosby, M. J. and Harwood, R. A. (1974). Vibration Control Using Semi-Active Force Generators, Journal of Engineering 

for Industry, Transactions of ASME, May 1941, pp.619-626. 

Midonkawa M., Azuhata T., Ishihara T, Matsuba Y., Mtsushima Y. and Wada A. (2002) Earthquake response reduction of buildings by 

Monta 

rocking structural systems, Proc.of SPIE's 9th. International Symposium on Smart Structures and Materials, Paper No. 4696- 

33 

K, Teshigawara M., Isoda H., Hamamoto T. and Mita A. (2001) Damage Detection Tests of Five-Story Steel Frame with 

Simulated Damages, Proc of SPIE's 6th International Symposium on NDE for Health Monitoring and Diagnostics, Paper No. 

4335-18 

Otani S., Hiraishi H. r Midonkawa M., Teshigawara M., Saito T. and Fujitani H. (2000) 

Development of Smart Systems for Building Structures (Invited paper), Proc.of SPIE's 7th. International Symposium on Smart 

Structures and Materials, Paper No. 3988-01.




PREDICTIVE STRUCTURAL VIBRATION CONTROL 

USING SOFT COMPUTING 

H. Furuta 1 and Y. Nomura 2 

Department of Informatics, Kansai University 

Takatsuki, Osaka, Japan 

2 Graduate School of Informatics, Kansai University, 

Takatsuki, Osaka, Japan 

INTRODUCTION 

In recent years, there has been a tendency that structures become higher and longer due to the advance 

of civil engineering technology. For such large and long structures, active structural control attracts 

attention to reduce the amplitude of vibration and to improve their safety and serviceability. In this 

study, an attempt is made to develop new structural control systems that can reduce the vibration of 

structures by introducing such soft computing technologies as recurrent neural network and predictive 

fuzzy control. 

By introducing the recurrent neural network, it is possible to realize the learning ability and to deal 

with continuous outputs. It is reported that the recurrent neural network can learn various patterns in 

higher speed than the usual Hopfield's neural network. In the integration of the output from each cell, 

the feedback system is adopted, in which every cells except the input layer are mutually connected so 

that the output is used as an input for the next step. This implies that the recurrent neural network is 

suitable for the analysis of time series data and for the structural vibration control. 

Predictive fuzzy control is also adopted to improve the accuracy of fuzzy active control. The external 

force at the next step is predicted based upon the time-series data by applying the deterministic 

nonlinear prediction method with fuzzy reasoning using neighborhood's difference. Several 

numerical examples are presented to demonstrate the efficiency of the control systems developed here. 

STRUCTURAL VIBRATION CONTROL USING RECURRENT NEURAL NETWORK 

Recurrent neural network is effective in processing of time series data (Douya, 1991) In a recurrent 

neural network, the output value as opposed to the input value of time t turns into the input value of 

time t+1. Since each input data is dealt with as having mutual relations, the recurrent neural network 

can learn simultaneously not only an input-and-output pattern but also the time change pattern during 

each input and output. 

In time t, the output value of unit i is expressed by Z t (t). In case that this unit i is in the input layer, the

414 

output value is defined as jq (t) and it is y t (t) in the case other than this unit i is in the input layer. 

where I, H, and 0 denote the input unit, intermediate unit, and output unit, respectively. In time t+1, 

yj (t+1), the output value of unit i^HUO, is expressed by Eqn. 2, where 85 (t+1) means the internal 

state expressed by Eqn. 3. Then, w, j is a weighting coefficient between unit i and unit j. The function 

f t is the monotonous increasing function that can be differentiated, and a sigmoid function is used here 

(Eqn. 4) (Kitagawa et aL, 1997). 

f^t + l)), ieHuO (2) 

!>.*,«+ £*W;« = 2X.Z-W (3) 

jel jeH(jO ys/utfuO 

, ' ; l - + exp(-x) -7—\ 

As typical learning methods of the recurrent neural network, BPTT (Back Propagation Through Time) 

that performs error propagation in the direction of reverse time, and RTRT (Real Time Recurrent 

Learning) that performs error propagation in the direction of positive time, are developed. Since the 

purpose of this study is the real time learning, RTRT is employed here. The error between the output 

value from a network and teaching data in time t is calculated as Eqn. 5 and Eqn. 6. 

E(t)=iZ(e t (t)) 2 (5) 

(t)=fy k (t)-d t (t), 

if keo 

kW [0 otherwise 

where d k (t) is the teaching data for the input value in time t. 

between the output value and the teaching value. 

In the output unit, e k (t) is the difference 

In this study, the input values of the neural network are current wind velocity, displacement and 

velocity of structure, and the control force at the previous time step. Then, learning of recurrent 

neural network is performed as follows: the input values in time t 0 are the wind velocity, the 

1 

displacement and velocity of the structure. Then, the input values at the previous time step are set to 

be zero, because ^ is the starting time. The learning is not performed and only the output value is 

calculated in order to initialize the information in a neural network. In time t+1, the input data in 

time ^ are used as the data before 1 step and set new data with current input value. The next output 

values are calculated in the same way and the learning of neural network starts with the minimization 

of the square error between output values and teaching data. The similar process is executed in real 

time learning method by numerical computation. However, it is difficult for the vibration control to 

obtain the teaching data for the learning so that virtual teaching data are used for learning, which are 

obtained by solving an approximate differential equation. Fig.l shows learning process of recurrent 

neural network.

415 

Wind Velocity (t J 

Relative Displacement (tj 

Relative Velocity (t n) 

Wind Velocity (t n_,) 

Relative Displacement^) 

Relative Velocity (t n.,) 

When applying this system to the structural 

control, it is assumed that the external force, 

the velocity and displacement are measured 

and normalized by 5% of each maximum 

value. 

Fig. 1 Learning Process of Recurrent Neural Network 

Err 

6QOE-02 

500E-02 

4DOE-02 

First, 1,000 times learning is executed by 

using 100 wind velocity data obtained at 

every 0.05 second as the input data. The 

degree of sigmoid function, learning 

coefficient, and the node number of 

intermediate layer are supposed to be 1.0, 

0.01, and 10, respectively. Fig. 2 shows the 

convergence process of the square error in 

learning. 

300E-02 

200E-02 

1OQE-02 

ODOE+00 

Fig.2 

200 400 600 

800 1000 

TineStep(005sec) 

Convergence Process of Square Error 

The calculated displacement and velocity are presented in Fig. 3 and Fig. 4, and the maximum velocity 

and displacement are presented in TABLE 1 and the average velocity and displacement are presented 

in TABLE 2. In Fig. 3 and Fig. 4, the dotted line shows the displacement or velocity without control, 

and the solid line shows the results obtained by the structural control using the recurrent neural 

network. 

0.06 

0.04 

0.02 

0 

-002 

-0.04 

-0.06 

. 

^ 

Jblg. 3 .Displacement alter Learning 

-. 

Fig.4 Velocity after Learning

416 

No control 

RNN 

No control 

RNN 

TABLE 1. MAXI1VIUM VALUES 

Velocity 

4 8143 fm/s) 

43197(m/s) 

Displacement 

0 6353 (m) 

0 6277 (m) 

TABLE 2. AVERAGE VALUES 

Velocity 

0 4733 (m/s) 

0 1215 (m/s) 

Displacement 

0 0442 (m) 

0 0236 (m) 

PREDICTIVE FUZZY CONTROL FOR STRUCTURAL VIBRATION 

Fuzzy Active Vibration Control 

In essential the fuzzy control is employed here as a basic control method m which fuzzy rules are tuned by the 

prediction results of wind velocity As a fuzzy control rule the following If-Then rules are used 

If (wind load) and (relative velocity), then (control force) 

where the antecedent and consequent parts are defined m terms of seven membership functions 

respectively Representative rules are presented m TABLE 3 

TABLE 3. REPRESENTATIVE RULES 

A 

NB 

NM 

NS 

ZR 

PS 

PM 

PB 

NB NM NS ZR PS PM PB 

PB 

PB 

PB 

PB 

PB 

PB 

PB 

PM 

NB 

NM 

NS 

PS 

PS 

PM 

PB 

NB 

NM 

NS 

ZR MS NM 

PS 

PM 

PB 

NB 

NB 

NB 

NB 

NB 

NB 

NB 

Identification of Structural Characteristics Using Neural Network 

Structural characteristics are identified by using a layered type neural network The identification of structural 

characteristics is considered as one of approximation of nonlinear functions Using the data from 5001 to 10000 

steps, wind load and control force are learned As output values the relative wind velocity at the next step is 

obtained bv implementing the neural computing with values of wind load, relative structural displacement and 

velocity at the previous step, control force at the previous step as input variables For learning, back-propagation 

method is employed 

Time Series Prediction by Chaos Theory 

If the data sampled is chaotic the behavior is regard to be governed by a deterministic rule Then, if a 

non-linear deterministic rule is estimated, it is able to predict the data of near future until the 

deterministic rule does not work due to the sensitivity to the initial state Fig 5 and Fig 6 show the 

attractor and the result of Lyapunov exponent analysis

417 

Fig. 5 4ttractor (time lag=l) Fig. 6 Result of Lyapunov Exponent Analysis 

When a time series shows a chaotic behavior it may follow a deterministic rule Then it is possible to 

predict the near future behavior of the time series with the aid of chaos theory Here it is attempted to 

predict the wind velocity given by a time series record using Local Fuzzy Reconstruction Method 

(LJFRM) (lokibe et al, 1994) and Dtermimstic Nonlinear Prediction Method using Neighborhood s 

Difference (NDM) (Sakawa et al 1998) Fig 7 and Fig 8 show prediction results by LFRM and 

NDM 

Application Example 

In the proposed system the structural characteristics are determined by the neural computing and the 

wind velocity at the next step is predicted by the chaos theory Then the wind load at the next step is 

calculated through the neural computing by using the vibration states and control force at the previous 

step as the input data Introducing the wind load calculated, fuzzy active control is implemented The 

procedure can provide the control force at a step before than the usual fuzzy control This enables to 

realize a more flexible and effective vibration control 

70 80 90 00 

Fig. 7 Prediction Result by LFRM 

Fig. 8 Prediction Result by NDM 

Figs 9 thorough 11 present the relative displacements of the cases without control, with fuzzy control 

and with the proposed system, respectively Figs 12 through 14 depict the changes of relative 

velocity TABLE 4 shows the mean displacement and mean velocity It is seen that the proposed 

system can provide the best result among the three methods to reduce the vibration 

CONCLUSIONS 

In order to realize a more accurate and practical control for structural vibration of structures, it is

418 

necessary to update control rules in real time. It is, however, difficult to apply the conventional 

vibration control svstem for the structures with the variation of structural characteristics. In this study, 

Fig 8. Relative Displacement without Control 

Fig 11. Relative Velocity without Control 

1600 I BOO 2000 

Fig 9. Relative Displacement with Fuzzy Control Fig 12. Relative Velocity with Fuzzy Control 

Fig 10. Relative Displacement with Proposed System Fig 13. Relative Velocity with Proposed System 

TABLE 4. MEAN VALUES OF RELATIVE VELOCITY AND DISPLACEMENT 

No control 

Fuzzy 

Proposed 

Mean Velocity 

0.2234 (m/s) 

0.13021 (m/s) 

0.08183 (m/s) 

Mean Displacement 

0.0205 Cm) 

0.0143 Cm) 

0.0109 (m) 

an attempt was made to develop a vibration control system of structures with the ability of selfadjustment. 

By applying recurrent neural network, it is possible to update control rules automatically, 

because the recurrent neural network has the ability of learning time series data and correcting 

weighting coefficients of neural network in real time. From the numerical simulation, it was

419 

concluded that the real time recurrent learning can reduce the intensity of vibration. This implies that 

the real time system identification of the dynamic characteristics of the structure can be sufficiently 

done. 

In this paper, another attempt was made to develop a structural vibration system that can reduce the 

displacement and velocity of the structure with more efficiency. Introducing the prediction of wind 

velocity at the next step, it is possible to realize a more effective control of structural vibration. In this 

study, the structural characteristics are identified by using the layered-type neural network. For the 

learning data, the control results given by the fuzzy control are used. Although the structural model is 

very simple, the identification result is quite satisfactory. 

In the development of predicting method of wind velocity, the chaos theory was useful in short-term 

prediction. In the utilization of the chaos theory, it is necessary to prove the chaotic behavior of the 

wind velocity record. The chaotic behavior was proven by the Lyapunov exponent analysis. The 

prediction of wind velocity could be done with more accuracy by employing appropriate parameters 

whose values were determined empirically. The proposed method may be promising for the structural 

vibration control due to earthquake excitations. 

The authors would like to acknowledge Osaka municipal office and Hitachi Zosen Corp. for providing 

the data of wind velocity measured at Osaka bay. 

REFERENCES 

Douya, K. (1991), Learning algorithm of recurrent network, /. of Instru. And Control, Vol.30, No.4, 

pp.296-301. (in Japanese) 

Kitagawa, K., T. Honma and K. Abe (1997), Emergent learning method of recurrent neural network, J. 

oflnstru. and Control Vol. 33, No. 11, pp. 1093-1098. (in Japanese) 

Takens, R (1981), Dynamical Systems and Turbulence, Springer, Berlin, pp.366-381 

lokibe, T., M. Kanke, Y. Fujimoto and S. Suzuki. (1994), Short-term prediction on Chaotic Timeby 

local fuzzy reconstruction method, Proc. of Brazil-Japan Joint Symposium on Fuzzy Systems, 

pp. 136-139 (in Japanese) 

M.Sakawa, M., K.Kato and K.Ooura. (1998) A deterministic nonlinear prediction method through 

fuzzy reasoning using neighborhood's difference and its application to actual time series data, J. 

of Japan Society for Fuzzy Theory and Systems, Vol. 10, No.2, pp381-386 (in Japanese)




DAMAGE CONTROL OF HYSTERETIC STRUCTURES BY USING 

INSTANTANEOUS OPTIMAL CONTROL ALGORITHM WITH A 

SPECIAL WEIGHING MATRIX 

LI Hui 1 , PENG Jun-yi l , Suzuki Yoshiyuki 2 , GUO An-xin 1 

School of Civil Engineering, Harbin Institute of Technology, Harbin 

150090, China 

2 Disaster Prevention Research Institute, Kyoto University, Kyoto, Japan 

ABSTRACT 

This paper studied the damage reduction of hysteretic structures by using the instantaneous optimal 

control strategy with a special weighing matrix. The reduction performance of several control 

algorithms is compared through numerical studies. In addition, several cases, which include all 

responses could be observed, or only part of responses could be observed and other responses were 

estimated by the Kalman filter and the extended Kalman filter (EKF), respectively, with the structural 

hysteretic model prior-known are also studied and compared. The simulation results indicated that the 

instantaneous optimal control strategy with the proposed weighing matrix established a further 

reduction on the damage under the same control force to compare with that obtained by LQR and the 

instantaneous optimal control strategy with conventional proposed weighing matrix. Simulation results 

also showed that the semi-active control based on the instantaneous optimal control algorithm with the 

proposed weighing matrix is more trackable than that based on conventional instantaneous optimal 

control and LQR strategies. The estimated responses by Kalman filter based on the initial structural 

parameters results in large discrepancy of the responses and control forces from that based on all state 

observer or the extend Kalman filter, so the identification of structural hysteretic model parameters are 

very important for the control implement. 

KEY WORDS: structural control; instantaneous optimal control; semi-active control; hysteretic 

structure; nonlinear system; seismic response 


Since the concept of structural control was first introduced into civil structures application, a wide 

variety of active control strategies and hardware configurations have been developed to protect 

structures against wind and seismic hazards (Housner et al, 1997). The linear quadratic regulator

422 

(LQR/LQG), H M and the instantaneous optimal control strategies are frequently employed as active 

control and semi-active control strategies. However, the weighing matrices in the performance index 

are more difficult to choose. The stirrhess and mass matrices were ever proposed and widely used as 

the weighing matrix with the strain energy and kinetic energy of the controlled building being 

minimum in nature. 

However, damping matrix or stifihess matrix can be chosen as the linear quadratic objective weighing 

matrix for the instantaneous optimal control algorithm due to the special expression of the control 

force. The control force will provide a classical proportional damping to the controlled building when 

the mass multiplying by damping matrix (MC) is selected as the weighing of the instantaneous optimal 

control algorithm that results in a reasonable damping distribution for the controlled structures. 

Simulation results indicated that the control forces obtained from the instantaneous optimal control 

algorithm with the suggested weighing matrix can further reduce the drifts and absolute responses 

significantly than that obtained from the instantaneous optimal control algorithm with the conventional 

weighing matrix (MM) and LQR under the same control force. In addition, the control strategy is more 

suitable for semi-active variable damping control (such as variable damping damper and MR damper 

and ER damper) because it just acts like passive damper with adjustable parameters in nature. The 

simulation results showed that the semi-active control based on the instantaneous optimal control 

algorithm with proposed weighing matrix is more trackable than that based conventional instantaneous 

optimal control and LQR strategies. 

2, INSTANTANEOUS OPTIMAL CONTROL WITH FULL STATE OBSERVER 

For a structure with n degrees of freedom (DOFs), the equation of motion is given by 

)= -MX S (f)+ DU(t) (2.1) 

whereMand Care the nxn mass and damping matrices, respectively, X(t),X(t),X(t) are the 

relative displacement, velocity and acceleration vectors, respectively, U(t) is the pxl active 

control force vector, D is the n x p distribution matrix that relates the effect of each control force 

with each DOF, X g (t) is the earthquake ground acceleration vector at each DOF, and F(x(t),X(t)) 

is the restoring force of the structure and can be described by many hysteretic models, herein, 

Bouc-Wen model is employed to investigate the control strategy in this study as follows 

F(x(t\ X(t))= ofcc + (1 - a}kz (2.2) 

where k is the pre-yielding stiffiiess, a is the ratio of post-yielding stirrhess to pre-yielding 

stiffness, and (l-a)fc is the hysteretic part of the restoring force in which z is described by the 

following nonlinear differential equation 

~ (2.3)

423 

with the parameters A,fi,y and n govern the shape of the hysteresis loop. *andi are the 

interstory drift and velocity, respectively. The state-space representation of Eqn. 2.1 can be written in 

the form 

(2.4) 

where A- 

-M~ I F(X,X] 

0 X 

-M' I CX\' 

matrix, I denote the identity matrix, and -1 is a matrix with -1 as the elements. 

The instantaneous optimal control strategy for nonlinear structures can be obtained as follows 

(2.5) 

where A/ is the sampling time, and superscript -1 denote the inverse matrix. The control force in Eqn. 

2.5 will miriirnize the following objective 

J(t) = Z T (i)QZ(t)+ U(t} T RU(t) (2.6) 

where Q and R are positive-semi-definite and positive-definite weighing matrices, respectively. 0 

is selected as follows in general 

M 

(2.7) 

The weighing matrix Q in Eqn.2.7 will make the kinetic energy and strain energy of controlled 

structure minimum However, another type of weighing matrix Q has showed more effectively to 

reduce responses of linear structures subjected to earthquakes and described here 

(2.8) 

in which Q n can be arbitrary positive-semi-definite matrix. The above weighing matrix Q 

generated following control force (MC for short) 

(2.9) 

The weighing matrix R is frequently selected as diagonal matrix, so the control force in Eqn.2.9 

would supply the optimal control force with classical proportional damping matrix to the controlled 

building when the diagonal elements of R are equal to each other. As a result, the value of the 

elements in R implies the additional damping ratio to the controlled building by the control devices 

and can be conveniently determined according to the response reduction demand. The additional

424 

damping ratio provided by the control devices is described 

where f a and C are the additional and the original structural damping ratios, respectively, and s is 

the diagonal element of the weighing matrix R . The instantaneous optimal control force in Eqn. 2.9 is 

(2.11) 

The control force described in Eqn. 2.11 provides more reasonable damping distribution and is more 

suitable for semi-active fluid control devices considering semi-active fluid control devices acting 

essentially as passive dampers with adjustable parameters. 

The control force in Eqn. 2.1 1 can further reduce the linear responses of structures to compare with the 

linear optimal control force and other type of control force, however, it will be further investigated in 

this study that the control force in Eqn. 2.11 can establish further reduction for the nonlinear structures 

with hysteretic restoring force in comparison with the linear optimal control strategy and other type of 

control strategies. The optimal linear control force (LQR or LQG) can be obtained in Eqn. 2.12 with 

the positive definite symmetric matrix P determined by the matrix Riccati equation. 

= -R' l B T PZ(t) (2.12) 

For simplicity, A will be replaced with A^, and 

0 

where K is the 

pre-yielding stiffness matrix. And the weighing matrices Oand R in Eqn. 2.12 and Riccati equation 

will be determined by Eqn. 2.7 and the demanding control force, respectively. The matrix in Eqn. 2.7 

will also be used to determine the instantaneous optimal control force in the Eqn. 2.5 (MM for short) in 

order to compare the performance with that established in Eqn.2.11. 

3. INSTANTANEOUS OPTIMAL CONTROL WITH OBSERVER 

It is impossible to install sensors at all DOFs to measure all responses from the practical application 

view. However, it is a realistic way that few sensors are only installed at limit locations to measure 

responses and other unobserved responses could be estimated by some filter methods, such as the 

Kalman filter and the extended Kalman filter (EKF). The Kalman filter is a stochastic filter that allows 

the estimation of the states of a system based on a linear state space model and the EKF uses a local 

linearisation to extend the scope of the Kalman filter to systems described by nonlinear ordinary 

differential equations (Maybeck 1982). 

Since the Kalrnan filter is based on linear model that inherently implies the structure to be always in 

elastic phase, it is not suitable for the nonlinear structure control problem. Comparing with the Kalman 

filter, the EKF is based on nonlinear model, its time varying model parameter can track the stiffness

425 

change continuously. In this paper, the EKF is employed to estimate the states of the nonlinear 

structure in order to implement the structural control. Suppose the nonlinear dynamic equation of 

structure with active control devices subjected to additive plant noise w may be 

Z = f(z 9 U,X g ,w) (3.1) 

And the measurement equation with noise v given by 

y v ~C v Z + v (3.2) 

where C v is the measurement matrix and y v are the observe structural states, w and v are 

statistically independent zero mean Gaussian white noise with covariance O w and R v , respectively. 

The EKF implements the state estimation through a recursion algorithm that is more suitable for online 

simulation. Every calculating step is composed of time update and measurement update equations. The 

recursion algorithm is given as follows: 

(3-3) 

(t M |O (3.4) 

l 

(3.5) 

Z(t k \t k ) = Z(t t Kj+£(Oy-C,Z(f t Kj (3-6) 

P(t k \t k } = [l-K(t k )C,]P(t k \t k J (3.7) 

In which Z(t k \ t k ] is the state estimation at t k ,P(t k \ rj is covariance matrix of error 'mZ(t k \ t t ), 

K(t k ) is Kalman gain at time t t , $(t M \t t ) is state transfer matrix from t t to t M , and 

^fe+ilO is ^ discretirized plant noise input matrix. For the sampling time A, matrix

426 

by Eqn. 2.9 and Eqn. 2.12, respectively, based on the estimated states obtained by the EKF 

4. DAMAGE EVALUATION CRITERIA 

In order to compare the performance of various control strategies, some criteria should be given. The 

damage of the hysteretic structures subjected to strong earthquake is the most critical criteria to 

evaluate the performance of the structures. Lots of damage evaluation criteria have been given for 

hysteretic structures (Cosenza et al 2000). As a whole, the peak value of the interstory drift, the energy 

dissipation, and the damage index are the most common criteria to indicate the destroy and collapse of 

structures. The Park and Ang's damage index will be employed in this study. 

In addition, the control force that is related with the power for active control devices will be another 

criteria. The ratio of the energy dissipation supplied by the control devices to that by the structural 

hysteretic characteristics is also a criterion to evaluate the efficiency of the control strategy. 

5. NUMERICAL STUDIES 

5.1 Full State-Feedback Controller 

Consider a 5-story shear building model with active control devices attached at each floor. The 

uniform lumped mass and floor stiffness are 2000kg and 10 6 N/m so that the first period of the linear 

structure is 0.99s. The critical damping ratio is 0.02 for the first two modes and the Rayleigh damping 

matrk keep constant regardless of whether the structure is linear or nonlinear. The parameters to 

describe the hysteretic characteristics are assumed to be ,4 = 1,a = 0.05,^ = 0.5 and 7 = 0.5. The yield 

displacement is 2cm, ultimate displacement is 14cm and £ =0 15 used in Park and Ang's damage 

index is assumed. The control forces in Eqns. 2.12 and 2.5 with the weighing matrix in Eqn. 2.7, and in 

Eqn. 2,11 are used in the numerical example, and the additional damping ratio in Eqn. 2.10 is assumed 

to be 0.03, 0.08, 0.13 and 0.18. The matrix R in Eqn. 2.12 and 2.5 was determined according to the 

demand of identical control force for three control strategies. EL Centro, Kobe and Northridge ground 

motion records with peak values 3.417m/s 2 , 8.18m/s 2 and 8.27m/s 2 , respectively, are used as inputs. 

The earthquake intensity is changed by multiplying coefficient ranged from 0.05 to 3.5 for EL Centro 

earthquake and 0.05 to 2 for Northridge and Kobe earthquakes with increment 0.1. 

The interstory drifts of the building subjected to various intensity earthquakes are shown in Fig. 5.1. It 

was shown that the response reduction of the first story drift is the nearly same by the three control 

strategies, MC, MM and LQR for small and moderate intensity EL Centro, however, the further 

reduction for the first story drift was established by the proposed control strategy (MC) than that 

achieved by MM and LQR for high intensity EL Centro, Furthermore, the proposed control strategy 

(MC) established significant further reduction of the 1 st and 5 th story drifts than MM and LQR by using 

the same value of control force for both Northridge and Kobe earthquakes over all intensity range, 

especially more significant further reduction established over high intensity range. 

Fig. 5.2 show the 1 st floor damage index under Northridge earthquake. The results indicate that MC 

strategy decreased the damage of the structures more significantly than MM and LQR.

427 

Shown in Fig. 5.3 is the ratio of active devices dissipated energy to the structural dissipated energy by 

three control strategies under El Centro earthquakes. In this figure, the upper lines inply that the larger 

additive damping ratio £ a . We can observed from the figure that the larger f a can dissipate more 

earthquake input energy compared to the dissipated energy by structure itself, and the latter are 

comprised of viscous damping dissipated energy and hysteretic behavior dissipated energy. However, 

the trend will weaken in the situation of larger intensity earthquakes. 

05 10 15 20 25 30 

(a) the l sr mterstory drift 

10 15 20 25 30 

(b) the 5 th tnterstory drift 

Fig. 5.1 The interstory drifts by three control strategies under El Centro earthquake 

1 0-, 

08- 

06- 

04- 

02- 

00- 

intereity 

Fig.5.2 The 1st story damage under 

Northridge earthquake 

Fig. 5 3 The ratio of dissipated energy 

under El Centro earthquake 

5.2 The Control Force with the State Estimation by EKF 

Consider the same model above, however, only the fifth floor displacement is observed and other 

states will be estimated by EKF and the Kalman filter, respectively, to implement the active control 

strategy in this case. 

full state feedback 

the Kdlnan filter estimated 

state feedback 

— full state feedback 

O the EKF estimated 

state feedback 

0 10 20 30 40 50 0 10 20 30 40 50 

(a) 

(b) 

Fig.5.4 Comparison of the 5 th relative displacement response (MC 

controller , £ a = 0.08, with the intensity =2,0) of three types of state feedback cases 

under ElCentro earthquake

428 

Fig.5.4 shows the comparison of the 5 th relative displacement response between the full state feedback, 

the EKF estimated state feedback and the Kalman filter estimated state feedback, respectively. The 

results indicate that the Kalman filter could not track the real structural response for the higher 

intensity earthquake since the nonlinear behavior of the hysteretic structure. However, the responses of 

the structure with full state feedback controller and with estimated state controller (EKF) are similar, 

which indicated that EKF can accurately estimate the nonlinear structural response. 

Although it is difficult to on-line implement the nonlinear structural state estimation during the control 

process, EKF is still valuable to further study on the online estimation for nonlinear structures with 

respect to its precision. 

6, CONCLUSIONS 

The proposed control strategy can establish further reduction of the drift and damage by using the 

same control force subjected to various earthquakes than LQR and instantaneous optimal control with 

conventional weighing matrix. In addition, the semi-active control determined by the proposed active 

control strategy could more exactly follow the performance tracks established by the corresponding 

proposed active control Finally, the active control for nonlinear structure with observation can be 

accurately implemented by the EKF state estimation online. 

REFERENCES 

Housner, G W. et al (1997). Structural control: past, present, and future. Journal of Engineering 

Mechanics, ASCE, 123:9, 897-971. 

Cosenza, E. et al (2000). Damage indices and damage measures. Progress Structure Engineering 

Material 2, 50-59 

Cosenza, E. et al (1993), The use of damage functional in earthquake-resistant design: a comparison 

among different procedures. Struct Dyn. & Ear. Engng 22, 855-868 

Soda, S. (1996). Role of viscous damping in nonlinear vibration of buildings exposed to intense 

ground motion. Journal of Wind Engineering and Industrial Aerodynamics 59, 247-264 

Yang, J. N., Li, Z. (1991) Instantaneous optimal control for linear, nonlinear and hysteretic structures: 

stable controllers. NCEER-91-0026, Nat. Or. For Earthquake Engrg, Res., State University of New 

York, Buffalo, NY. 

Maybeck, P. S. (1982). Stochastic models, estimation and control. New York: Academic Press.




42g 

AN ENERGY FRAMEWORK FOR DECENTRALIZED MARKET- 

BASED STRUCTURAL CONTROL 

Jerome Peter Lynch and Kincho H. Law 

Department of Civil and Environmental Engineering, Stanford University 

Stanford, CA, USA 

ABSTRACT 

The current state-of-practice in structural control uses the centralized Linear Quadratic Regulator 

(LQR) approach to determine optimal control forces. The centralized architecture of the LQR control 

approach does not easily scale to complex systems characterized by high sensor and actuator densities. 

In this paper, a decentralized control approach is proposed for application in structural control 

systems. Termed energy market-based control (EMBC), the control system's decision process is 

modeled after the laws of supply and demand that govern free-market economies. The demand 

function of system actuators reflect the real-time kinetic and strain energy response of the structural 

system while the supply function of the system power sources are influenced by the structure's 

external input energy- The approach is shown to yield control results comparable to those obtained 

from the centralized LQR solution. 

INTRODUCTION 

The concept of using control systems in the structural engineering domain has been proposed for about 

three decades (Yao 1972). In 1989, the first building to use an active-control system for limiting 

structural deflections during wind and seismic loadings was constructed (Kobori et al. 1991). Over 

twenty buildings, primarily in Asia, have since been constructed using a variety of structural control 

system designs (Nishitam 1998). Early control system types are termed active because one or two 

large actuators apply control forces to the structure directly. WJiile success was attained using active 

control systems, they suffered from many technological limitations such as high costs and high power 

consumption demands. In response to these limitations, most recent research efforts have been 

centered upon the design of semi-active control systems. In semi-active control, passive energy 

devices are modified to possess variability in their response properties for service as indirect system 

actuators. Examples of semi-active control devices include stiffness control devices and variable 

dampers (Takahashi et al. 1998). Semi-active control devices are reliable, compact, require power on 

the order of tens of watts, and are cheaper to manufacture and operate (Symans and Constantinou 

1999).

Innovation is driving control device designs towards smaller, cheaper, and more power efficient 

actuators for structural control. The technological improvement gained in adopting semi-active 

control serves as an example of this evolutionary trend. The trend suggests that control systems of the 

future will potentially employ up to hundreds of actuators and sensors, resulting in a large-scale 

control problem defined by high actuation and sensing densities. Current state-of-practice employs a 

centralized controller for the calculation of control forces based upon measurements obtained from 

system sensors. A centralized controller is generally not adopted in large-scale control systems 

because control force computations increase at faster than a linear rate with increases in system 

dimensionality (Lunze 1992). 

As an alternative to the centralized controller, decentralized control techniques can be considered for 

adoption in large-scale control problems. Application of decentralized control solutions result in the 

reduction of the global system into an interrelated collection of smaller subsystems. A vanety of 

decentralized control approaches can be considered for adoption in a structural control system. 

Modifications of the centralized linear quadratic regulator (LQR) in order to produce optimal 

decentralized controllers for structural control have been explored (Lynch and Law 2002). From a 

performance standpoint, it has been shown that no penalty is incurred in choosing a decentralized 

control solution. 

Most recently, the explosion in development of MEMS sensing and actuation systems has resulted in 

many large-scale control problems. With the reliability of MEMS sensors and actuators lower than 

conventional counterparts, adaptive and flexible control methods are required with decentralized 

control solutions most popular. Researchers have explored using free-market concepts as one 

approach for controlling large-scale MEMS systems (Guenther et al. 1997). By modeling the control 

system as a free-market economy, where actuators are market buyers and power source are market 

sellers, an optimal control solution can result. Market-based control (MBC) methods have also been 

applied for controlling the computational load of microprocessors and for load-balancing in data 

networks (Clearwater 1996). 

Lynch and Law (2001) have proposed a market-based control (MBC) approach specific to application 

in a structural control system. The MBC method was derived using linear demand and supply 

functions for behavioral definition of market participants. Although excellent control performance 

was attained, the linear market functions lacked a physical rationale. The scope of this research is to 

revisit the MBC derivation and to develop a more rational framework by incorporating the naturally 

occurring measures of energy in the system. Termed energy market-based control (EMBC), the 

approach will be tested using the semi-active controlled Kajima-Shizuoka building as an illustrative 

example. The EMBC control performance will be compared to that of the centralized LQR controller. 

430 

OVERVIEW OF MARKET-BASED CONTROL 

The free market economies are an efficient means of allocating amongst market participants scarce 

resources, such as labor and goods. Free markets are decentralized in the a priori sense because the 

market mechanisms operate without knowledge of the global system. The historically poor 

performance of centralized economies underscores the efficiencies of the decentralized marketplaces. 

The competitive mechanisms of a free market can be extended for application to the control paradigm. 

First, the marketplace is defined by a scarce commodity such as control power, control forces, or 

control energy, just to name a few potential quantities. In the control marketplace, the role of market

uyers and sellers are assumed by system actuators and power sources respectively. The behavior of 

buyers is defined by individual utility functions, UB, that measure the amount of utility derived by the 

buyer from purchasing the market commodity. Utility is a function of the pnce per unit commodity, p, 

the amount of commodity purchased, CB, and response measures of the dynamic system, y. Similarly, 

the sellers are governed by individual profit functions, Us, that measure the amount of profit derived 

by the seller from selling the commodity. Profit is modeled as a function of the price per unit 

commodity, p, and commodity sold, C$. 

The goal of market buyers is to maximize their utility. In doing so, maximization of their utility 

functions is constrained by limiting the total purchase cost, pCs, to be less than their instantaneous 

wealth, W. Maximization of the market sellers' profit functions is constrained by the maximum 

amount of commodity they possess, 

431 

max U SI (C SI , p) subject to C sl < C MAXl 

maxII S2 (C S2 ,p) subject to C S2

432 

STRUCTURAL ENERGY DURING VIBRATIONS 

The energy balance of a structural system dunng a seismic disturbance can easily be derived. First 

consider the equation of motion of an n degrees-of-freedom structural system subjected to a seismic 

disturbance and using controls to limit responses that would result: 

My 00 + Cx(t] 4- Kx(t) = Du(t) (2) 

The displacement response vector of the system is x(t), the control forces applied to the system by m 

actuators are represented by u(t), and the absolute displacement isy(t). The absolute displacement of 

the system, y(t), is simply the input ground displacement, x s (t), added to each term of the relative 

displacement vector, x(t). The mass, damping, and stiffness matrices are nxn'm dimension and are 

denoted by M, C, and K, respectively. It is assumed that the mass, damping and stiffness matrices are 

symmetric. D is the n x m location matrix for the application of control forces. 

Equation (2) represents the equilibrium balance of forces in the structural system at any point in time. 

Integrating the forces over the response path from the initial position, x 0 , to the final position, ,r/, yields 

the energy of the balanced system (Wong and Yang 2001). 

[*' y T Mdx+ J v x T Cdx + J V X T Kdx = J"' U T D T dx (3) 

The first term on the left-hand side of Equation (3) reflects the kinetic energy of the system while the 

third term represents the strain energy of the system. Both measures of energy are based upon 

conservative forces and are therefore path independent. Their measure is only dependent upon the 

current position and initial position of the system. Assuming the system is initially at rest, the kinetic 

and strain energy of the system can be rewritten and Equation (3) updated. 

The four terms of the left-hand side of Equation (4) represent respectively, kinetic energy (KE), 

damping energy (DE), strain energy (SE) and control energy (CE). These four energies balance the 

input energy (IE) resulting from the ground motion as shown on the right-hand side of Equation (4). 

DERIVATION OF ENERGY MARKET-BASED CONTROL 

The derivation of energy market-based control (EMBC) is centered upon a marketplace allocating the 

scarce commodity of control energy. The method begins with the selection of demand and supply 

functions that reflect measures of energy in the system. The demand and supply functions will each 

contain a "tuning" constant that can be used to vary their sensitivities. 

Derivation of Demand and Supply Functions 

In the energy marketplace, the scarce commodity of control energy is used to determine the magnitude 

of control forces applied to the structural system. The form of the demand function is selected to 

reflect two intentions of the market buyers. First, when the price of control energy is zero, the demand 

of the market buyer is equal to the input energy of its degree-of-freedom. Second, the demands of the

433 

Remaining Control Energy in Power Source 

p, price 

(a) Market Buyer Demand Function 

p, price 

(b) Market Seller Supply Function 

Figure 1 - Energy market-based control (EMBC) demand and supply functions 

market buyers asymptotically converge toward zero at infinite prices. To encapsulate these two 

characteristics, an exponential demand function for the i th market buyer, is proposed: 

(5) 

The y-axis intercept is equal to the instantaneous input energy of the ground motion at a particular 

degree-of-freedom multiplied by the market buyer's wealth, W t . The exponential decay of the demand 

function is dependent upon the kinetic and strain energy of the system as depicted in the denominator 

of the exponential term. As the response of the system increases due to greater kinetic and strain 

energy, the rate of decay decreases. The tuning constant, a, is provided to control the sensitivity of the 

demand function. Figure l(a) illustrates the behavior of the modeled demand function. 

The control system's battery sources represent the market sellers whose actions are described by 

supply functions. Each market seller has in its possession a certain amount of control energy. Again, 

two observations of the market seller's behavior are required before specifying a suitable supply 

function. First, if the price of power is set to zero, no market seller is willing to sell. Second, as the 

price grows to infinity, each market buyer would be willing to sell all of its remaining control energy 

denoted by L t . As a result, the following supply function is proposed: 

Equation (6) provides an origin intercept in addition to an asymptotic convergence to the remaining 

battery life at very large market prices. The constant J3 is used to provide a means of adjusting the 

supply function. Figure l(b) presents a graphical interpretation of the market seller supply function. 

Figure 2 - Determination of the competitive equilibrium price of control energy

434 

Equilibrium Price of Power 

With the demand and supply functions for all market participants established, the Pareto optimal price 

at each time step can be readily determined. The aggregate demand function is set equal to the 

aggregate supply function to determine the competitive equilibrium price of energy for a given time 

step. A graphical interpretation, as shown in Figure 2, is the price of control energy of the point where 

the global demand and supply functions intersect. It can be shown that this intersection point always 

exists. The solution represents a Pareto optimal price of control energy for the marketplace. 

The amount of control energy that is purchased by each system actuator is used to determine the 

applied control force. Given the instantaneous control energy purchased by an actuator, the control 

force u t can be determined from Equation (7). 

CE = M,Ax, (7) 

After control energy has been purchased, the energy is evenly subtracted from the system power 

sources. Similarly, the amount of energy purchased by an actuator times the market price per unit 

power determines the amount of wealth removed from each actuator's total wealth. 

EXAMPLE EMBC IMPLEMENTATION IN THE KAJIMA-SHIZUOKA BUILDING 

The Kajima-Shizuoka Building is used to illustrate the implementation of the derived EMBC control 

solution. The structural details of the building are presented in Figure 3 (Kurata et al. 1999). A total 

of ten semi-active hydraulic dampers, capable of changing their damping coefficient in real-time, are 

installed in the structure's weak longitudinal direction. Each SHD control device is capable of 

producing a maximum control force of 1,000 kN. 

To quantify the performance of the EMBC solution, the structure is controlled for the El Centro, Taft, 

and Northridge seismic disturbances. For the three earthquake records selected, peak absolute ground 

velocities have been normalized to a value of 50 cm/sec. The performance of the EMBC controller 

will be directly compared to that of a centralized LQR controller. 

I 

5F 

V 

VIVVIV 

1 st to 5th 

1st to 5th 

Floor 

Floor 

cj r 1 T rl1 rT r 

run 

1 

1F 

Floor Seismic Mass (kg) Story Stiffness (kN/m) 

1 215,200 147,000 

2 209,200 113,000 

3 207,000 99,000 

4 204,800 89,000 

5 266.100 84,000 

Figure 3 -The Kajima-Shizuoka Building, Shizuoka, Japan

For the implementation of the EMBC controller, the demand and supply constants, orand ft, are set to 

unity. It is determined that values of unity for these two constants makes the supply and demand 

functions sufficiently sensitive to yield excellent control results. Each floor of the structure that 

contains two actuators is provided with an equal amount of initial wealth. 

W l = 1000; W 2 = 1000; W, = 1000; W 4 = 1000; W 5 = 1000 (8) 

The total amount of power initially provided by the system power source is roughly calculated based 

upon the observation that the control system battery in the Kajima-Shizuoka Building is designed to 

last for 8 continuous minutes with 10 SHD devices each drawing 70 W of power. As a result, the total 

energy provided to the system battery sources is set to 1.25 x 10 10 J. 

L T =1.25 x 10 i0 J (9) 

An LQR controller is also implemented for the structure. The Q and R weighting matrices of the LQR 

controller are chosen to weigh with heavier emphasis on the absolute velocity response of the system 

degrees-of-freedom. 

Figure 4 presents the maximum absolute interstory drift of the Shizuoka Building when no control is 

used and when the LQR and EMBC controllers are employed. As shown, both the LQR and EMBC 

controllers are effective in reducing the drift response of the structure, with minimal differences 

between the two control performances. Only for the Northridge seismic disturbance does the LQR 

controller exhibit slightly superior performance when compared to the drift response of the EMBC 

controller at the 2 nd and 3 rd stones. 

(10) 

435 

E! Centra (50 cm/s) Taft (50 cm/s) Northridge (50 cm/s) 

"0 0.035 

Drift (m) 

0 035 0 07 

Drift (m) 

- Revised MBC Control -e- Centralized LQR Control 

Figure 4 - Maximum absolute interstory drift using LQR and EMBC controllers

436 

CONCLUSION 

The scope of this research focused upon extending the concepts of decentralized market-based control 

(MBC) using an energy framework. The decentralized EMBC controller was implemented in the 

Kajima-Shizuoka Building with excellent control results observed compared to the centralized LQR 

controller. Other market models may exist and provide more superior control performances. 


This research is partially sponsored by the National Science Foundation under Grant Numbers CMS- 

9988909 and CMS-0121S42. 

REFERENCES 

Clearwater, S. H. (1996). Market-based control: a paradigm for distributed resource allocation. World 

Scientific Press, Singapore. 

Guenther, O., Hogg, T., and Huberman, B. A. (1997). "Controls for unstable structures." SPIE Smart 

Structures and Materials: Mathematics and Control in Smart Structures, SPIE, v.3039, 754-763. 

Kobori, T., Koshika, N., Yamada, K., Dceda, Y. (1991). "Seismic response controlled structure with active mass 

driver system - Part 1." Earthquake Engineering and Structural Dynamics, 20:2, 133-149. 

Kurata, N., Kobori, T., Takahashi, M., Niwa, N., Midorikawa, H. (1999). "Actual seismic response controlled 

building with semi-active damper system." Earthquake Engineering and Structural Dynamics, 28:11, 1427- 

1447. 

Lunze, J. (1992). Feedback control of large-scale systems. Prentice Hall, New York, NY. 

Lynch, I. P., and Law, K. H. (2001). "Formulation of a market-based approach for structural control" 

Proceedings of the 19 th International Modal Analysis Conference. Society of Engineering Mechanics, Bethel, 

CT, 921-927. 

Lynch, J. P., and Law, K. H. (2002). "Decentralized Control Techniques for Large-Scale Civil Structural 

Systems." Proceedings of the 2Cf h International Modal Analysis Conference. Society of Engineering 

Mechanics, Bethel, CT. 

Mas-Colell, A., Whinston, M. D., and Green, J. R. (1995). Microeconomic theory. Oxford University Press, 

New York, NY. 

Nishitani, A. (1998). "Applications of active structural control in Japan." Progress in Structural Engineering 

and Materials, 1:1, 301-307. 

Symans, M. D., and Constantinou, M. C. (1999). "Semi-active control systems for seismic protection of 

structures: A state-of-the-art review." Engineering Structures, 21:6, 469-487. 

Takahashi, M., Kobori, T., Nasu T T., Niwa, N., and Kurata, N. (1998). "Active response control of buildings for 

large earthquakes - seismic response control systems with variable structural characteristics." Smart Materials 

and Structures, 7:4, 522-529. 

YaoJ.T.P. (1972). "Concept of structural control." Journal of Structural Division, 98:7, 1567-1574.




TOWARD THE REALIZATION OF SEISMICALLY ISOLATED 

LARGE-SPAN SPATIAL STRUCTURES: 

MODEL TEST AND SIMULATION 

T. Matsui 1 , E. Sugiyama 2 , F. Qiao 3 and T. Hibino 4 

'Department of Environmental Engineering and Architecture, 

Graduate School of Environmental Studies, Nagoya University, Nagoya, Japan 

2 Ito Architects & Engineers Inc., Nagoya, Japan 

3 0gawa Tec Corporation, Tokyo, Japan 

Department of Architecture, School of Engineering, Nagoya University, Nagoya, Japan 

ABSTRACT 

Until recently few research efforts have been directed toward the application of seismic isolation 

system in large-span spatial structures. This is due to the light weight and relatively long natural period 

of this type of structure which makes it difficult to apply the commonly used isolation system 

composed of laminated rubber isolators and dampers for multi-story buildings. In this paper, the use of 

a seismic isolation system with sliding mechanism is proposed for large-span spatial structures. In 

order to confirm its effectiveness in reducing the seismic response, shaking table tests are performed 

using a small-scale arch model. The test results are utilized to validate the simulation method, which is 

then applied to investigate the seismic response of a 200m span latticed dome supported by the baseisolation 

system. It is confirmed that the proposed seismic isolation system is effective in reducing 

substantially the horizontal and vertical accelerations as well as member stresses due to horizontal 

ground motions. 

INTRODUCTION 

Large span spatial structures such as gymnasia or auditoria are requested to have higher level of safety 

against earthquakes because they always accommodate many and unspecified spectators and are often 

used as regional refuges once the disaster occurs. In order to protect the structure from severe damages 

and to prevent the drop of hanging equipments or ceiling panels due to strong earthquake motions, it 

seems to be effective to employ a seismic isolation system. Especially, in view of the brittle nature of 

the failure of this type of structure and the difficulty of repair of damaged members over the large span 

roof, it is desirable to retain all the members within the elastic range even under extremely rare strong 

earthquake. This is only possible by employing the seismic isolation system. The seismic isolation is 

also effective to release thermal stresses, and to prevent the failures caused by unequal movement of 

the supporting structures which were frequently observed at Kobe Earthquake in 1995.

438 

The effectiveness of the seismic isolation system in multi-story buildings has been well proved and 

realized. However, until recently few research efforts have been directed toward the application of 

seismic isolation system in spatial structures. 

The most commonly used seismic isolation device for multi-story buildings is a combination of 

laminated rubber isolators to elongate the natural period apart from the dominant period of earthquakes 

and dampers to absorb vibration energies (AD, 2001). Kato et al (1998) attempted to apply such an 

isolation system in spatial structures and confirmed its effectiveness analytically in reducing seismic 

response of a 300m span dome. However, the dynamic properties of laminated rubber isolator is 

known to vary depending on many factors, such as the weight of super-structure, strain level, 

temperature and so on. It is generally difficult to elongate the natural period of commonly used seismic 

isolation devices sufficiently long to be applied to spatial structures with light weight and relatively 

long natural period (typically 1 second or so). In order to lengthen the natural period of the isolator 

without a restriction on the mass of super-structure or mechanical properties of the materials of device, 

Kawaguchi and Tatemichi (2000) proposed the "paddle isolator", a kind of rocking pendulum, and 

realized the isolator with a natural period up to 4 seconds. 

In the present paper, the use of a seismic isolation system with sliding mechanism is proposed for 

large-span spatial structures. Such an isolation system can be realized, e.g. by an elastic sliding device 

composed of a laminated rubber isolator with Teflon sheet attached to the bottom and placed on 

stainless plate, which is usually combined with commonly used laminated rubber isolators to give 

restoring forces after sliding. In order to confirm its effectiveness in reducing the seismic response, 

shaking" table tests are performed using a small-scale arch model. The test results are utilized to 

validate the simulation method, which is then applied to investigate the seismic response of a 200m 

span latticed dome supported by the base-isolation system. The effectiveness of the proposed seismic 

isolation system is demonstrated to reduce the dome response under strong earthquake motions. 

SHAKING TABLE TESTS FOR SMALL-SCALE ARCH MODEL 

Test Model 

In order to confirm the effectiveness of the 

proposed seismic isolation system, shaking 

table tests were performed using a 

small-scale arch model composed of 

polyvinyl chloride (PVC) sheet and a pair of 

aluminum tie rods, as shown in Fig. 2.1. The 

arch is hinge-supported along the bottom 

sides on a pair of horizontal edge beams 

which are roller-supported on a stainless 

plate and connected through 4 springs to the 

shaking table. A twin model of non-isolated 

arch was also manufactured which is 

hinge-supported directly on the bed of the 

shaking table. The material and mechanical 

properties of the model were selected such 

that the fundamental natural period of the 

model is close to that of this type of structure. 

Roller: 

Shaking tabl* 

FIG. 2.1 

TEST MODEL

439 

stiffness of the springs added to give restoring 

forces after sliding was tuned such that the 

natural period of the isolation system after 

sliding is 3.0 seconds. Young's modulus of PVC 

material and the friction coefficient of the roller 

were evaluated by measurement. The damping 

ratios were evaluated by the random decrement 

method from the records of forced oscillation 

tests (Vandier, 1982). The principal dimensions 

and parameters of the model are listed in Table 

2.1. 

Method of Measurement 

Tests were performed using the bi-axial shaking 

table at Nagoya University which can generate 

horizontal and vertical motions simultaneously. 

The NS and UD components of El Centro 1940 

acceleration records which were normalized such 

TABLE 2.1 

PRINCIPAL PARAMETERS OF TEST MODEL 

Span 

Rise 

Radius of curvature 

Width 

Thickness 

Half subtended angle 

Total mass of arch 

Total mass of sub-structure 

Young's modulus of PVC 

Damping ratio 

non-isolated model 

isolated model 

Stiffness of a spring 

Friction coefficient of rollers 

780mm 

212.5 mm 

480.6 mm 

300mm 

I mm 

54.2° 

0.409 kg 

0.329 kg 

2.90 kN/mnf 

0.809 N/m 

0.03 

that the maximum horizontal acceleration is 300 gal were adopted as input horizontal and vertical 

ground motions, respectively. The accelerations of the shaking table were measured by means of 

strain-gauge type acceleration meters, and the displacement responses were measured at 7 points along 

the arch section and a point on the shaking table by means of an optical image tracking system. 

Test Results 

Prior to forced oscillation tests using earthquake records, sweep oscillation tests were performed to 

measure the natural frequencies of the non-isolated arch model, of which the results are shown in 

Table 2.2. 

Fig. 2.2 shows the measured time histories of 

displacement responses at the point A of the nonisolated 

and isolated models. These are the 

displacements relative to the base-isolation system 

for the isolated model and to the shaking table for 

the non-isolated model. It is observed that large 

displacement responses are excited not only in the 

horizontal but also in the vertical directions of the 

non-isolated arch. This is due to the fundamental 

TABLE 2.2 

MEASURED AND PREDICTED NATURAL 

PERIODS OF NON-ISOLATED MODEL 

Mode 

Measured 

Predicted 

1 st mode 

(antisymmetric) 

0.632 s 

0.637 s 

2 nd mode 

(symmetric) 

0.172s 

0.180s 

(s) 

Horizontal disi 

Vertical displacanent 

Mon-isolated 

Non-isolated 

— Isolated 

FIG. 2.2 

MEASURED DISPLACEMENT RESPONSES 

AT POINT A OF ARCH

440 

anti-symmetnc mode excited by the horizontal ground motion. The large displacement responses of 

the non-isolated arch are found to be reduced drastically by introducing the base-isolation system, 

demonstrating the effectiveness of the proposed base-isolation system. The seismic isolation system 

possesses only a mechanism to isolate the horizontal ground motion. In the present case the vertical 

displacement of the arch is governed by the contribution from the anti-symmetric mode excited by the 

horizontal ground motion for which the base-isolation system is effective. This is the reason why the 

seismic isolation system is effective in reducing the displacement responses not only in the horizontal 

but also in the vertical directions. 

In order to investigate the effectiveness of 

additional springs in reducing the residual 

displacement of the base-isolation system at the 

termination of the earthquake, tests were also 

performed for the isolated model without the 

springs. Fig. 2.3 shows the comparison between 

the displacement responses of the base-isolation 

system with and without the springs. It is FIG. 2.3 

obvious that the additional springs are very MEASURED DISPLACEMENT RESPONSE 

effective in reducing the residual displacements 

OF BASE-ISOLATION SYSTEM 

SIMULATION ANALYSIS FOR EXPERIMENTAL ARCH MODEL 

Simulation Method 

The earthquake response analysis is based on the sub-structure method which divides the whole system 

into the arch and the base isolation system. Assuming that it remains within the elastic range, the 

response of the arch is expressed as a superposition of free vibration modes in the absence of the base 

isolation system. The base isolation system is idealized as a single d. o. f. system with bi-linear 

hysteresis characteristics which is movable only in the horizontal direction. This makes it easy to 

implement the parametric analysis of the coupled arch-base isolation system and to treat the different 

damping mechanisms (non-proportional damping) for the arch and the base isolation system. In the 

simulation below the measured viscous damping ratios were adopted for the arch, while no viscous 

damping was assumed for the base isolation system. The reduction of the stiffness of the arch due to 

the axial compression introduced in the process of forming the arch by bending a flat PVC sheet was 

taken into account as geometric stiffness matrices. Time-domain simulation was performed by 

employing the linear acceleration method with a time interval of 0.01 seconds. 

Comparison between Model Test and Simulation 

Prior to earthquake response analysis the free vibration analysis was performed for the non-isolated 

pin-supported arch. The predicted natural periods for the lowest two modes are shown in Table 2.2 

together with the measured values. Satisfactory agreement is observed between the prediction and 

measurement. Figs. 3.1 and 3.2 show the comparison of the measured and simulated time histories of 

displacement responses at the point A. Agreement is seen to be satisfactory between the simulation and 

model tests except the large discrepancies in the peak values of vertical displacement of the nonisolated 

model. These discrepancies may be attributed to the geometrically non-linear effect (biharmonic 

response) that is not considered in the simulation.

441 

-— «oaei test 

~^^\f^ 

Horizontal displacement 

Smulatian 

Model test - 

Sinulation 

Model test - 

^ 

10 Li 30 2i 30 

(s) 

Vertical displacement 

(s) 

Vertical displacenent 

FIG. 3.1 

MEASURED AND SIMULATED 

DISPLACEMENT RESPONSES AT POINT A 

OF NON-ISOLATED MODEL 

FIG. 3.2 

MEASURED AND SIMULATED 

DISPLACEMENT RESPONSES AT POINT A 

OF ISOLATED MODEL 

SIMULATION ANALYSIS FOR A 200M SPAN LATTICED DOME 

Analysis Model 

As analysis model a single-layer latticed dome of 200m span and 46.63m rise is selected, as shown in 

Fig. 4.1. The dome is composed of pin-jointed steel pipes and pin-supported on the base isolation 

devices at the bottom of a tension ring which is assumed to be so stiff that the isolation devices can be 

idealized as a single d. o. f. system movable only in the horizontal x-direction. In Table 4.1 are shown 

the principal dimensions of the dome which were so determined as to satisfy the allowable stress 

design criteria against the gravity loads, wind loads and temperature change of 40°C (AIJ, 1973). 

TABLE 4.1 

PRINCIPAL DIMENSIONS OF DOME 

FIG. 4.1 

ANALYSIS MODEL 

Span 

Rise 

Radius of curvature 

Half subtended angle 

Gravity load 

Total mass of dome 

Young's modulus of steel 

Poisson's ratio of steel 

Section of roof member 

Section of tension ring 

200m 

46.63 m 

130.54m 

50° 

1.96 kN/nr 

76701 

0.206 MN/mm 3 

0.3 

0500x20 

0900x50

442 

Base Isolation System 

The base isolation system 

considered herein is a combination 

of elastic sliding devices and 

commonly used laminated rubber 

isolators to give restoring forces 

after sliding, for which bi-linear 

hysteresis characteristics may 

reasonably be assumed In Tables 

42 and 4.3 are shown the principal 

parameters and specifications of 

the base isolation devices which 

were designed according to ALT 

recommendation (ALT, 2001). In 

order to prevent the sliding occur 

under ordinary wind condition the 

yield displacement was set larger 

than the maximum displacement 

2.24cm of the base isolation 

system, predicted by the spectral 

modal analysis under the design 

wind loads in the return period of 

100 years (Matsui et al, 2001). 

Free Vibration Analysis 

TABLE 4.2 

PARAMETERS OF BASE-ISOLATION SYSTEM 

Elastic natural period 

Natural period after yielding 

Initial horizontal stiffness 

Secondary horizontal stiffness 

Yield displacement 

Yield shearing force coefficient 

Friction coefficient 

Item 

1.0s 

4.0s 

303 kN/mm 

18.9kN/mm 

0.025 m 

0.10 

0.16 

TABLE 4.3 

SPECIFICATIONS OF BASE-ISOLATION SYSTEM 

Number of devices 

Diameter 

Rubber thickness 

Shear modulus of rubber 

Shape coefficient S l 

Shape coefficient S 2 

Horizontal stiffness 

Vertical stiffness 

Surface pressure 

Elastic sliding 

device 

40 

600mm 

5mmx5 

0.628 N/mm 2 

30.0 

24.0 

7. 10 kN/mm 

12.0 MN/mm 

4.43 N/mnr 

Laminated 

rubber isolator 

20 

600mm 

6 mmx22 

0.441 N/mm 2 

25.0 

4.6 

0.95 kN/mm 

2.06 MN/mm 

4.43 N/mm 2 

Prior to earthquake response 

analysis the free vibration analysis was performed for the non-isolated pin-supported dome. Symmetry 

with respect to the x-axis was exploited to analyze only a half of the dome. The natural periods and 

corresponding free vibration modes for the lowest few modes are shown in Fig. 4.2. It is noted that the 

natural periods with significant participation factors are closely spaced over the period range between 

0.58-0.25 seconds and that not only horizontal but also vertical motions are excited by the horizontal 

ground motion (anti-symmetric modes). In the earthquake response analysis below the lowest 64 

modes with significant participation factors were adopted. 

1st mode/0.580s 

(anti-symmetric) 

3rd mode / 0.528 s 

(anti-symmetric) 

4th mode / 0.524 s 

(symmetric) 

FIG. 4.2 

NATURAL PERIODS AND FREE VIBRATION MODES OF NON-ISOLATED DOME

443 

Earthquake Response Analysis 

Earthquake response analysis was performed for 

the non-isolated and isolated domes. Both the NS 

and UD components of El Centro 1940 

acceleration records which were normalized such 

that the maximum horizontal velocity is 50 kine 

were simultaneously inputted as the horizontal 

and vertical ground motions, respectively. In 

order to investigate the residual displacement of 

the isolation system at the termination of the 

earthquake, the accelerations were reduced 

gradually from 30 seconds approaching zero after 

35 seconds. Time domain simulation was based 

on the linear acceleration method with a time 

interval of 0.01 seconds for duration time of 40 

seconds. The viscous damping ratio was assumed 

to be 1.5% for the dome and 1.0% for the base 

isolation system. 

Fig. 4.3 shows the distribution of the maximum 

acceleration responses along the x-axis. It is 

observed that the large horizontal and vertical 

acceleration responses of the non-isolated dome 

are reduced drastically by introducing the baseisolation 

system. This is because the antisymmetric 

vibration modes excited by the 

horizontal ground motion are reduced by 

introducing the seismic isolation system. 

Co-latitude (dec) 

Co-latitude (dec) 

Vertical acceleration 

* Non-isolated| 

ees Isolated 

FIG. 4.3 

MAXIMUM ACCELERATION RESPONSES 

ALONG X-AXIS TO EL CENTRO 1940 

Fig. 4.4 shows the distribution of the maximum 

axial stresses (bending stresses not included) of 

1 6 10* 1 

-U 10* - L 

Displaceraent (cm) 

FIG. 4.5 

HYSTERESIS LOOP FOR BASE 

ISOLATION SYSTEM 

Co-latutude 

lagonai neabers adjacent to x-axis 

FIG. 44 

MAXIMUM AXIAL STRESS RESPONSES 

TO EL CENTRO 1940

444 

each member along or adjacent to the x-axis to which the axial stresses due to the gravity load are 

added. It is again observed that the axial stresses of the members of the non-isolated dome are reduced 

drastically by introducing the base-isolation system. 

Fig. 4.5 shows the hysteresis behavior of the base isolation system. It can be understood that the 

present isolation system with sliding mechanism absorbs vibration energies by hysteretic damping. 

The maximum displacement of the base isolation device remains within 17cm (130% of the height of 

isolators) which may be considered allowable. The residual displacement of the base isolation device 

at the termination of the earthquake remains within 2.4cm due to the effect of laminated rubber 

isolators giving restoring forces after sliding. 

It can be concluded from the above results that the proposed seismic isolation system is effective in 

reducing the dome response under strong earthquake motions. 

CONCLUSION 

A seismic isolation system with sliding mechanism was proposed for reducing the seismic response of 

large-span spatial structures. The proposed isolation system is a combination of elastic sliding devices 

with commonly used laminated rubber isolators to give restoring forces after sliding. In order to 

confirm its effectiveness in reducing the seismic response, shaking table tests were performed using a 

small-scale arch model. The test results were utilized to validate the simulation method, which was 

then applied to investigate the seismic response of a 200m span latticed dome supported by the baseisolation 

system. The effectiveness of the proposed seismic isolation system was confirmed to reduce 

the dome response under strong earthquake motions. 

References 

Architectural Institute of Japan (1973). Design Standard for Steel Structures, AIJ, Tokyo, Japan. 

Architectural Institute of Japan (2001). Recommendation for the Design of Base Isolated Buildings 

(2 nd Edition), AIJ, Tokyo, Japan. 

Kato, S., Nakazawa, S., Ueki, T., Uchikoshi, M. and Osugi, F. (1998). Earthquake response of domes 

implemented by hysteresis dampers for earthquake isolation. Lightweight Structures in Architecture, 

Engineering and Construction, lASS/IEAust/LSAA International Congress, Sydney, Australia, 451- 

459. 

Kawaguchi, M. and Tatemichi, I. (2000). Seismic isolation systems and their application in space 

structures, Bridging Large Spans from Antiquity to the Present, IASS Symposium Istanbul Turkey, 

217-228. ' 

Matsui, M., Qiao, F., Moribe, Y., Sugiyama, E. and Esaka, Y. (2001). Responses of seismically 

isolated large span domes to fluctuating wind loads, Theory, Design and Realization of Shell and 

Spatial Structures, IASS Symposium, Nagoya, Japan, TP120. 

Vandier, J.K. et al (1982). A mathematical basis for the random decrement vibration signature analysis 

technique, J. Mech, Design 104, 307-313. °




APPLICATION OF SEMI-ACTIVE DEVICES IN STRUCTURES 

SUBJECTED TO EARTHQUAKES, WIND AND VIBRATIONS 

Oliver Nagel 

Maurer Soehne GmbH & CO KG 

Frankfurter Ring 193 

D 80 807 Munich, Germany 

www.rnaur@r-soehne.de, nagel@mcrin.maurer-soehne.de 

ABSTRACT 

During the European community research programme SPACE, "smart" devices for mitigating the 

influence of vibrations on structures are developed. This paper describes the development of a 

magnetorheological damper and its practical application for the damping of a pedestrian bridge in 

Germany. Beyond that, in the project the improvement of the seismic behaviour of a steel railway 

bridge by replacing the existing passive hydraulic dampers by MR-dampers was investigated. Reason 

for this development is that the occurring earthquakes in the recent past were much stronger than the 

design earthquakes. Furthermore, a new kind of magnetic valve was developed which led to a 

considerable improvement of the dampers properties regarding the response force. Additionally, 

shake-tabie-tests with a 2-floor steel structure model (scale 1:5) were performed. 


The SPACE Project "Semi-active and Passive Control of the dynamic behaviour of structures 

subjected to Earthquakes, wind and vibrations", financed by the European Union, has its main 

objective in developing devices to mitigate the effects of earthquakes and other disturbing vibrations. 

The following systems are in the process of being developed in the course of this project: 

• A semi-active damper, based on magneto-rheological fluids, especially suited to seismic highdemand 

applications 

• An improved passive-floor isolation system, based on high damping elastomeric bearings 

• A hybrid system, combining both passive and semi-active devices 

To achieve the named objectives, it also has to be worked out (among others): 

• Suitable materials (e.g. stable magnetorheological fluid) 

• Suitable real time control algorithms and electronic hardware for different types of structures 

This paper concentrates on the semi-active damper and the developments in this connection.

446 

2 PROJECT PARTICIPANTS 

Concerning the semi-active damper, the following partners are involved in the project; 

Industrial representatives: 

• Maurer Soehne GmbH & Co KG as bridge equipment manufacturer 

• Bilfinger + Berger AG as construction company 

• Thales Underwater System as a designer / manufacturer of electronic and acoustic underwater 

equipment 

Universities: 

• KTH Polymerteknologi, largest technical university of Sweden, for the development of MR fluids 

and materials to be used in the manufacturing of the MR dampers. 

• Universita di Roma 3, for the development of mathematical control algorithms, analytical and 

numerical modelling of device behaviour 

Private research institutions: 

• ISMES, a contract research organisation, performs tests on MR devices 

• ENEA, the Italian National Agency for New Technologies, Energy and Environment, is entrusted 

with the development and validation of numerical models of the devices. 

• ENEL, one of the world's largest utilities companies, is involved in the numerical modelling of 

structures and mock-ups as well as in dynamic analysis. 

3 MAGNETORHEOLOGICAL DAMPERS 

Energy dissipation is a very important measure to mitigate the action of earthquakes, wind and other 

vibrations on structures. However, passive devices are designed on the base of clearly defined 

parameters regarding load, velocity etc. As soon as the occurring real event deviates from the designevent, 

the efficiency of the installed passive devices is diminished. 

Semi-active devices have the capability to modify their stiffness and/or damping characteristics and 

adapt their response behaviour to the real event. That means for every single event, the energy 

dissipation can be optimised and the damper can be adapted to different excitations. 

The essential characteristic of controllable fluids is their ability to change within milliseconds from a 

free-flowing, linear viscous fluid to a semisolid with a controllable yield strength when exposed to an 

electric (ER fluids) or magnetic (MR fluids) field. From today's point of view, the use of 

magnetorheological fluids has the following advantages over electrorheological fluids: 

• MR fluids are not sensitive to impurities which can arise from manufacturing and usage 

• MR fluids can be mixed with a wide choice of additives to enhance stability, seal life etc. 

» MR fluids have a maximum yield stress of up to 100 kPa (ER fluids: 3-3,5 kPa) 

» MR fluids can operate at temperatures from - 40°C to + 150°C with only slight variations in yield 

stress 

On the other hand, the employment of MR fluids makes it necessary to overcome the following 

obstacles: 

» The generation of an evenly distributed magnetic field is more complicate than the generation of an 

electric field 

• The occurring effects of sedimentation are much stronger in case of MR-fluids than in ER-fluids.

447 

3.1 Activities completed 

3.LI New magnetic valve 

The development of a new kind of magnetic valve could be completed. The improvement with regard 

to a constant response force of the damping device -independent from the movement velocities- is 

shown in the following chart (Fig. 3.1). It is obvious, that the response force area of the 2 nd step device 

is considerably increased compared to the previous device. The maximum response force already is 

achieved at low movement velocities and it stays constant, i.e. independent from the movement 

velocity. At a movement velocity of app. 10 mm/sec, the response force of the damper could be 

increased from 29 kN to 700 kN by applying a current of 4,2 amperes and 22 volts. That means: with a 

small amount of energy (less than 100 watts) we can control a power of 7,000 watts. 

^ __ t ^possible response 

force graph 

velocity [mm/s] 

Fig 3.1: comparison of adjustable damper force - 1 st and 2 nd design step of magnetic valves 

The following picture (Fig. 3.2) shows the new designed piston with a special flow channel inside and 

different adapted materials. 

Fig. 3.2: Hydraulic piston with new magnetic valves

448 

3.1.2 Investigations at Seyhan Bridge, Turkey 

The Seyhan Metro Bridge is a 3-span steel structure. The total length of the railway bridge is app. 180 

m (50 m + 80 m + 50 m), the height of the piers is 10m. The bridge was opened in 1999 with a 

seismic protection system consisting of two non-linear passive hydraulic dampers in longitudinal 

direction at each abutment, elastomeric bearings (on the piers) and unidirectional sliding pot bearings 

(on the abutments, to avoid transverse relative movements between bridge deck and abutments). 

The dampers are characterized by a quasi-constant force level, i.e. the response force of the damper 

stays constant and is independent from the movement velocity of the damper. 

The protection system was designed according to Turkish national anti seismic rules for bridge design, 

considering a peak ground acceleration of 0,36 g. 

However, the Bolu earthquake (Turkey, 1999) showed a peak ground acceleration of 0,8 g, which is 

more than the double value of the considered design earthquake. 

The goal of the investigation was to find out, how the existing seismic protection system could be 

adapted to earthquakes exceeding the design earthquake. 

After modelling the bridge and the protection devices, non-linear step-by-step dynamic analyses were 

made to single out an appropriate semi-active protection system which is capable to adapt to a broad 

spectrum of earthquakes. 

end piers 1 and 4 central piers 2 and 3 

£ 20- 

actual semi- semi- semi- actual semi- semi- semipassive 

active 1 active 2 active 3 passive active 1 active 2 active 3 

Fig. 3.3: Seyhan bridge: actual passive versus different semi-active control algorithms 

Regarding the graph of Fig. 3.3, it becomes clear that the replacement of the passive dampers by semiactive 

devices (i.e. semi-active 3), the maximum deformations of both end piers and central piers can 

be kept below the maximum design values - even in case of earthquakes much stronger than the 

design earthquake. 

4 CURRENT APPLICATIONS 

4.1 Vibration control of a footbridge by means of a magnetorheological tuned mass damper 

Tuned mass dampers (TMD) represent a powerful tool to control dynamic vibration of footbridges 

caused by excitation of crossing pedestrians. For economical and esthetical reasons, these bridges 

often are carried out as light weight structures with low natural frequencies and low structural 

damping. To avoid any damage resulting from vandalism as well as to ensure a certain comfort for 

crossing pedestrians, damping devices have to be installed.

449 

A topical example for a vibration-sensitive structure is the suspension pedestrian bridge in Forchheim, 

a town in the south of Germany (Fig. 4.1). The bridge deck was installed in summer 2002 and 

investigations about the dynamic behaviour of the bridge are made at the moment (08/2002). 

The deck is supported by 8 cables, meeting in the top end of an inclined steel pier. All main structural 

members (pier and superstructure beams) are made of steel, whereas the pedestrian deck is covered 

with wooden planks. The total length is 117.5 m, the width of the deck is 4.25 m. The height of the 

inclined pier is app. 30 m. 

Fig. 4.1 Sketch of the footbridge Forchheim, Germany 

A numerical modal analysis of the bridge showed the natural frequencies in the range between 1,0 Hz 

and 3,5 Hz. Experimental testing at the bridge pointed out that the structure is not only sensitive to one 

natural frequency, but at least two of them could be excited by a group of people crossing the bridge. 

This is the reason why the installation of one passive device would not be sufficient. 

For maximum efficiency the properties of a TMD should be chosen according to the real dynamic 

characteristics of the bridge which are usually obtained by experimental tests. However, in case the 

system properties may change (e.g. variable bridge mass due to variable live loads) or several different 

frequencies with similar eigenforms shall be dampened, the properties of the damper shall vary. The 

effectiveness of a passive TMD is considerably reduced in case the system properties deviate from the 

design properties (e.g. deviation in frequency). Therefore, instead of the conventional passive device, a 

semi-active magnetorheological damper is installed in the TMD in order to adopt the latter to the 

current frequency and to increase the device's effectiveness in different load cases. 

4.2 Characteristics of the TMD 

The mechanical properties of the semi-active magnetorheological TMD will be identical to an optimal 

passive TMD, according to the Den Hartog criteria. The passive viscous damping device will be 

replaced by a magnetic field-controlled MR damper. 

To avoid transverse movements and rotations of the oscillating damper mass, two MR-dampers are 

installed in the TMD. The damping mass of the TMD can be varied by the number of steel plates, 

inserted into the damper (Fig 4.2). This step allows a variation of the damper frequency in order to test 

the efficiency of the semi-active behaviour.

450 

Fig. 4.2: principle drawing of a semi-active tuned mass damper 

The output of the MR damper will be controlled by a special algorithm which aims at minimising the 

bridge displacements. Based on a set of continuously measured system state variables during a 

dynamic excitation, the algorithm evaluates an optimal force which will be approximated by varying 

the strength of the magnetic field. The maximum current supply for the MR device will be only a few 

amperes and therefore can be realized by a small battery and a solar panel. 

4.2.1 Numerical analysis and results 

The analysis of the footbridge was carried out using a simplified structural model (2 DOF system, 

representing the corresponding eigenform and the TMD) and the non-linear numerical model for the 

simulation of the MR damper. The investigations were performed implementing a passive and a semiactive 

TMD. The dynamic loading induced by a pedestrian was assumed by a stationary harmonic load 

(4.2.1) 

at maximum of the corresponding eigenform. The amplitudes A, represent dynamic components of a 

walking, running or jumping pedestrian (G = 1,0 kN) with corresponding phase angle 0 t . For running, 

the sum of these amplification factors can reach about 2.5 times the static load. As it is commonly 

known that pedestrians synchronize their movement due to the frequency of the bridge, the angular 

frequency (Ob is assumed to be the one of the corresponding bridge eigenform. 

In a first study the dynamic response of the footbridge in terms of its displacement (at maximum of the 

corresponding eigenform) without TMD, with a passive TMD in the optimal configuration and with a 

semi-active TMD with properties of the optimal passive one was investigated. 

As the dynamic behaviour of the footbridge with both types of TMD (optimal passive and semi-active) 

are more or less coincident, it can be concluded that no improvement (but also no deterioration) is 

obtained by implementing a semi-active damper - in case the TMD already exposes optimal properties

451 

due to the Den Hartog criteria. The effectiveness of the TMD could be observed as the maximum 

displacements are reduced to 1mm in comparison to the undamped motion with maximum value of 

about 7mm. 

When performing a numerical modal analysis of a structure, natural frequencies higher than the real 

ones are obtained in general, as the masses and material stiffness are assumed on the save side (usually 

the mass is overestimated and the material stiffness is underestimated). These assumptions finally 

result in calculated frequencies to be about 10-20% higher than the real occurring frequencies. 

Experimental testing on the footbridge always is recommended to optimize the TMD. Nevertheless, in 

order to show the influence of higher real system stiffness than the assumed one (resulting in a nonoptimal 

configuration of the TMD), the dynamic response of the bridge is shown in Fig. 4.3 for a 

running pedestrian. 

-no TWO 

~TMD with optimal tuned passive damper 

0004 

0005 

""TMD with subopttmal tuned passve damper 

-TMD with suboptimal tuned semi-active damper 

Fig. 4.3 Bridge response with non-optimal TMD properties (stiffness 

variation) 

It can be observed that in case of a non-optimal tuned TMD, the amplitudes are app. 35% higher than 

in case of optimal tuning. In case of a non-optimal tuned semi-active TMD, the maximum 

displacements (and accelerations) do not exceed the values of the optimal tuned passive TMD. This 

shows the excellent performance of a semi-active controlled TMD. 

Due to presence of live loads on the bridge deck, the mass of the structure may significantly vary 

which again results in a non-optimal configuration of the TMD. In case the total mass of the bridge is 

very high, only little influence on the natural frequency has to be expected, as it varies proportional to 

the square root of the mass. However, in case of the footbridge in Forchheim, the live loads may 

increase the total bridge mass by a factor of 2. Thus, the passive TMD cannot be tuned optimal for 

different load cases. 

In order to show this effect, the bridge mass of the simplified model was increased by a factor 1.6 

(taking into account that the corresponding eigenform only represents a part of the total mass). The 

resulting response of the bridge is shown in Fig. 4.4 for the undamped case as well as for the cases 

with a passive TMD and with a semi-active TMD.

452 

no TMD 

— TMD with optimal tuned passive dampen 

— TMD witn suboptimai tuned passive damper 

— TMD wirn sudoptimal tuned semi aeuve damper 

Fig. 4.4: Bridge response with non-optimal TMD properties (variation of mass) 

Again, the passive TMD strongly is losing its effectiveness with a non-optimal configuration, resulting 

in an increase of the displacements and accelerations by a factor higher than 2. The non-optimal semiactive 

TMD performs much better than the passive one, even if it does not reach the minimum values 

obtained for an optimal TMD. However, this result is also due to the effect that the mass ratio between 

damper and bridge masses is decreasing, which in general reduces the effectiveness of any TMD. 

5 CONCLUSION 

In the course of the Projekt SPACE, the following steps were made: 

The effectiveness of MR dampers for upgrading or retrofitting of seismic protection systems for 

bridges was investigated in the course of non-linear dynamic analyses. The developed hardware (i.e. 

new magnetic valves) was tested in laboratory and showed its broad range of variability regarding the 

response force. The practical application is made these days by installing semi-active dampers in a 

TMD of a pedestrian bridge. The application of MR-dampers for seismic protection now requires an 

open-minded client with a suitable bridge project. 

REFERENCES 

Medeot, R, The EC-fimdedproject "SPACE", 7 th International Seminar on seismic isolation, passive 

energy dissipation and active control of vibrations of structures, Assisi, Italy, October 2-5, 2001 

Bachmann, H. et al, Vibration problems in structures. Practical guidelines, Birkhauser Verlas 

Basilea, 1995 

ISO/DIS 10137, Bases for Desing of Structures. Serviceability of Buildings against Vibrations, Ginebra 

BS-54QO, Steel, Concrete and Composite Bridges. Part 2 Specifications of Loads, 1978. 

DEN HARTOQ J. P., Mechanical Vibrations, McGraw Hill, 4 th edition, 1956 

SERINO, G, OCCHIUZZI, A., Semi-active MR dampers in TMD sfor vibration control of footbridges, 

Part 1: Numerical Modeling and Control Algorithm, Proceedings of Footbridge 2002, Paris, 2002 &




ACTIVE CONTROL STUDY OF CABLE-STAYED TING KAU 

BRIDGE UNDER STOCHASTIC EARTHQUAKE EXCITATION 

Y. Q. Ni 1 , J. M. Ko 1 , Z. L. Huang 2 , J. Y. Wang 1 

1 Department of Civil and Structural Engineering, The Hong Kong Polytechnic University, 

Hung Horn, Kowloon, Hong Kong 

2 Department of Mechanics, Zhejiang University, Hangzhou 310027, P. R. China 

ABSTRACT 

Structural control provides an efficient means for protection of cable-stayed bridges against 

earthquakes, wind and other hazards. Although the concept of active control of cable-stayed bridges 

has been proposed at the end of the seventies, studies on active seismic response control of cablesupported 

bridges referring to detailed bridge models have evolved only in recent years. In this paper, 

a stochastic optimal control strategy is developed for seismic response mitigation of cable-stayed 

bridges with use of active mass drivers (AMDs). The cable-stayed Ting Kau Bridge in Hong Kong is 

selected as a paradigm to demonstrate the seismic control efficacy of the proposed strategy. Simulation 

studies show that seismic response of the bridge can be significantly reduced by implementation of 

AMDs, and the control design can be performed with use of only a few dominant modes. It is found 

that installation of AMDs on top of the bridge towers and at the middle of the main spans can achieve 

good control effectiveness. The control effectiveness for the deck displacements and internal forces 

under transverse seismic excitation is better than that under longitudinal seismic excitation. 

INTRODUCTION 

Due to high flexibility and low damping, a long-span cable-stayed bridge is vulnerable to dynamic 

loading such as earthquakes and typhoons. Structural control provides an efficient means for protection 

of cable-stayed bridges against earthquakes, wind and other hazards. Although the concept of active 

control of cable-stayed bridges has been proposed as early as the end of the seventies (Yang & 

Giannopolous 1979a,b), substantial research on this subject did not progress in the eighties. Studies of 

applying active control to cable-supported bridges have revived in the last decade. While active control 

of wind-induced flutter vibration has been extensively studied using both mechanical and aerodynamic 

measures, active seismic control of cable-supported bridges has essentially focused on exploring active 

cable tendon control and decentralized control techniques by referring to simple structural models. 

Studies on active seismic response control of cable-supported bridges referring to detailed bridge 

models have evolved only in recent years (Miyata et al 1996; Shoureshi & Bell 1996; Paulet- 

Crainiceanu 1998; Schemmann & Smith 1998a,b). A benchmark problem for seismic response control 

of the cable-stayed Cape Girardeau Bridge has been recently developed (Dyke et al 2000). A state-ofart 

review of active/semiactive control for cable-supported bridges is available (Ni et al 2001a).

454 

In this paper, a stochastic optimal control method is developed for seismic response control of cablestayed 

badges with use of active mass drivers (AMDs). Numerical simulation studies of applying the 

proposed method to the cable-stayed Ting Kau Bridge (TKB) are then carried out to demonstrate the 

control effectiveness and efficacy. Based on a precise three-dimensional finite element model of the 

bridge, a control-oriented reduced-order modal model suitable for control design is developed. 

Different AMD installation configurations (number and location) are designed for the bridge, and the 

random earthquake excitation is assumed to act in the longitudinal, transverse and 45° directions, 

respectively. The stochastic optimal control strategy based on the dynamical programming principle 

and stochastic averaging method is devised to command the operation of AMDs. Twelve evaluation 

cnteria are formulated to evaluate the control effectiveness and efficiency. Structural deflection and 

internal force responses of the bridge without and with active AMD control are obtained and compared 

under different excitation conditions. 

PROPOSED CONTROL STRATEGY 

Consider an rc-degree-of-freedom structure incorporated with active mass drivers. The governing 

equation of motion of the integrated system can be expressed as 

MX + CX + KX = ~x s ME - PDU (1) 

in which M, C , K are mass, damping and stiffness matrices of the structure, respectively; E is the 

n-dimensional vector associated with external excitation; P is an nxm matrix indicating the location 

of control devices; D is a diagonal matrix comprising the masses of AMDs used. V denotes a vector 

of relative accelerations where AMDs are attached; x g is the random ground acceleration excitation, 

its spectrum being taken herein as the non-white Kanai-Tajimi power spectral density function with the 

expression 

where % g and & g represent the damping coefficient and predominant frequency, respectively, of the 

ground motion; 5 0 is a constant power spectral density. 

Based on the mode superposition method, displacement response of the structure can be expressed in 

terms of modal transformation as 

where & c , Q c are the dominant modal matrix and displacement vector, respectively. 

Equation (1) can then be transformed into the following reduced modal coordinate equations 

a+^â+^G^-A^w-v, (i=i,2,...,o - (4) 

where / is the reduced modal order; Q l9 CD I and £ are modal displacement, frequency and damping

455 

ratio of the zth mode; /, -{ T cME} t is the participation factor of the z'th mode; v, = { 

denotes the control force corresponding to the ith mode. 

The seismic response control of the structure can be achieved through the corresponding energy 

control of structural modal vibration. By applying the stochastic averaging technique for quasiintegrable-Hamiltonian 

system (Zhu & Lin 1991; Zhu et al 1997), the averaged Ito stochastic 

differential equations with respect to modal energies can be obtained in the form of 

_ 

dH t ^[m l (H)--^v ] }dt + a i (E}dW l (t) (f,;=l,2,...,/) (5) 

dQj 

in which the model energy vector H , the reduced drift vector m(H) and the diffusion matrix a(H) 

are represented as 

H,=U2+a>Q)l2 

m,(H) = -2^ca,H, +Lfis t «o t ) 

(6a) 

(6b) 

and W, (r) is unit Wienner process. 

a;(H) = ft-S l (a),)H, (6c) 

The objective of the study is to find out an optimal feedback control law to minimize a finite horizon 

performance index. Assuming the ground excitation be stationary and ergodic, the performance index 

of the stochastic optimal control in the finite time horizon ( r 0 , T ) is expressed in the form of 

J = \imlL(Q c (T\Q c (r),U(r))dr (7) 

7"_»oo J 

0 

where L denotes a continuous differentiable convex function. 

In order to minimize the average energy diffusion of the structure, the following stochastic Hamilton- 

Jacobi-Bellman (HJB) equation for optimal ergodic control is established according to the stochastic 

optimal dynamical programming principle (Stengel 1986) 

_i t 2 o H t 

(8) 

where V represents a value function with respect to H corresponding to optimal control force. 

The optimal control force U* is then obtained by minimizing the right hand side of Equation (8). 

When the convex function has the form 

the expression of U* can be obtained as 

U (9)

456 

The dynamical programming equation is finally obtained by substituting Equation (10) into Equation 

(8) and completing the averaging with respect to Q c and Q c as follows 

+ £ [ JH (H) 4^ 

where A u = [0 T C PR' 2 P T , Jt ff,#A + o(||uf ) (12a) 

on l y=i 

the optimal control force is obtained as 

with

457 

The exact stationary probability density for modal energy H can be obtained from solving the reduced 

Fokker-Planck-Komogorov (FPK) equation associated with the averaged stochastic differential 

equation (14). That is 

p(H) = C,exp[-] A.*. 

J A a iLJ> 

ft- e (, 

where C n is a normalization constant. 

H t H,] 

' 

(16) 

After obtaining the root mean square responses of structural deflection and internal forces, the control 

effectiveness is evaluated using the following performance index 

K = - 

-ov 

(17) 

where subscripts u and c denote the uncontrolled and controlled structure; s represents the structural 

response variables, which can be displacement, velocity, acceleration or internal force. 

A total of twelve evaluation criteria have been defined to assess the control effectiveness and efficacy. 

They include: maximum and average displacement reduction; maximum and average base force 

reduction; maximum and average bending moment reduction; maximum displacement of AMDs; 

maximum control force of AMDs; ratio of total control force to weight of bridge; etc. 

CASE STUDY OF TING KAU BRIDGE 

The Ting Kau Bridge (TKB), as shown in Figure 1, is a three-tower cable-stayed bridge with two main 

spans of 448 m and 475 m respectively, and two side spans of 127 m each (Bergermann & Schlaich 

1996). The bridge deck has been separated into two carriageways with three monoleg towers located 

between. Eight longitudinal stabilizing cables with length up to 465 m have been used to strengthen 

the slender central tower. Due to adopting these special design configurations, the TKB is one of the 

most flexible cable-stayed bridges in the world. 

For modal and seismic response analyses, a precise three-dimensional finite element model of the TKB 

has been developed that contains over 15,000 degrees of freedom (DOFs). The huge number of DOFs 

in a full structural model for cable-stayed bridges makes it difficult to create computationally effective 

Fig. 1 Cable-stayed Ting Kau Bridge

458 

control strategy. Model reduction is necessary to achieve a reduced-order model suitable for control 

design purpose. Based on both modal analysis and observability/controllability evaluation, a controloriented, 

reduced-order model for the 1KB has been formulated (Ni et al 200Ib). The proposed 

modeling strategy consists of three level models. The Level I model is the three-dimensional finite 

element model that aims to depict spatial configuration and dynamic properties of the bridge. The 

Level II model, which is reduced from the Level I model and contains only the so-called important 

modes, is a mode-space model used for response prediction, state estimation and control efficacy 

evaluation. The Level III model containing only limited dominant modes is used for control synthesis 

and analysis. The scale of the Level II and III models and the selection of important and dominant 

modes are dependent on specific bridge under consideration. 

Numerical verification of effectiveness of the proposed control method for seismic mitigation of the 

TKB has been conducted under different AMD configurations and various earthquake excitations. In 

this study, the Level III model consisting of 22 dominant modes is used for control design, while the 

Level //model consisting of 47 important modes is used for response prediction and control 

evaluation. The earthquake excitations are considered to apply to the TKB in the longitudinal, lateral 

TABLE 1 

RMS DISPLACEMENT RESPONSES AND REDUCTION OF TKB USING AMDS 

Uncontrolled 

Controlled 

Reduction 

(%) 

Uncontrolled 

Controlled 

Reduction 

(%) 

Ting Kau 

tower top 

&(m) 

0.1018 

0.0614 

39.73 

1 

459 

Uncontrolled 

Controlled 

Reduction 

(%) 

179.62 

132.89 

26.02 

920.32 

514.17 

44.13 

195.41 

133.64 

31.61 

1.95 

1.74 

11.06 

9.58 

6.27 

34.49 

1.88 

1.47 

21.50 

M and Q denote the moment and shear force, respectively, 

and 45° (horizontal) directions, respectively. In this example, the random earthquake excitation 

spectrum is taken as the Kanai-Tajimi power spectral density function with the parameters % g = 0.25, 

0) s = 5 rad/s, and S 0 = 0.01 m 2 -rad/s j . For each seismic excitation, five AMD implementation schemes 

are investigated. The number of the AMDs, each with mass of 8xl0 4 kg, ranges from 5 to 20. The 

AMDs are installed on top of the three towers, at the half heights of the towers, on one or two sides of 

mid-span and quarter-span positions of the two main spans. Due to the limited space, only the control 

results of one AMD configuration under the earthquake excitation in 45° direction are reported here. 

The AMD implementation is designed as follows: one AMD on each tower top and mid-points of Ting 

Kau and Tsing Yi main spans both in lateral and longitudinal directions. As a result, a total of ten 

AMDs are utilized in this case. Figure 2 shows the reduction of cable tension forces after active control 

using the AMDs. Tables 1 and 2 list the displacement and internal force responses and the response 

reduction of the TKB after using the AMDs achieved with the proposed control strategy. It is observed 

that the seismic response of the TKB can be significantly reduced by using AMDs in conjunction with 

the proposed stochastic optimal control method. The reduction of displacement responses is larger than 

that of internal forces. Simulation studies also show that the control effectiveness for the deck 

displacements and internal forces under transverse seismic excitation is better than that under 

longitudinal seismic excitation. It is found that installation of AMDs on top of the bridge towers and at 

the middle of main spans can achieve relatively good control effectiveness. 

o 

as 

» 20 

DC 

5 

n 

:- 

: * 

: | 

: 

i 1 

: j 

r - >«• 

, 

: 

n : : ft 

' *• " 

' j 1 ' ' j 

-•-.-JV--.H 

'"•:i^ j'ii 

: i 

• f • - t * r v 

-,; 

l\ 

• 

' 

-;- 

l 

• 

' 

-.- 

\ ' 

-!-' 

! i 

; * f • 

" " i \ ' 

li '' 

::;!'.L: 

•* •• f i • 

'-•';- ' ' 

,--.-- 

1 

1 * 

\f 

, , , i , 

| I- /• 

^* 1 '• • • * 

j -i- -,-.-•- 

1*•„ |. i . •. . . 

, , , 

f\[ 

• ' • 

, * * 

•T ''f\ 

• 

; 

I , . ! , , , , ! , , , 

' 

' 

s i 

\ 

460 

(AMDs) and the proposed control strategy for seismic response mitigation of the 1KB has been 

conducted. Simulation results show that installation of AMDs on top of the bridge towers and at the 

middle of main spans of the TKB can achieve good control effectiveness both in longitudinal and 

lateral directions. Both the displacement and internal force responses at the bridge towers and deck can 

be significantly reduced by using only limited dominant modes in control design, especially for the 

central tower. The fluctuation of cable tension forces can also be reduced remarkably by implementing 

the AMDs. 


The work presented in this paper was supported by a grant from The Hong Kong Polytechnic 

University through the Area of Strategic Development Programme (Research Centre for Urban Hazard 

Mitigation). This support is gratefully acknowledged. 

References 

Bergermann, R. and Schlaich, M. (1996). Ting Kau Bridge, Hong Kong. Structural Engineering 

International 6, 152-154. 

Dyke, S.J., Turan, G., Caicedo, J.M., Bergman, L.A. and Hague, S. (2000). Benchmark control 

problem for seismic response of cable-stayed bridges. Research Report, Washington University in St. 

Louis (http://wusceel.cive.wustl.edu/quake/), 42p. 

Miyata, T., Yamada, H. and Paulet-Crainiceanu, F. (1996). Active structural control for cable bridges 

under earthquake loads. Proceedings of the 3rd International Conference on Motion and Vibration 

Control, Chiba, Japan, 2: 53-58. 

Ni, Y.Q., Spencer Jr., B.F. and Ko, J.M. (200la). Active/semiactive seismic response control of cablesupported 

bridges: current research status and key issues. Earthquake Engineering Frontiers in the 

New Millennium, B.F. Spencer Jr. and Y.X. Hu (eds.), Rotterdam, Netherlands, 299-304. 

Ni, Y.Q., Spencer Jr., B.F. and Ko, J.M. (2001b). Feasibility of active control of cable-stayed bridges: 

an insight into Ting Kau Bridge. Smart Structures and Materials 2001: Smart Systems for Bridges, 

Structures, and Highways, S.C. Liu (ed.), SPIE Vol. 4330, 387-398. 

Paulet-Crainiceanu, F. (1998). Seismic response control of cable-stayed bridges. Proceedings of the 

2nd World Conference on Structural Control, Kyoto, Japan, 2: 959-964. 

Schemmann, A.G. and Smith, H.A. (1998a). Vibration control of cable-stayed bridges-part 1: 

modeling issues. Earthquake Engineering and Structural Dynamics 27, 811-824. 

Schernrnann, A.G. and Smith, H.A. (1998b). Vibration control of cable-stayed bridges-part 2: control 

analyses. Earthquake Engineering and Structural Dynamics 27, 825-843. 

Shoureshi, R.A. and Bell, MJ. (1996). Vibration control of cable-stayed bridges: control system 

development and experimental results. Engineering Mechanics: Proceedings of the 11th Conference, 

Y.K. Lin and T.C. Su (eds.), ASCE, New York, 2: 902-905. 

Stengel, R.F. (1986). Stochastic Optimal Control, Wiley, New York. 

Yang, J.N. and Giannopolous, F. (1979a). Active control and stability of cable-stayed bridge. ASCE 

Journal of the Engineering Mechanics Division 105, 677-694. 

Yang, J.N. and Giannopolous, F. (1979b). Active control of two-cable-stayed bridge. ASCE Journal of 

the Engineering Mechanics Division 105, 795-810.

461 

Zhu, W.Q., Huang, Z.L. and Yang, Y.Q. (1997). Stochastic averaging of quasi-integrable-Hamiltoman 

systems. ASME Journal of Applied Mechanics 64, 975-984. 

Zhu, W.Q. and Lin, Y.K. (1991). Stochastic averaging of energy envelope. ASCE Journal of 

Engineering Mechanics 117, 1890-1905.




HYBRID FRC-ENCASED STEEL TRUSS BEAMS FOR SEISMIC 

UPGRADING OF REINFORCED CONCRETE STRUCTURES 

Gustavo Parra-Montesinos 1 , Subhash C. Goel 1 , and Steve Savage 2 


University of Michigan, Ann Arbor, Michigan, USA 

2 Coughlin, Porter and Lundeen, Inc., Seattle, Washington, USA 

ABSTRACT 

This paper presents an innovative scheme using hybrid beams consisting of steel trusses encased in 

fiber reinforced concrete (FRC) for seismic upgrading of RC frames with inadequate beams or slabcolumn 

frames. The behavior of various connection details between the hybrid beams and RC columns 

was evaluated through testing of four beam-column subassemblies under simulated earthquake 

loading. In this paper, experimental results are described in terms of load vs. displacement response, 

cracking pattern, beam rotations and energy dissipation capacity of the subassemblies. Test results 

indicate that moment connections consisting of either a combination of longitudinal beam bars passing 

through the RC column plus external steel plates, or the use of only external steel rods behave 

satisfactory under large displacement reversals. 

INTRODUCTION 

During the last few decades, several seismic retrofit schemes have been used in old non-ductile 

reinforced concrete (RC) buildings that include the addition of structural walls, dampers, steel braces, 

wrapping or jacketing of RC members with steel or composite materials, and base isolation. With the 

development of innovative hybrid structural systems that combine the use of steel and RC elements, 

new possibilities for seismic upgrading of RC structures are also becoming available. Hybrid beams 

that consist of steel trusses embedded in fiber reinforced concrete (FRC) for use in buildings with slabcolumn 

frames or inadequate RC beams is one scheme. These hybrid beams have been shown to 

exhibit excellent behavior under large displacement reversals (Khuntia and Goel 1998). In addition, 

hybrid FRC-encased steel truss beams are attractive from a construction viewpoint because they could 

be prefabricated or constructed on site with little or no shoring required. This can be done by first 

connecting the steel truss to the RC columns and then suspending metal or wood forms from the truss 

during the FRC casting process. 

The addition of hybrid FRC-encased steel truss beams for seismic retrofit was recently considered for 

the historic King County Courthouse in Seattle, Washington. The King County Courthouse is an Pishaped 

11-story RC building with a plan of approximately 240 by 240 feet and height of 180 feet.

464 

Square and octagonal columns are supported by individual piers and pyramid spread footings. The 

floor system consists of RC beams in the north-south direction that support the joists and 4 in. thick 

one-way slabs in the east-west direction. The beam-column system in the north-south direction 

constitutes a poorly detailed moment frame and the lack of beams in the east-west direction implies a 

very weak and soft system in that direction. Based on FEMA-356 Life-Safety performance criteria 

(FEMA 2000) for a 10% in 50 year event, the lateral-force-resisting system of the building was found 

to be inadequate when the current renovation design was undertaken. 

Results from linear and non-linear 2-D and 3-D computer analyses of the structure conducted by 

Coughlin, Porter and Lundeen, Inc. led to the selection of fluid viscous dampers in conjunction with 

hybrid FRC-encased steel truss beams for strengthening the structure in the north-south direction, 

while Yielding Steel Braced Frames were selected for the east-west direction. For the addition of the 

hybrid beams, appropriate connection details needed to be developed in order to assure adequate 

transfer offerees between the new hybrid beams and the existing RC columns. In this paper, a testing 

program aimed at evaluating various details for connecting hybrid FRC-encased steel truss beams to 

the existing RC columns of the King County Courthouse building structure is described. 

EXPERIMENTAL PROGRAM 

The experimental program consisted of testing of four beam-column subassemblies under 

displacement reversals. Specimens 1 and 2 represented interior connections, while Specimens 3 and 4 

represented exterior subassemblies. Fig. I shows a sketch of the test setup used in the experimental 

program. The four test specimens were subjected to a displacement pattern similar to that specified in 

the ACI ITG/T1.1-99 document (ACI 1999) with story drifts ranging from 0.2% up to 3.5%. 

Additional cycles, with an amplitude of approximately 4.5% drift and 5.0% drift, were applied to 

Specimens 1 thru 3 and Specimen 4, respectively, after the completion of the cycles to 3.5% drift. 

The column and beam cross section dimensions were the same for all four specimens. The column 

cross section had an octagonal shape with a side length of 8.25 in. and a 20 in. square beam cross 

section was used for all test specimens. These dimensions translated into an approximately 2/3 scale 

compared to the existing columns and the hybrid beams planned for the building. Lateral 

displacements were applied through a hydraulic actuator connected to the top of the column and to a 

strong reaction wall (Fig. 1). A small axial load was applied to the column through hydraulic jacks, 

amounting to an axial stress of approximately 150 psi. Three connection details were evaluated for 

achieving proper moment connection between the beam and column: 1) use of beam longitudinal 

reinforcing bars passing through and epoxy-bonded to the column, 2) use of beam longitudinal 

reinforcing bars combined with external steel reinforcement, 3) use of only external steel 

reinforcement for full moment transfer. In all cases, shear connection between the embedded steel truss 

and the existing column was provided to resist the full shear demand. 

Test Specimens 

Four beam-column subassemblies were designed and tested to evaluate the behavior of the three 

moment connections mentioned above. In the following, a brief description of beam, column and 

connection details used in the test specimens is given.

465 

Specimen 1 

Specimen 1 represented an interior beam-column subassembly with beams framing into the column 

from two opposite sides (Fig. 1). As mentioned above, a 20 in. square beam and an octagonal shape 

column with a side length of 8.25 in. were used in this specimen. The column was constructed prior to 

the beam to simulate the existing columns of the structure. The FRC beam had steel hooked fibers with 

a volumetric fraction of 1% and the beam reinforcement consisted of 4 #5 top and bottom longitudinal 

bars, #3 closed stirrups, and a steel truss embedded in FRC (Fig. 2). In this specimen moment transfer 

in the connection region was achieved by passing the longitudinal beam bars through holes drilled in 

the RC column, which were later epoxy injected. The chords of the steel truss were connected to a steel 

plate with slotted holes that was in turn bolted to the column (Fig. 2), and thus the truss was not 

expected to contribute to moment strength in the beam plastic hinge region adjacent to the column 

face. The truss, however, was designed to resist the shear force demand in the plastic hinge in addition 

to supporting the permanent steel forms during the concrete casting process. 

The original column design in Specimen 1 included 12 #5 longitudinal rebars representing a steel ratio 

of 1.1%. However, because core drilling of the column would be performed in this retrofit scheme, 

four bars were eliminated to account for possible cutting of longitudinal reinforcement during the 

drilling process. The column transverse reinforcement consisted of a #3 spiral with a 4 in. pitch. The 

spiral reinforcement was cut at different locations in the joint region to simulate possible cutting during 

column core drilling. To maintain adequate concrete confinement after cutting of the column 

transverse reinforcement, the column was wrapped with glass fiber sheets in the connection region. 

Specimen 2 

Specimen 2 was similar to Specimen 1. Moment transfer in the connection region was achieved by 

passing the beam longitudinal reinforcement through holes drilled in the RC columns, as for Specimen 

1, in combination with external steel plates (Fig. 3). The purpose of these external plates was to control 

the opening of gap at the beam-column interface, as well as to increase the beam moment strength in 

the region adjacent to the column, forcing the plastic hinge to form away from the column face. 

Connection between the FRC beam and the external reinforcement was achieved through a 1.25 in. 

steel rod that was embedded in the beam section (Fig. 3). Standard holes were drilled in the steel plates 

to connect them to the beam through-bolts. To decrease the moment capacity in the relocated plastic 

hinge region so as to maintain a shear demand comparable to that in Specimen 1, the beam longitudinal 

reinforcement was reduced to 2#5 and 2#4 top and bottom rebars. 

Specimen 3 

This specimen represented an exterior beam-column subassembly and used a moment connection 

consisting of internal beam rebars and external steel plates, as in Specimen 2. A beam stub was used in 

the opposite (back) side of the column for anchorage of the beam longitudinal reinforcing bars. 

Because the use of hooks at the bar ends would pose construction difficulties, the beam longitudinal 

bars were terminated with a mechanical anchor. In Specimen 3, 2 in. through-bolts were used for 

transfer of forces between the FRC beam and the external reinforcement. In addition, tight holes were 

used in the external steel plates to avoid any slip that would lead to "pinching" in the load vs. 

displacement hysteretic response (Fig. 3). With regard to beam transverse reinforcement, #3 stirrups 

were used only adjacent to the through-bolts and reinforcement mechanical anchors. No transverse

466 

reinforcement was used over the remaining part of the beam and beam longitudinal reinforcement was 

kept the same as for Specimen 2. 

Specimen 4 

Test Specimen 4 represented an exterior subassembly similar to Specimen 3. However, no reinforcing 

bars were used in the hybrid beam, and thus the steel reinforcement consisted only of a steel truss 

embedded in the FRC beam. Moment transfer in this specimen was achieved through the use of 

external steel rods. The use of a moment connection through only external reinforcement eliminates 

the need for drilling holes in the RC column, and thus the risk of cutting longitudinal and transverse 

reinforcement, which required wrapping of the column. A sketch of the connection details used to 

transfer the forces from the steel truss to the external rods is shown in Fig. 4. 

Material Properties 

Ready-mix concrete from a local supplier was used in all four specimens. For the FRC used in the 

beams, 1.2 in. long hooked steel fibers with a diameter of 0.02 in. were added at a volumetric fraction 

of 1%. Concrete compressive strength for the columns ranged between 4100-5600 psi, while that of the 

FRC ranged between 3500-4100 psi. Grade 60 steel was used for all reinforcing bars in the columns 

and beams. A3 6 steel was used for the steel truss members and Grade B7 bolts were used for the 

through-rods in Specimens 2 and 3, and the external rods in Specimen 4. 

EXPERIMENTAL RESULTS 

The behavior of the test specimens was evaluated in terms of their load vs. displacement hysteretic 

response, cracking pattern, beam rotations, and energy dissipation capacity. 

Load vs. Displacement Behavior and Cracking Pattern 

The load vs. displacement response for Specimens 1, 3 and 4 is shown in Fig. 5. Specimen 1, with 

moment connection relying on epoxy injection of the longitudinal beam bars passing through the 

column, experienced significant pinching due to slippage of the longitudinal beam bars. Yielding of 

these bars began at 1.0% drift and for story drifts up to 1.4%, pinching in the hysteretic loops was 

moderate. However, at larger drift levels, the amount of pinching increased severely due to an almost 

total loss of bond between the longitudinal beam bars and the column concrete, which led to the 

opening of a large gap at the beam-column interface. Flexural cracking in the FRC beams was 

observed at early stages of the test. However, concentrated rotations at the beam ends due to the 

opening of the gap dominated beam rotations, and thus beam flexural cracks did not open significantly 

during the cycles performed at larger drifts. In terms of lateral strength, Specimen 1 maintained its 

strength up to 4.5% story drift with little loss of stiffness during repeated cycles at the same drift level 

Specimen 3 5 which had the moment connection with combined external steel plates and epoxy 

injection of the longitudinal beam bars passing through the column, exhibited a stable hysteretic 

response with good stiffness retention and energy dissipation capacity, as shown in Fig. 5b. In this 

specimen, the use of external steel plates with tight holes connected to the beam through-bolts

467 

increased the beam moment strength in the region adjacent to the column face, and thus lead to 

spreading of the beam inelastic deformations away from the column face, as shown in Fig. 5c. In 

addition, the external steel plates controlled the opening of the gap at the beam-column interface, 

reducing pinching in the load vs. displacement hysteretic loops compared to Specimen 1 The strength 

of Specimen 3 was governed by the beam moment strength, which was increased due to the use of 

steel FRC. For all drift levels, no decay in strength was measured and only little decay hi stiffness 

during repeated cycles at the same drift level was observed. Specimen 2, which had the same details as 

Specimen 3 but with standard holes in the external steel plates for connection with the beam throughbolts, 

exhibited a response somewhat between those of Specimens 1 and 3 due to slip of the external 

plates that did not control the opening of the gap at the column face as effectively as the plates with 

tight holes used in Specimen 3. 

Specimen 4 had no longitudinal beam bars, and thus the steel reinforcement was provided by the 

embedded steel truss. The moment connection in this specimen was achieved entirely through external 

steel rods as shown in Fig 4. Specimen 4 showed a load vs. displacement response characterized by 

full hysteretic loops (Fig. 5d). The strength of this specimen was governed by yielding of the truss 

chords, which occurred in the region adjacent to the embedded steel tubes connecting the steel truss 

with the external steel rods. In Specimen 4, tlexural cracking concentrated at this location and at large 

drift levels a significant opening of a single crack was observed in the FRC beam. Severe local 

buckling of the truss chords was observed at 5% drift, leading to a significant decay in the specimen 

strength. 

Energy Dissipation Capacity 

The energy dissipation capacity exhibited by the test specimens was evaluated by comparing the area 

of the hysteretic loop corresponding to the last cycle at a particular drift level with that of an equivalent 

elasto-plastic system with a stiffness corresponding to the initial stiffness of the specimen and the 

strength measured during the first cycle at the story drift level of interest. Fig. 6 shows the energy ratio 

(specimen vs. equivalent elasto-plastic system) for the four test specimens. As can be seen, Specimen 1 

exhibited poor energy dissipation capacity with energy ratios below 12% for displacement levels larger 

than 1.0% drift. On the other hand, Specimen 4, with external steel rods, exhibited energy ratios as 

high as 47%. Specimens 2 and 3 exhibited energy ratios of approximately 15% and between 19-25% at 

story drifts larger than 2%, respectively. 

Beam Rotations 

Rotations in the FRC beams were measured with linear potentiometers over a length of 22 in. from the 

column face (1.1 x beam depth). In Specimen 1, beam rotations primarily concentrated at the beamcolumn 

interface due to significant slip of the beam longitudinal reinforcing bars. In Specimen 2, due 

to the use of standard holes in the external steel plates, the opening of the gap at the beam column 

interface was not adequately controlled, and thus beam rotations consisted of both concentrated 

rotations at the beam end and in the beam region adjacent to the column face. However, in Specimen 3 

with a combination of internal beam bars and external steel plates with tight holes, significant rotations 

were measured in the region adjacent to the through-bolts due to relocation of the plastic hinge region 

and an adequate control of the opening of the gap at the beam-column interface. In this specimen, 

rotations in excess of 4% were measured over approximately one depth from the column face. In 

Specimen 4, beam rotations concentrated primarily in the beam region connected to the external steel

468 

rods where the damage concentrated at large drift levels. In this specimen, rotations exceeding 6% 

were measured at the end of the test, as shown in Fig. 7. 

SUMMARY AND CONCLUSIONS 

Results from tests of four hybrid FRC-encased steel truss beam-RC column subassembhes are reported 

in this paper Three connection details were evaluated for moment transfer in the connection region. 1) 

use of beam longitudinal reinforcing bars passing through and epoxy-bonded to the RC column, 2) use 

of beam longitudinal reinforcing bars combined with external steel reinforcement, and 3) use of only 

external steel reinforcement for full moment transfer Experimental results indicate that connections 

consisting of beam longitudinal reinforcement in combination with external reinforcement, or only 

external steel rods, exhibit a stable response under large displacement reversals with good strength, 

stiffness and energy dissipation capacity The use of external reinforcement in the beam region 

adjacent to the RC column controlled the opening of a gap at the beam-column interface and allowed 

the relocation of the beam plastic hinge away from the column face. 

REFERENCES 

American Concrete Institute, ACL (1999). Acceptance criteria for moment frames based on structural 

testing. Report ACI1TG/T1 1-99 

FEMA, Federal Emergency Management Agency (2000). Prestandard and commentary for the seismic 

rehabilitation of buildings. FEiMA-356, Washington, B.C. 

Khuntia, M., and Goei, S. C. (1998). FRC-encased steel joist composite beams under reversed cyclic 

loading. ASCE Journal of Structural Engineering 124:10, 1115-1124. 

Fig. 1 - Test Setup

469 

3 

4@8 

~T / C 

4 #5 #3 stirrups 

86 

Beam Reinforcing Bars 

v Dlxl/2 

PL 17x7\i/4 

\ \ \A325 0=3/4' 

\ \ 14x3x3/8 

\ HILTIHSLM12/25 

k v ' 

. Dl/2

470 

" 

5 -i 3 - 2 - 1 

Story Drift (%) 

b) Specimen 3 

1 Z 3 4 S 

Story Drift {%) 

c) Cracking in Specimen 3 d) Specimen 4 

Fig. 5-Specimen Load vs. Displacement Response and Cracking Pattern 

Story Dnft {%) 

•Q 08 -0 OS -0.04 -0.02 0 0 02 Q 04 0 06 0 08 

Fig. 6 - Energy Ratio 

Fig. 7 - Load vs. Beam Rotation Response 

(Specimen 4)




SEISMIC RESPONSE OF MULTISTOREY MASONRY 

BUILDING WITH RESTRICTED BASE SLIDING 

M. Qamaruddin, S. Ahmad, H. Irtaza and S. M. Waseem 

Department of Civil Engineering, Z. H. College of Engineering and Technology, 

Aligarh Muslim University, Aligarh 202002, India 

ABSTRACT 

In the present investigation, a multistory masonry building system is considered in which a 

smoothened surface is created at the top of the substructure and the superstructure rests at this level 

and restricted sliding is permitted with fnctional resistance. The restricted sliding base system is 

idealized as a discrete mass mathematical model with two degree of freedom for computing its seismic 

response. The shear walls provide the spring action in the system. The total mass of the superstructure 

of the building, that is, the mass of all the stones, except the bottom half of the ground story mass, is 

lumped as top mass in the model and mass of the remaining bottom half portion of the ground story is 

lumped as bottom mass in the model. The bottom mass is assumed to rest on a plane with dry frictional 

damping to permit restricted sliding of the system. The seismic response of multistory masonry 

building with restricted sliding subjected to Koyna and El Centro earthquakes is computed employing 

this mathematical model treating the frictional resistance as rigid plastic. It turns out from this study 

that the restricted sliding base system is effective in reducing the seismic force acting on the building 

with low values of coefficient of friction. The seismic response of such a system is much reduced 

compared to the fixed base buildings with the same parameters. 


In past earthquake in India, unengineered masonry buildings were mostly damaged 

due to a variety of factors, such as. their short fundamental period and mostly lying in the dominant 

frequency range of strong earthquakes; very low tensile and moderate shear strength; some times not 

sufficiently high compressive strength; usually poor workmanship, etc. To improve the behavior of 

masonry structures under seismic forces, strengthening measures have been proposed. In spite of all 

these measures, masonry buildings can still be cracked during the earthquake. Such damage might be 

avoided if major part of the seismic energy is dissipated during earthquake One possibility to achieve 

this objective is by means of base isolation technique, which consists of decoupling the structures from 

the damaging effect of the earthquake. In the last few decades, various researchers have proposed 

varieties of isolation devices. 

In several major earthquake occurrences in developing countries like India and China, a beneficial 

behavior of the low-rise buildings which could slide as rigid bodies over their foundations was 

observed during some past severe earthquakes, viz., the Dhubri Earthquake in Assam in 1930 [2], and 

in Bihar-Nepal Earthquake in 1934. The damage study by Gee [2] and the observations made in China 

by LiLi [3], showed that those buildings in which the possibility of movement existed between the 

superstructure and the substructure suffered less damage than those buildings in which no such 

freedom of movements existed. Arya, Chandra and Qamaruddin [1], LiLi [3] and Qamaruddin [5] have

472 

investigated a base isolation system in which the isolation mechanism is purely sliding friction. Such 

system utilizes pure friction to allow some parts of a structure to slide relative to the others. 

A mathematical model was introduced by Qamaruddin [5] and Arya et al. [1] to compute the seismic 

response of masonry building with friction base isolation. A new concept has been proposed for the 

construction of brick building in which a clear smoothened surface is created just above the dampproof 

course at plinth level without any mortar, and the superstructure simply rests at this level and is 

free to slide except for frictional resistance. The concept of such system was further strengthened by 

the damage studies made by Li Li[3] after the Xintai (1966), Bohai(1969) and Tangstan (1976) 

earthquakes in which it was found that adobe buildings which were free to slide on their foundations 

(by accident) survived with little or no damage whereas others which were tied on their foundations 

collapsed. Researchers have made experimental and theoretical studies to incorporate such a system in 

masonry buildings economically to achieve a collapse free if not a damage free performance during the 

earthquakes. Further studies have also been made by Mostaghel et al. [4] and Qamaruddin et al. [6 and 

7] with encouraging results. 

2. MATHEMATICAL IDEALIZATION 

2.1 Mathematical Model 

A two-degree spring mass model shown in Fig. 2.1 represents the concept of multistory building with 

restricted base sliding. A layer of suitable material of known coefficient of friction is laid between the 

contact surfaces of the bond beam of the superstructure and plinth band of the substructure. The spring 

action in the system is assumed to be provided by the shear walls. Internal damping is represented by a 

dashpot that is parallel with the spring. The total mass of the superstructure of a multistory building, 

i.e., the mass of all the stories, except the bottom half of the ground story is lumped at the top of the 

system and the mass of the remaining bottom half portion of the ground story is lumped at the level of 

the band beam. The lower mass is assumed to rest on a plane with dry frictional damping to permit 

sliding of the system. Restricted base sliding can occur at the contact surface without overturning or 

tilting. The building is subjected to only one horizontal component of ground shaking at a time, the 

effect of vertical ground motion is not considered here. The stopper is considered as rigid. 

2.2 Equations of Motions 

Phase I: Initially, bottom mass moves with the base so long as sliding force does not overcome the 

frictional resistance. So the building behaves as a single degree of freedom system and therefore 

equation of motion is: 

M t x t +C s (Z t -Z b )+K s (Z t -Z b ) = 0 (2.1) 

orZ t -r2o)^(Z t -Z b )4-03 2 (Z t -Z b ) =-y(t) (2.2) 

where, C s = coefficient of the viscous damper; K s = spring constant; M t = top mass; X{, Z{ = 

absolute and relative acceleration of the top mass; y(t) = ground acceleration at time t; Z b, Z t = 

lateral relative displacement of masses M b and M b respectively; M b - bottom mass; Z b , Zt = 

relative velocity of masses M b and M t, respectively; co = natural frequency and £ = fraction of critical 

damping.

473 

Phase 2: The sliding of the bottom mass begins when the sliding force overcomes the frictional 

resistance at the plinth level. The force to cause sliding S t, is given by 

S f = C s (Z t -Z b ) + K s (Z t -Z b )-M b X b (2.3) 

Sliding of bottom mass occurs if | Sf | > JJL Mjg (2.4) 

where, g = acceleration due to gravity; MT=Mb + M t and}x = friction coefficient. In this phase, the 

building acts as two degree of freedom system shown by the following simplified equations of motion: 

F =-y(t) (2.5) 

Z, +2 2 (Z, -Z h ) = -X') (2.6) 

where, F = |ig(l + 0)Sin(Z5); Z b = relative acceleration of bottom mass; Sin(Z b ) =+1 if (Z b ) 

is positive; Sin(Z b ) = -1 it (Zj-,) is negative and 0 = —— = mass ratio 

M b 

Phase 3: If the stopper is positioned in such a way that the peak displacement of the system is more, 

then the system will strike the stopper at any time. Therefore, the motion of the system will be reversed 

if the frictional resistance at the plinth level is overcome by the sliding force (S'f ) at the time of strike. 

This force is given by: 

n 

M t 

X b /p (2.7) 

where, p=l/(2+0). The sliding of the bottom mass occurs backward if | S'f | > ^ M T g and the 

equations of motion are given as follows: 

Z b -2coêp(Z t -Z b ) - (B 2 9|3(Z t -Z b ) + Fp = -y(t) (2.8) 

Z t +2o^(Z t -Z b ) + 03 2 (Z t -Z b ) = -y(t) (2.9) 

Phase 4: At any time during the motion of the system if | Sf | or | S'f j < JJL Mxg, then the sliding of 

the bottom mass is stopped but the top mass continues to vibrate. Therefore, the system will behave as 

single degree of freedom and its equation of motion is same as given in phase 1. 

2.3 Solution of Equations of Motion 

The equations of motion for the different phases are solved by Runga-Kutta fourth order method for 

obtaining the complete seismic response. A computer program has been developed to compute the 

time-wise earthquake response of multistory masonry building with restricted sliding base system. 

3. PARAMETRIC STUDY 

The parameters that are mainly considered in the study include the time period, mass ratio, dry 

coefficient of friction and viscous damping for estimating realistic forces and displacements of 

multistory building with restricted sliding type system. The response has been computed for two 

actually recorded severe earthquakes viz., Koyna Earthquake (India) of December II, 1967 

(longitudinal component) and El Centra Earthquake (USA) of May 18, 1940 (N-S component). The

474 

quantities of interest are: (1) absolute acceleration which determines the forces acting on the shear 

walls; (2) the maximum relative displacement of the superstructure (in case of free sliding); and (3) the 

residual relative displacement which will indicate the position of the superstructure at the end of the 

ground motion. The following range of values of different parameters has been estimated, that would 

cover a wide variety of multistory masonry buildings: 

TABLE 3.1 

DATA FOR COMPUTING SEISMIC RESPONSE 

Time Period 

(TP) sec. 

0.40 

0.50 

0.60 

0.70 

0.80 

Mass Ratio 

(9) 

6 to 8 

8 to 10 

10 to 12 

12 to 14 

14 to 16 

Damping 

Coefficient 

(f) 

0.05 to 0.15 

0.05 to 0.15 

0.05 to 0.15 

0.05 to 0.15 

0.05 to 0.15 

Coefficient 

of friction 

(M.) 

0.10 to 0.25 

0.10 to 0.25 

0.10 to 0.25 

0.10 to 0.25 

0,10 to 0.25 

Restricted 

Sliding base 

range (A) mm 

2 to 130 

3 to 140 

2 to 150 

9 to 145 

2 to 165 

It is assumed that a coefficient of friction less than 0.10 in sliding will be difficult to obtain in actual 

building construction, and for a value greater than 0.25, no sliding motion may occur in most real 

earthquakes and the system may act just like a fixed-base structure. For four-story building, top three 

and half portion of the total mass of the building is assumed to be lumped as the top mass in the 

mathematical model. In the same way, as the story of the building increases, the mass ratio increases. 

In estimating the mass ratio, it is assumed that mass in each story is same. 

4. DISCUSSION OF RESULTS 

The influence of various parameters on maximum response of the free sliding (without stopper), 

restricted base sliding (with stopper) and fixed base multistory masonry buildings subjected to Koyna 

and El Centra shocks is presented in the present investigation through representative Figures 4.1 to 4.6. 

A comparison of the absolute acceleration response in the restricted base sliding has been made with 

that of the free sliding and fixed base buildings. The results of this study are discussed in the following 

paragraphs. 

Figs. 4.3 to 4.6 shows the typical acceleration response of the masonry buildings subjected to Koyna 

and El Centra earthquakes for different parameters. In these figures, the upper dashed line shows the 

maximum acceleration response in the fixed base structure, whereas the bottom dashed lines show the 

maximum acceleration developed in the building with free sliding base system during the seismic 

ground motion at different mass ratios. The firm lines show the maximum acceleration response 

computed in the buildings with the restricted base sliding system during the earthquakes at different 

mass ratios. 

4.1 Influence of Viscous Damping 

The representative acceleration response curves show that at a particular mass ratio, time period and 

coefficient of friction in the case of restricted base sliding buildings, an increase in viscous damping 

(from 5% to 15%) decreases the acceleration response (Figs. 4.1 and 4.2). This is perhaps because of 

increasing energy dissipation in the internal friction of the system as the damping coefficient increases. 

In view of this well-established trend, the acceleration response for other cases has not been studied.

475 

4.2 Effect of Coefficient of Friction 

It is seen from Figs. 4.3 to 4.6 that for a particular time period and coefficient of critical damping, as 

the coefficient of friction increases, the peak displacement decreases in the case of building with free 

sliding base. This is logical, since the resistance against sliding of the system decreases as the 

coefficient of friction between the sliding surface decreases and the buildup of a large inertia force in 

the superstructure becomes restricted. Thus, the spectral acceleration of such a system decreases. 

Similar trend is generally true in the buildings with restricted base sliding system. The slight departure 

in this trend observed in few cases is due to the fact that positions of stopper are different in different 

building parametric cases. 

4.3 Time Period Effect 

It is observed from Figs. 4.3 to 4.6 that for both the El Centre and Koyna shocks, the acceleration of 

the restricted base sliding buildings -decreases in general as the time period increases for a particular 

viscous damping coefficient and coefficient of friction. For same building systems, the acceleration 

values are more when they are subjected to Koyna shock than that in the case of El Centre shock. But, 

the acceleration response of the free sliding system is independent of its time period for different 

values of coefficient of friction. 

4.4 Influence of Stopper's Position 

It is seen from Figs. 4.3 to 4.6 that if the stopper is placed close to the building (before its sliding), then 

its acceleration approaches to the corresponding values as obtained in the case of fixed base. But, if the 

stopper is placed at a distance approximately equal to the peak displacement of the free sliding system, 

far away from the building, then its acceleration is very near to that of the corresponding values of the 

free sliding system. This trend is observed in all the building systems subjected to both the earthquakes 

(Koyna and El Centro). 

Table 3.1 shows the position of the stopper for different time periods of the system ranging from 0.4 to 

0.8 seconds subjected to Koyna and El Centro shocks. These are the most feasible ranges of stopper 

positions for different mass ratios. This study shows that if the stopper is placed anywhere in this 

range, then the absolute acceleration of the top mass of the system is much less than the corresponding 

fixed base system. The occupants of the multistory buildings will have psychological comfort by 

providing stopper in the sliding base system. 

4.5 Effect of Mass Ratio 

From Figs, 4.3 to 4.6, it may be observed that as the mass ratio increases, generally, the acceleration 

response decreases in all the cases of parametric combinations for a particular time period and 

damping coefficient in the case of free sliding base. But, generally, no definite pattern of acceleration 

response variation has been observed in the case of buildings with restricted base sliding with varying 

values of the mass ratio. 

5. CONCLUSIONS 

The following conclusions are drawn from the seismic response study of multistory masonry building 

with restricted base sliding system:

F b 1 MrtTHEMAHCAL MODEL FDR MULT I 

Mrt ^Dt P BUILDING WITH RL-TRILTED BAGE 

The restricted sliding base svstem is effective in reducing the effective seismic force acting on the 

multistory masonry building with low value of coefficient of friction The restricted sliding base 

system will provide psychological comfort to the occupants of the multistory buildings 

6 REFERENCES 

(1) Arva, A S , Chandra B and Qamaruddin, M (1981) A new concept for resistance of masonry 

buildings in severe earthquake shocks, Journal of the Institution of Engineers (I), 61 PtC16, 302 - 308 

(2) Gee E R (1934) Dhubn earthquake-1930, Memories of geological survey of India, LXV, 1 - lOo 

(3) Li Li (1984) Base isolation measures for a seismic in China, Eight World Conference on 

Earthquake Engineering San Francisco California, IV,791 -798 

(4) Vlostaghel N , Hejazi, M and Tanbakuchi, J (1983) Response of sliding structures to harmonic 

motion Journal of Earthquake Engineering and Structural Dynamics, 11, 355 - 366 

(5) Qarnaruddm, M (1978) Development of brick building systems for improved earthquake 

performance', Ph D Thesis University ofRoorkee Roorkee 

(6) Qamaruddm, M Arva, A S and Chandra, B (1986) Seismic response of brick buildings with 

sliding substructure, Journal of Structural Engineering ASCE, 122 3, 558 - 572 

(7) Qamaruddm M, Rasheeduzzafur, Arya, A S and Chandra, B (1986) Seismic response of 

masonry buildings with sliding substructure, Journal of Structural Engineering ASCE 122 9 2001 - 

2011

477 

-ZETA=Q 05 

-ZETA=010 

-ZETA=015 

03 

II 

015 4 

70 80 90 100 110 

POSITION OF STOPPER (mm) 

FIG 4 1 VARIATION OF ABSOLUTE ACCELERATION WITH 

DIFFERENT POSITION OF STOPPER FOR ELCENTRO SHOCK 

(TP=040MR=7p=01) 

03 

028 

CO 

2 026 

O P 024 

!u 022 

02 

-ZETA=0 05 

-ZETA=01Q 

-ZETA-015 

a 016 

S 014 

03 

478 

1 

c095 

^09- 

§085 

j= 03 

go 75 

3 ° 7 1 

goes 

o as 

479 

031 

en 03 

0029 

P 

|Q 28 

~*—MR=6(WS) 1 

-»—MR=7(WS) 

-*—MR=8(WS) 

FIXED BASE 

MR=6(WOS) 

MR=7(WOS) 

MR=8(WOS) 

3027- 

2o26 

I- 

^025 

u) 

SO 24 

023 

7 8 9 

POSmON OF STOPPER (mm) 

FIG 4.5. VARIATION OF ABSOLUTE ACCELERATION WITH DIFFERENT 

POSITIONS OF STOPPER FOR EL CENTRO SHOCK 

01 

Z 

9


Advances and New Challenges in Earthquake ° 


SLIDING MODE SEMI-ACTIVE BASE ISOLATION USING THE 

SWITCHING HYPERPLANE DESIGNED BY DISTURBANCE- 

ACCOMMODATING BILINEAR CONTROL 

Norihiro SANTO, Kazuo YOSHIDA 

l Keio University, Deparment of System Design Engineering, 3-14-1, 

Hiyoshi, Kouhoku-kn, Yokohama, Japan 

2 Keio University, Deparment of System Design Engineering, 3-14-1, 

Hiyoshi, Kouhoku-ku, Yokohama, Japan 

ABSTRACT 

In vibration isolation control problem, the semi-active control methods in which coefficient of viscous 

damping is varied effectively have been proposed and expected to realize a high performance. However, the 

semi-active system is classified into a bilinear system which belongs to a nonlinear system. In this study, the 

switching hyperplane of sliding mode control is designed by the disturbance-accommodating bilinear optimal 

control theory, and it is applied to a semi-active base isolation problem. The proposed method adopts the 

damping force in criteria function, and the controller takes into account the disturbance dynamics by difining an 

augemented bilinear system. This study deals with a ten-degree-of-freedom structural model with base isolation. 

And the computer simulation is carried out by taking account of the delay of switching the coefficient of 

dampers. As a result, the usefulness of the present method was demonstrated. 

INTRODUCTION 

Recently the technologies on active or semi-active vibration control methods have widely spread and some 

of them have been put into pracice.tn the field of building structure, a semi-active vibration control of high-rise 

building was put into practice 1979 for the first time, and then an active vibration control of pencil building was 

realized in 1989 for the first time. These vibration control systems belong to a type of dynamic vibration 

absorber which is installed at the top ofbuilding. The number of buildings with active vibration absorber 

exceeded 20. Since the Kobe earthquake occurred in 1995, passive base isolation systems with lead damper, 

high damping rubber and so on have been used as a damping element. These passive dampers are designed 

for the large seismic wave. Therefore, the base isolation system is not necessarily effective resonance transmissibility, 

while it raises the transmissibility in the higher frequency range. 

Semi-active control method which is a method to control the parameters of control device such as spring

482 

stiffness or coefficient of viscous damping, has recently drawn more attention [Kamopp, D., 1995, Dyke, S. 

J., Spencer, Jr., B. R, 1997, Yoshida. S. and Yoshida. K., 1997, Fujio. T. and Yoshida. K., 1999] than the 

active control methodology which needs much power. Semi-active control is characteristic of less power then 

active control and higher performance than passive control. Especially, variable dampers have already been 

put into practice in the field of automobile, and a skyhook damper is approximately realized by using variable 

damper. Since the suppression of actuator power is an important issue for active structural control, in this 

stud\, a new semi-active control method for vibration base isolation is presented by a sliding mode using the 

switching hyperplane designed by disturbance-accommodating bilinear control. 

One of the authors proposed an bilinear potimal control with feedforward link for a semi-active dynamic 

vibration absorber and then applied it to a base isolation problem [Fujio. T. and Yoshida. K., 1999]. This 

study presents a sliding mode controller using the switching hyperplane designed by disturbance-accommodating 

bilinear control. In order to investigate the performance and the effectiveness of the proposed method 

by numerical calculation and computer simulation, it is applied to a model of an actual structure with 10 degree-of-freedom. 

CONTROLLER DSEIGN 

Fundamental Model for Controller Design 

hi this study, a ten-storied structural model with semi-active base isolator subjected to seismic excitation is 

treated as a control object A variable damper is installed between the base and the first story of structure. As a 

result of the eigenvalueanalysis, first natural frequency of system is 0.25Hz and damping ratio is 2%, respectively. 

Then this control object is represented by the following bilinear equation. 

V" ;r Y" •' • A TC TC V • • • T I 

l A 0 A 10 "hj 

"S L A 10 

where .r is state variable vector and z is disturbance vector, u denotes the vector of the control input, that is, the 

damping coefficient of damper given by the semi-active control. 

Controller with Feedforwrad Link 

In this stud}, it is assumed that there is a disturbance with flat spectral density at the controlled frequency 

range in order to adapt to various earthquakes. Accordingly, by restricting the frequency range of disturbance, 

we assume that the characteristic of power sepectal density is the same as white noise within the restricted 

frequency range. Then, we assume that the disturbance is such a colored noise that has the transfer function of the 

characteristic of a low pass filter as shown in Fig. 1, whose transfer function is give by

483 

li£l = 2 

W(s) r+-2O 

(2) 

In this case, the state equation of the disturbance is written by 

(3) 

where 

o i o 

o o i 

0 -co. 1 -If 

> A* = 

(4) 

Utilizing Eqs.( 1) and (3), we have the following augmented state equation 

B = #, (5) 

Switching Hyperplane Designed by Disturbance-Accommodating Bilinear Control 

In this chapter, by appling the bilinear optimal control theory to the agumented system, we derive the 

switching hyperplane. Generally, such semi-active control as parameter control is classified into a bilinear 

system which belongs to non-linear systems. It is considered that that design restriction in semi-active damper 

is a damping power rather than a damping coefficient. From this viewpoint, the maximum damping power of 

damper design is the designed parameter. In this study, term of the damping power is introduced to the 

criterion function. 

We consider the following control strategy to minimize the criterion 

+« T (0^TW^(0^(0«(0]*' 

(6) 

where Q and R are weighting functions. And the optimal control input U Q which satisfies the following equation 

is synthesized 

J r (w°) = minJ r (n) (7) 

Then the dynamic programming method is applied to obtain the optimal control input u°. This problem is

484 

solved for the Hamilton Jacobi-Bellman equation of the optimality condition, and the switching hyperplane is 

expressed by the following equation. 

S = F = X m lty'B T usm P (8) 

where P in (8 ) denotes the positive semi definite matrix which satisfies the following equations. 

(9) 

Sliding Mode Controller Design by Disturbance-Accommodating 

Here, we obtain the continuous sliding mode controller based on the switching hyperplane by disturbanceaccommodating 

bilinear control. The block diagram of the proposed control system is shown in Fig. 2. The 

control law consits of the feedback link and the feedforward link. 

This control object is represented by the following bilinear equation. 

We assume an sliding mode is defined in n-dimensional state space as follows : 

(10) 

485 

where 8 t (i = 1 , 2 , . .. , n] denote the sliding margins. These controllers will have the state move from an 

arbitrary initial state to the sliding mode and then move along the user-specified sliding mode to origin. 

And the control input at boundary layer is assumed as follows equation. 

u c - -Kx = -(A^ + & 2 jc 2 + .. . + k n x n ) (13) 

+,!-/; xinf a, -£,) £ f >o 

The boundary layer is a hypothetical band centred on the sliding mode cr = 0 , The width of the boundary layer, 

A represents the tolerance or allowable deviation from the ideal sliding motion due to parameter uncertainty and 

desturbances. The contol input where the boundary layer equivalence is considered can be reprezented by 

where (f> represents parameter uncertainty and other disturabances. w is the classical sliding mode given by 

disturbance-accommodating. 

In the above treatment, the semi-active force is supposed to be realized by variable dampers. Actually, the 

damping coefficients should be positive, and remain within a certain range for practical use. Therefore, the 

damping coefficients of semi-active dampers are limited by the inequality. By supposing that the control force 

generated by the semi-active dampers under the above constraint is as close as possible to the control force of 

active control, the damping coefficient of semi-active damper is expressed as follows: 

C = 

NUMERICAL CALCULATION 

In order to investigate the feasibility and the effectiveness of the semi-active control using variable dampers 

which is designed by the method proposed in this study, the numerical calculation was carried out. 

And the control characteristics of the proposed semi-active base isolation control are examined in both time 

and frequency domains. Since the semi-active control is essentially nonlinear, the vibration transmissibilityunder 

semi-active control is obtained as a ratio in frequency range of the spectral densities of input and output. And a 

white noise is used as an input. 

For the semi-active contol based upon the disturbance accommodating sliding mode using the bilinear optimal 

switching hyperplane, an observer used VSS observer was, since the measured physical values are sup-

486 

posed to be the accelerations of the base and the top floor of the structure and the relative displacement between 

the base and the first story of the structure. 

The specification of the 10 degree-of -feedom structure model shown in Fig. 3 are shown in Table. 1. The no 

isolation case means the case that its first damping ratio is 5%. Then the passive case 1 measns no control, and 

the passive case 2 means the case where its first damping ratio is much higher than the passive 1, the passive case 

3 measns on isolation with lead rubber bearing, the skyhook case means that skyhook control method is adopted 

as a semi-active control The disturbance-accommodating sliding mode control (DA-SMC) case and skyhook 

cace deal with the case of installing variable viscous damper between the base and the first story. The frequency 

range is assumed to be less than 7 Hz. 

Numerical calculation was carried out for the case that the absolute velocity is adopted as an objective 

function as follows: 

O lx , 1 0 0] 

(16) 

And in order to consider the dynamics of oil damper, Maxwell model is introduced in the numerical calculation. 

Figs. 4-8 and Table.2 show the time history of the top floor acceleration and the first floor dispacement 

under El Centro earthquake,and so on. From these figures the effectiveness of the DA-SMC was demonstrated. 

And it was seen that the DA-SMC can reduce the maximum response of acceleraion at the top floor more than 

the passive control and also reduces the residual vibration which appears in no-control. 

Figure 9 shows the numerical results of the vibration transrnissibility for respective controllers. In the case of 

the skyhook control the performance at the frequency range is affected, while DA-SMC is effective at the whole 

frequency range. 

Table, 1 Parameters of the sturucture 

Fig. 3 10-DOF structure model possessing an semiactive 

base isolator 

Layer 

1 

2 

3 

4 

5 

6 

1 

8 

9 

10 

Mass [t] 

4981.40 

3438.20 

2490.60 

1826.40 

2033.10 

2050.00 

2036.90 

2037.10 

2066.40 

2499.90 

Spring constant 

ft/cm] 

68.20 

2320.00 

2820.00 

2020.00 

1840.00 

1850.00 

1600.00 

1410.00 

1180.00 

1020.00 

Coefficient of 

viscous damping 

[t s/cm] 

0.90 

12.94 

1971 

13.34 

11.15 

11.05 

10.55 

10.15 

8.48 

8.71

487 

| 05, 

I 

20 25 30 

(a) Passive 3 (LRJB) 

Time fsl 

(bj Skyhook 

(c) DA SMC 

Fig. 4 Acceleration of the 10th floor for the El Centro earthquake 

04| 

-, 031- 

— Passive 2 (NRB+POD) j 

°-03 

-04J 

0 S 10 15 20 25 30 

Traefs] 

S° 

S oi 

J

488 

vlaxacccleiation [my 

O O O O —_.-._. 

No Isolaion 

Passive 1 (NRB) 

Passive 2 (NRB+POD) 

Skyhook 

DA-SMC 

Passive 1 Passive 2 Passive 3 Skyhook DA-SMC 

(NRB) (NRB+-POD) (LRB) 

Fig. 8 The max acceleration comparison of 10th 

foor for the El Centre eathquakes 

Fig. 9 Vibration transmissibihty 

CONCLUSIONS 

In this study, the disturbance-accommodating sliding mode controller of variable damper was presented and 

it was applied to the building structure with base isolation. As a result, the conclusions were obtained as 

follows. 

(1) The semi-active damper has higher control performance on the vibration transmissibility than the passive 

control and skyhook control at a wide frequency range. 

(2) For El Centra earthquake as well as the other earthquakes, the present method reduces residual vibration 

which appears in the uncontrolled case and also reduces the maximum response of the first story of the structure 

in comparison with the passive control whose first mode damping ratio is the same as the semi-active 

control 

From to these result, the capability of D A-SMC for semi-active vibration isolation system is demonstrated 

REFERENCES 

Yoshida. S. and Yoshida. K.( 1997). Semi-Active Vibration Isolation Control Using Feedforward Information of 

Disturbance. Proc ofASm Design Eng. and Tec. Conf. (CD-ROM), Paper No. DETC97/VIP-3820. 

Dyke, S. J., Spencer, Jr., B. F. (1997). A Comparion of Semi-Active Control Strategies for the MR Damper. 

Proceedings of the IASTED International Conference, Intelligent Information Systems. 

Singh, M. P., Matheu, E. E.(1997). Active and Semi-Active Control of Structures Under Seismic 

Excitation. Earthquake Eng Sturuct Dyn.26, 193-213. 

Karaopp, D, (1997). Active and Semi-Active Vibration Isolation. Trans, of the ASMS, Special Anniversary 

Deign Issue, 117,177-185. 

Yoshida. K. and Fujio. T.(1999). Semi-Active Base Isolation for a Building Structure, Proceedings of ASME 

Design Engineering Technical Conference(CD-ROM), Paper No. DETC99/MOVIC-8427. 

Kang, S. and Yoshida, K. (1993). Vibration Isolation Control with Feedforward Link using ETCntrol Theory. 

Proceedings of 'the JSME International conference on Advanced Mechatoronics, 645-649.




EVALUATION OF VIBRATION CONTROL EFFECT 

ON EQUIPMENT USING LOW-STIFFNESS AND 

HIGH-DAiMPING RUBBER 

1 

HIGUCHI Shoko ', OKUTA Kenji 2 , DOHI Hiroshi ' 

OZAWA Teruhiko ! and TSUJII Yasuhito 3 

Research and Development Headquarters, NTT Facilities, Inc., 

NTT Building Technology Institute, 

Architectural Management Headquarters, NTT Facilities Inc., 

Tokyo, JAPAN 

ABSTRACT 

The vibration control device of low-stiffness and high-damping laminated rubber has been 

developed for the anti-seismic measurement for the information equipment. This vibration 

control device can reduce the effect of seismic acceleration against the equipment by installing 

between the bottom of the equipment and the floor. Thus, the equipment can be prevented from 

falling or being damaged by the seismic motion. This paper evaluates this vibration control 

effect on the equipment by using the laminated rubber against seismic motion based on the 

shaking table test's results. 

Center of gravity 

\ steel plate tf=25VV 

Fig. I. 

OVERVIEW OF THE TESTING MODEL 

Side view

490 

INTRODUCTION 

In recent years, the large earthquakes have frequently occurred near the urban area and brought 

about enormous damage to not only building structures but also lifeline system. Especially the 

damage to information equipment has great impact on city functions with the advance of 

information society. Therefore it is important to take effective measures to keep the safety of 

the equipment against the earthquakes. 

In the previous paper *\ the excellent vibration control effect on the equipment by the device of 

high-stiffness and high-damping rubber had been shown where the equipment was installed on 

the upper floor of information buildings. In this study, the vibration control device of 

low-stiffness and high-damping rubber has been considered through the shaking table tests 

using the artificial wave made from the lower floor response of buildings against earthquakes 

with seismic intensity scale MM8. 

VIBRATION CONTROL DEVICE 

The vibration control device is consisted of pieces of low-stiffness and high-damping laminated 

rubber sized to match the equipment and is attached to the bottom of the equipment. This 

rubber was designed to reduce the acceleration response at the time of an earthquake by 

absorbing the vibration energy. 

The characteristics of the equipment and level of the earthquake motion used in this shaking 

table test are shown in Table 1. The characteristics of each low-stiffness and high-damping 

rubber are shown in Table 2. Two kinds of rubber with different damping ratio were used as A 

and B. Their dynamic characteristics are described in Fig. 2. The equivalent stiffness is almost 

the same for type A and B. The equivalent damping factor of type A is larger than type B. The 

vibration control device, the equipment system, and the raised-floor were set up on the 

three-dimensional shaking table as in Fig. 1. 

TABLE. 1. 

CHARACTERISTICS OF THE EQUIPMENT AND TEST CASE 

Test Case 

Equipment size(mm) 

Equipment own weight (kg) 

Dummy weight (kg) 

Equipment Total weight (kg) 

Type of rubber 

X direction 

Natural Period (sec) 

Y direction 

Horizontal (m/sec 2 ) 

Earthquake motion level 

Vertical (m/sec 2 ) 

CaseN Case A CaseB 

_ 

0.23 

0.16 

700X900X2000 

190.5 

228.5 

419.0 

Type A 

0,24 

0.17 

4-6-8 

2-3-4 

TypeB 

0.25 

0.17

491 

SHAKING TABLE TEST 

Seismic Wave 

The seismograph data has been recorded by seismograph observatory for constant. These 

seismic records on the building floor obtained through the observations are dealt. Based on 

these records, the seismic waves were made artificially to have duration of 40 seconds and to 

have a floor acceleration response spectrum of medium-rise and low-rise buildings. The phase 

characteristic of the wave is provided by random value. 

The seismic waves are applied simultaneously in two horizontal directions and the vertical 

direction. As an example, the seismic wave-form applied in the shaking table test is shown in 

Fig. 3, and the floor acceleration response spectra for the horizontal wave and the vertical wave 

are shown in Fig. 4 (in the case of the floor acceleration of 6 m/s 2 ). 

-120-80-40 0 40 80 120 

Shear Strain (%) 

-120-80-40 0 40 80 120 

Shear Strain (%) 

Fig. 2. 

RESTORING FORCE CHARACTERISTIC RELATIONSHIP OF HIGH-DAMPING 

RUBBER (HORIZONTAL) 

TABLE. 2. 

CHARACTERISTIC OF LOW-STIFFNESS AND HIGH-DAMPING RUBBER 

Type of Rubber 

Type A 

Type B 

Size 

Equivalent Damping factor (%) 

Equivalent shear stiffness (N/cm) 

Diameter (mm) 

Height (mm) 

Shear strain=20% 






18 

24 

26 

1021 

590 

429 

74 

42 

8 

17 

21 

1000 

620 

452

492 

Since earthquake motion has been amplified in the buildings structure when it affects the 

equipment, the earthquake motion put in the equipment is estimated as the maximum 

acceleration level; 4 m/s 2 (AT04), 6 m/s 2 (AT06), and 8 m/s 2 (AT08) horizontally and vertical 

acceleration is half level of horizontal one. When the estimation is 8 m/s 2 horizontally and 4 

m/s 2 vertically, seismic intensity scale is MM8. Using this artificial earthquake motion as the 

one affecting the equipment, the authors made an investigation thorough testing in the 

following chapters. 

Shaking Table Test 

A shaking table test has been conducted using a three-dimensional shaking table with an actual 

equipment system set up on it together with the raised floor, in order to evaluate earthquake 

response characteristics of the equipment and vibration control effect achieved by the 

low-stiffness and high-damping rubber. Dummy weights were placed evenly on eight shelves of 

the equipment. In case A and B, the rubber type A and B are used. In case N, no rubber is used. 

20 

(1) Horizontal (2) Vertical 

Fig. 3 

ACCELERATION WAVE OF ARTIFICIAL EARTHQUAKE MOTION (10m/sec 2 ) 

S 15 Period (sec) 

0.1 1 

Period (sec) 

Fig. 4 

ACCELERATION RESPONSE SPECTRA OF ARTIFICIAL EARTHQUAKE MOTION 

(6m/sec 2 : AT06)

493 

TEST RESULTS 

Response Acceleration 

Fig. 5 shows acceleration response magnification at different points in the equipment, top, 

middle and bottom. The figure shows that installing vibration control device reduced response 

acceleration at the top of the equipment 13% in the X direction and 15% in the Y direction at 

AT04, and 21% in the X direction and 29% in the Y direction at AT06. At AT08, 36% reduction 

in response acceleration was found in the Y direction, but no vibration control effect was 

Top 

o-.. AT04-N ' 

A- - - AT06-N 

o--- AT08-N 

AT04-A 

AT06-A 

AT08-A 

AT04-B 

AT06-B 

AT08-B 

1 2 3 4 

Maximum Acceleration Response 

Magnification(X dir) 

1 2 3 4 

Maximum Acceleration Response 

Magnification(Y-dir) 

Fig. 5. 

MAXIMUM ACCELERATION RESPONSE MAGNIFICATION 

FOR EACH PART OF THE EQUIPMENT 

, 4 6 

Input Acceleration (m/sec 2 ) 

Fig. 6. 

SHEAR STRAIN LOW-STIFFNESS AND HIGH-DAMPING RUBBER

494 

confirmed in the X direction. Because the equipment had a small width in the X direction, 

rocking occurred more outstanding in the X direction than in the Y direction. 

The acceleration response magnification at the center of gravity was about 2.5 in case N, and 

about 2.0 in case A and case B. The acceleration response spectra in Fig. 4 indicate that the 

damping factor was about 5% in case N, and about 7% in case A and case B. 

Characteristic Difference In Two Types Low-Stiffness 

and High-Damping Rubber 

Fig. 6 shows shear strain of the low-stiffness and high-damping rubber at every level of seismic 

test. Shear strain increased as greater ground motion was applied in case A and B. Shear strain 

of type B was greater than type A. The shaking table test found no significant difference in 

vibration control effect of type A and type B. 

Physical Damage To The Equipment 

Physical damage to the equipment is discussed below. In case N, the welding, the major 

structural elements of the equipment, suffered damage, like cracking, at AT08. As a result, the 

acceleration response magnification at the top of the equipment at AT08 was lower than at 

AT06 (see Fig. 5). In case A and case B, neither permanent deformation of major structural 

elements nor cracking in the welding occurred at AT08. Thus, vibration control effect was 

confirmed. 

CONCLUSIONS 

The application of vibration control device that reduces horizontal seismic force, which affects 

the equipment, has been developed. Inserting low-stiffness and high-damping rubber between 

the bottom of the equipment and the floor could attain the reduction of the seismic force 

considered the lower floor response of buildings against earthquakes. Through the shaking 

table test, the following conclusions were obtained. 

(1) Installing low-stiffness and high-damping rubber between the bottom of the equipment and 

the floor could reduce response acceleration of seismic motion 4 m/s 2 and 6 m/s 2 by 13% 

to 29%. Damage to the equipment could be prevented against seismic motion 8 m/s 2 . 

(2) Vibration control will become more effective even at higher levels of input seismic motions 

by considering the way to reduce the rocking of the equipment. 

Reference 

INABA, T., DOHI, H., OKUTA, K., and SATO, T. (1998). Vibration Control of Equipment Using High 

Damping Rubber. Second World Conference on Structural Control, ABSTRACTS, JUNE, 28-JULY1, 170.




FRACTIONAL DERIVATIVE VERSUS GENERAL LINEAR 

MODELS OF VISCOELASTIC DAMPERS FOR 

SEISMIC ANALYSIS 

M. P. Singh 1 and T-S. Chang 2 

Department of Engineering Science and Mechanics 

Virginia Polytechnic Institute and State University 

Blacksburg, Virginia 24061, USA 

ABSTRACT 

For the design of structures installed with viscoelastic dampers it is important to work with analytical 

models that realistically capture the constitutive properties of the material and are easy to use 

analytically. A simple model with a spring and a dashpot in parallel is often used but it does not 

include the frequency dependence of the loss and storage moduli of the material. To include this 

dependence, the use of fractional derivative models has been suggested. However, these models are 

quite complicated to analyze. Herein, therefore, the use of a generalized linear model using a series of 

linear derivatives is explored. Such a model is capable of capturing the frequency dependency of the 

damper properties, and is also analytically convenient to work with. The paper presents a comparison 

of the numerical results obtained for a structure installed with viscoelastic dampers modeled by 

fractional derivative and by generalized linear models. Based on this study the use of the generalized 

linear models for the viscoelastic dampers is advocated. 

INTRODUCTION 

For response control of structural systems subjected to earthquake induced motions, the passive and 

active vibration control schemes have been considered, with passive schemes being used more 

commonly in practice. The passive schemes mainly consist of three different approaches: base 

isolation, tuned mass dampers, and energy dissipation approaches. Base isolation filters out the energy 

approaching the structure usually through a soft base isolation element. In earthquake engineering, this 

approach is more suitable for structures with higher frequencies such as low to medium height 

structures. The tuned mass dampers reduce the response by shifting of energy from the main structure 

to the tuned mass damper, and not by energy dissipation; they have received mixed support about their 

usefulness in seismic design applications. The energy dissipation systems on the other hand dissipate 

the energy in some discrete elements, called dampers, installed in the structure so that less is available 

to deform and damage the main structure. The classical viscous dampers, viscoelastic dampers, 

1 Preston Wade Professor 

2 Graduate Student

496 

yielding metallic dampers, and friction dampers are among the common energy dissipation systems 

that have been used in seismic design applications. This study is focused on some analytical issues 

primarily pertaining to the viscoelastic dampers. 

MODELS OF VISCOELASTIC DAMPERS 

For structural designs with an energy dissipation device, it is important to have the correct forcedeformation 

model of the device to represent the actual behavior of the material used in the device. 

The model should also be convenient for use in dynamic analysis to predict dynamic response as 

accurately as possible to examine different design alternatives. The viscoelastic dampers use a 

polymeric material that dissipates energy by shear deformations applied cyclically. These dampers 

contribute both to the damping and stiffness of a structure. To represent the load-deformation 

characteristics of these dampers, models with increasing degree of complexity have been used. One of 

the simplest models used is a classical Kelvin model with frequency-independent damping and 

stiffness elements arranged in parallel. The analysis of structures with such dampers is quite 

straightforward. However, this simple model does not capture the frequency dependence of the 

damping and stiffness characteristics of the material that are observed in the experiments. To capture 

the frequency dependence, the complex modulus relating the strain with stress has been used; it 

describes the relationship accurately for a purely harmonic motion. The real part of the complex 

modulus (called the storage modulus) represents the stiffness property of the material. The imaginary 

part (called as the loss modulus), on the other hand, is associated with the energy dissipation and thus 

damping. The linear structures installed with these dampers can be analyzed using the discrete Fourier 

transforms in frequency domain. For the response spectrum analysis of a structure installed with these 

dampers, the modal strain energy approach (linger and Kerwin, 1962) can be used (Soong and 

Dargush, 1997) to incorporate the frequency dependency of the parameters approximately. 

Equation Section 3 

Another model used to incorporate the frequency dependence is the fractional derivative model. It is a 

generalization of the classical Kelvin model but with a fractional derivative. It is defined by the 

following differential equation: 

f=ku + cD a (u}, 0

497 

The equation of motion of a structure installed with dampers modeled by fractional derivatives and 

excited by earthquake induced ground motion x g (r) will also have fractional derivative terms: 

M{3c} + CW + CD ff (W)+K f U} = -M{r}^(r) (3.4) 

where M and C, respectively, are the mass and inherent damping matrix of the structure; [x] is 

relative displacement vector; K, is the total stiffness matrix including the effect of the damper, and C 

is the damping matrix due to the contribution of the damper. The solution of these equations for an 

arbitrarily defined input such as earthquake acceleration time history (that are defined at discrete time 

points) is a complicated task. Only a few analyses with earthquake induced ground motions have been 

reported in the literature, and they have been primarily limited to the response calculation of single 

degree of freedom systems. Makris and Constantinou (1991) used the Fourier transform methods to 

obtain the numerical results. Koh and Kelly (1990) used the LI algorithm to solve the problem of a 

single degree of freedom base-isolated system in which the damping characteristics of the base 

isolation element was represented by a fractional derivative model. L-l signifies the use of the 

Luoiville-Riemann definition for the fractional derivative. The Gl algorithms using the Grundwald 

representation have also been used. These algorithms can be extended for direct integration of the 

equations of motion of the multi-degree of freedom (MDOF) structures installed with fractional 

derivative modeled dampers. In these schemes, the regular derivatives in the equations of motion are 

represented either by some finite difference schemes or by Newmark-/ approach. The development 

of these numerical integration schemes for MDOF systems can be found in Chang (2002). 

These equations can also be solved by a special eigenvector expansion approach, herein called as the 

modal approach similar to that used for linear systems with regular derivatives, if the exponent of the 

fractional derivative can be expressed as a rational number, that is a = l/m, i

498 

where the coefficients a m and b n with the derivatives in the model are chosen such that the model 

captures the frequency dependent properties of the material accurately. For this, the storage and loss 

moduli corresponding for this model are compared with the experimentally obtained moduli. The 

coefficients are adjusted to obtain a good fit over the range of frequencies of interest. The complex 

modulus for this model, which defines the storage and loss moduli, can be expressed as: 

(1- a, 

(3-6) 

Figure 3.1 Mechanical representation of a linear model with 7 parameters 

It can be shown that the system shown in Fig. 3.1 is equivalent to the model of Eqn. 3.5. The 

coefficients in Eqn. 3.5 can now be expressed in terms of the spring and damper coefficients in Fig. 

3.1. In this study, a third-order model has been used to obtain numerical results. For such a model, 

there are seven coefficients, which can be written in terms of the spring coefficients ^...£4 and 

damper coefficients c 2 , c 3 , and c 4 as follows: 

)+ c ; (*,-'+tf+k- 1 )+c 4 (k- 1 +£+*,-«)] 

;)+(^'+A 3 - 1 )(c I +c;)+(fc,- 1 +^')( C 3 +c 4 )l 

(3.7) 

"L 

4A ; 

499 

numerical approach. However, one can also formulate a more convenient self-adjoint system of state 

equations with the help of the physical model shown in Fig. 3.1. To describe the dynamic deformation 

of each damper completely, we introduce auxiliary displacement coordinates at the interface of each 

damping element as shown in Fig. 3.1. In terms of these auxiliary displacement coordinates and the 

displacement coordinates that define the motion of each structural degree of freedom, one can develop 

the kinetic energy function, potential energy function, and dissipation function. Using the Lagrange 

equations, the equations of motion can then be explicitly obtained. Using appropriate auxiliary 

equations, and after some re-arrangements of terms, these equations can be written in the self-adjoint 

state form. For a shear building with viscoelastic dampers installed in each story, these state equations 

can be written as follows: 

(4.1) 

where the state matrices are defined as: 

C 

0 

0 

-c 4 

M 

0 

C 2 

0 

0 

0 -( 

C+C 

0 

Ml 

0 

0 

0 

0 J 

,B = 

~~K 

Kfc 

0 

K 4 

_ 0 

K b 

-K,-K 

K 2 

0 

0 

K 2 0 0 

K 4 Ol 

ic 3 o 

-K 3 -K 4 0 

0 Mj 

(4.2) 

withC = C + C 4 , K=K + K 4 +K fl , C, ^ 

and 0 elsewhere, and the state vector is defined as {y} = {{*},fa},{z 2 },{zi},{x}] . It is noted that the 

state matrices are symmetric, and the system is self-adjoint. It is also noted that, for the third order 

model used here, the state vector is of size 5N. Usually a third order model would have enough 

parameters to change to obtain a close match with the experimentally obtained moduli. 

The system of equations, Eqn. 4.1 can be uncoupled using the eigen properties of the system. A modal 

analysis approach can be formulated, with all the advantages associated with such an approach. For 

example, one can develop a response spectrum approach to calculate the maximum displacement, story 

shear, floor acceleration, etc. for seismic design inputs defined in terms of response spectra. The 

details of this formulation are provided by Chang (2002). 

Equation Section 5 

NUMERICAL RESULTS 

In this section we present the numerical results of seismic analysis of a structure installed with 

viscoelastic dampers. To show that the higher order linear model of Eqn. 3.5 is as good as the 

fractional derivative model, we compare the numerical results obtained by the two formulations. 

A 4-story shear building with story masses of m^ ~ m^ = 40 Mg, m^ = m 4 = 36 Mg, and story stiffness 

coefficients of K : =18 MN/m , K 2 = £ 3 =12MN/m, K 4 = 10MN/m, was chosen to obtain the 

numerical results. To introduce inherent energy dissipation in the system, a damping matrix providing

500 

modal damping ratio of 2% in each mode was assumed. The undamped frequencies of the system are: 

1.07Hz, 2.89Hz, 4.39Hz, and 5.29Hz. The viscoelastic dampers were installed in the diagonal bracing 

of each story at 45° angles. The dimensions of the viscoelastic material used in the dampers were: 

area, A = 200 cm 2 , and material thickness, h = 3 cm. 

The model parameters for the fractional derivative model were chosen as: a = 3/5, G 0 = O.SxlO 5 

N/m 2 , GI - 7.2 x 10 5 sec* N/m 2 . For these parameters, the storage and loss moduli are defined by: 

G'(£y)=o.8-K7.2cos— )

501 

simpler and computationally very efficient compared to the solution of the uncoupled equations in the 

fractional derivative model case. 

Table 5.1 Comparison of maximum response 

Response 

Floor 1 

Floor 2 

Floor3 

Floor 4 

Relative displacement (m) 

No 

damper 

0.0315 

0.0804 

0.1242 

0.1598 

Fractional 

derivative 

0.0161 

0.0338 

0.0467 

0.0543 

Linear 

7-para 

0.0162 

0.0341 

0.0472 

0.0550 

Absolute acceleration (m/s~) 

No 

damper 

6.48 

6.10 

6.48 

10.0 

Fractional 

derivative 

3.28 

2.92 

3.40 

4.14 

Linear 

7-para 

3.26 

2.94 

3.50 

4.32 

No 

damper 

566 

590 

527 

359 

Story shear (kN) 

Fractional 

derivative 

290 

218 

160 

79.7 

Linear 

7-para 

291 

220 

164 

82.3 

The time histories calculated by the two approaches for displacement and acceleration responses were 

very similar, but not exactly the same. A more meaningful comparison is obtained when the maximum 

response quantities are compared. In Table 5.1 we compare the maximum values of the floor 

displacements, story shears, and floor accelerations calculated by the two approaches. It is noted that 

the response values calculated by different approaches are close to each other. Also shown are the 

values for the undamped case. The response reducing effect of the dampers is clearly evident. In 

Figures 5.2(a) and 5.2(b) we show the floor response spectra of the floors 2 and 4, again obtained by 

the two approaches. The difference between the floor response spectra calculated by the two 

approaches is minimal at higher frequencies. At lower frequencies, these is some difference, primarily 

because of the differences in the match of the frequency dependent moduli, Fig. 5.1. However, this can 

be improved. This verifies that the linear higher derivative model can provide an as accurate 

description of viscoelastic dampers as the fractional derivative model. 

70 

(a) Floor 2 (b) Floor 4 

0.2 1 10 100 0.2 1 10 



Figure 5.2 Floor spectra of pseudo acceleration at floor 2 and 4 

100

502 


The paper reviews the analytical models that have been proposed to model the frequency dependent 

behavior of the viscoelastic materials used in dampers. Among these different models, the fractional 

derivative model is considered to capture the frequency dependent characteristics quite well. However, 

the analytical complexities of this model are quite significant, and solution of the equations of motion 

of structures installed with these dampers can be quite tedious. Therefore, the use of a generalized 

Kelvin-Maxwell model with higher order derivatives is proposed. This model can capture the 

frequency dependencies of the material properties as well as the fractional derivative model. However, 

this model is analytically much simpler to work and numerically more efficient than the fractional 

derivative model. The use of this model facilitates the development of a response spectrum method of 

analysis, commonly used in earthquake engineering with linear systems. The optimal damper 

placement studies that are necessary for achieving performance-based design of buildings with these 

dampers can also be greatly simplified. 


This work is supported in part by the National Science Foundation through Grant No. CMS-99S7469. 

This support is gratefully acknowledged. The opinions, findings, and conclusions of this work are 

those of the writers and do not necessarily reflect the views of the National Science Foundation. 

REFERENCES 

Bagley, R. L. and Torvik, P. J. (1985). Fractional Calculus in the Transient Analysis of 

Viscoelastically Damped Structures. AIM Journal 23:6, 918-925. 

Chang, T-S. (2002). Seismic Response of Structures with Added Viscoelastic Dampers. Ph. D. 

Dissertation, Virginia Tech, Blacksburg, VA 24061, USA 

Chang, T-S. and Singh, M. P. (2002). Seismic Response of Structures with Fractional Derivative- 

Based Viscoelastic Dampers. Earthquake Engineering Research Institute, Proc, CD-ROM, 7 th U.S. 

National Conference on Earthquake Engineering, Boston, M.A., July 21-25. 

Makris, N. and Constantinou, M. C. (1991). Fractional-Derivative Maxwell Model for Viscous 

Dampers. ASCE Journal of Structural Engineering 117:9, 2708-2724. 

Koh, C. G. and Kelly, J. M. (1990). Application of Fractional Derivatives to Seismic Analysis of Base- 

Isolated Models, Earthquake Engineering and Structural Dynamics. 19, 229-241. 

Soong, T. T. and Dargush, G. F. (1997). Passive Energy Dissipation Systems in Structural 

Engineering, John Wiley and Sons, Chichester, U.K. 

Suarez. L. and Shokooh, A. (1997). An Eigenvector Expansion Method for the Solution of Motion 

Containing Fractional Derivatives. Journal of Applied Mechanics 64:3, 629-635. 

Unger, E. E. and Kerwin, E. M. (1962). Loss Factor of Viscoelastic Systems in Terms of Energy 

Concepts. Journal of American Acoustic Society 34, 954-957.




SEISMIC RESPONSE CONTROL OF A BENCHMARK 

CABLE-STAYED BRIDGE 

J. N. Yang 1 , S. Lin 1 and F. Jabbari 2 

Department of Civil and Environmental Engineering 

Department of Mechanical and Aerospace Engineering 

University of California, Irvine, CA 92697, USA 

ABSTRACT 

Control systems have been shown to be quite effective in reducing the seismic response of buildings and 

bridges. In this paper, two HZ based control strategies with energy-bounded or peak-bounded excitations 

are proposed and their applications to the seismic response control of a benchmark cable-stayed bridge 

are presented. For design loads specified by a class of "energy-bounded" or peak-bounded excitations, 

both controllers are derived by minimizing the upper bound of the HI performance. The design syntheses 

of these control strategies are developed and formulated within the framework of linear matrix 

inequalities (LMIs), so that the LMI toolbox in MALAB can be used effectively and conveniently. The 

performance of these two control strategies is demonstrated using a seismically excited benchmark 

cable-stayed bridge. Simulation results indicate that: (i) the performance of the proposed control 

strategies is excellent in comparison with that of the LQG sample controller, and (ii) the seismic 

response of the cable-stayed bridge can be reduced significantly through the application of the control 

technology. 

INTRODUCTION 

It has been shown that various control systems can be used effectively to reduce the response of civil 

engineering structures subject to strong winds, earthquakes and other service loads [e.g., Housner et al 

(1997), Spencer and Sain (1997), Yang et ai (2QOOb,c; 2002a,b)]. Benchmark problems for control of 

buildings and bridges have been established recently [ e.g., Spencer et al (1998), Ohtori et al (2000), 

Yang et al (2000a), Dyke et al (2000) ]. The response control of long-span cable-stayed bridges has 

attracted considerable attention [e.g., Yang et al (1979a,b), Warnitchai et at (1993), Achkire & Preumont 

(1996), Johnson et al (2002) ]. In particular, the connection between the bridge deck and towers affects 

the seismic response of cable-stayed bridges. Under seismic excitations, a strong connection will reduce 

the displacement response of the deck and increase the shear force and bending moment at the base of 

the towers. A weaker connection can be used as a compromise between the deck displacement and forces 

at the base of the tower to restrict the deck motion during small earthquakes. During strong earthquakes, 

the weaker connection will be damaged to allow a large deck motion thus preventing damages to the 

tower. Another approach is to dissipate the vibration energy of the bridge using active, passive,

504 

semi-active and hybrid control systems. An extensive review of various control systems, including their 

advantages and limitations, can be found in [ He et al (2002) ]. Based on the latter approach, a 

benchmark problem for the seismic response control of a cable-stayed bridge has been established 

recently [ Dyke et al (2000) ]. 

For the benchmark cable-stayed bridge, control systems, such as actuators, semi-active dampers, 

passive viscous dampers, etc , can be installed between the deck and the towers or piers to reduce the 

seismic response. A sample LQG control strategy has been used when actuators are installed in the 

bridge [ Dyke et al (2000) ]. In this paper, we present two H2 based control strategies for applications to 

the benchmark cable-stayed bridge and their performances in reducing the seismic response of the 

bridge are evaluated and compared with that of the LQG controller. 

PROBLEM STATEMENTS 

The equation of motion of an n-DOF structure equipped with control systems and subject to earthquake 

or wind excitations can be written as 

Mi(t) + Ci(t) + Kx(t) = Hu(t) + t|w(t) (1) 

in which x(t) is the displacement vector with the ith element xj being the displacement of the ith DOF 

relative to the ground; M, C and K are mass, damping and stiffness matrices of the structure; H is the 

location matrix of controllers; r\ is the influence coefficient matrix of the excitations; and u(t) and w(t) 

are the vectors of control forces and disturbances, respectively. For a structure subject to earthquake 

excitations, r\ is a (nxl) matrix and w(t)=x g (t) is a scalar denoting the earthquake ground 

acceleration. 

In the state space, Eq.(l) can be expressed as 

Z(t) = AZ(t) + Bu(t) + Ew(t) (2) 

in which Z(t) = {x T (t) , i (t)} T is a 2n state vector, A is a system matrix, B and E are appropriate 

matrices, and the superscript T denotes the transpose of a vector or matrix. In general, a ni -dimensional 

controlled output vector z\, a ^-dimensional constraint output vector zi, and a m-dimensional measured 

output vector y can be expressed, respectively, by 

Zl(t) = C 2l Z(t)-hD zl u(t)-hE zl w(t) 

(t) (3) 

where v is a measurement noise vector. Let z^t) be the ith element of the vector Z2(t), where all 

elements of za(t) denote the constraint variables (i.e., physical constraints). Since we are interested in 

limiting or penalizing the peak values of these variables individually and since each of Z2,i(t) is a scalar, 

the peak values of these physical quantities are the LQO norms, i.e., 

1 2 2,i W I oo = sup i Z2 >» (t) I ' 

for [ = ls 2) •"' n 2 ( 4 ) 

For a given structure in Eq.(2), our goal is to find a controller u(t) such that: (i) the closed-loop system 

is asymptotically stable, (ii) the Hi performance index J*2 is minimized, i.e., 

•7*2 = Kw(»|| 2 = [ ~ C trace [TÔJo) T ZlW (jco)] do> ] 1/2 (5) 

and (iii) the constraints for the peak values of all components of 22(1) are satisfied, i.e., 

|Ki(t)L^YiSi> i = l,2,...,n 2 , Vw(t)eW (6) 

In Eqs.(5>(6), T ZlW (s) is the transfer matrix from w(t) to zi(t), the superscript H is the conjugate and

505 

transpose of a matrix or vector, e, (i - 1, 2, ..., n 2 ) are specified design allowable values representing the 

constraints, such as peak control forces, peak strokes, peak drifts, etc., and YJ are design variables 

( weighting factors ) to be adjusted in order to obtain an admissible and good performance controller. 

Further, W can be either a class of "energy-bounded" excitations (design loads), i.e., 

or a class of peak-bounded excitations, i.e., 

w(t)€W = {w(t)|jw(t)| 2

506 

Iiz2i(t)l

507 

be adjusted and Eqs.(17)-(19) should be used. In this connection, the function "MINCX" in the LMI 

toolbox of MATLAB can be used conveniently to design both KfcB-EB and HaB-PB controllers. 

APPLICATION TO BENCHMARK CABLE-STAYED BRIDGE 

A benchmark model for a cable-stayed bridge has been established [Dyke et al (2002)] in order to 

investigate the effectiveness of control systems during an episode of earthquake. The benchmark 

cable-stayed bridge is the Missouri 74-fllinois 146 bridge spanning the Mississippi River near Girardau, 

Missouri, and a schematic drawing of the bridge is shown in Fig.l. A control system has been suggested 

and a sample LQG controller has been worked out in the benchmark paper to serve as a reference for the 

comparison of control performances with that of the other control strategies. In this control system, 

actuators are installed between the deck and two towers (a total of 4 locations) as well as between the 

deck at both ends and the pier and abutment, respectively. A total of 24 hydraulic actuators are installed 

symmetrically with respect to the centerline of the bridge at a total of 8 locations as shown by the dotted 

areas in Fig.2. The number of actuators installed at each location is indicated in parentheses next to the 

dotted area. All actuators are identical and each has a capacity of lOOOkN. Three earthquakes records 

have been considered as the input excitations in the longitudinal direction of the bridge, i.e., El Centro 

NS, Gebze NS and Mexico earthquakes. 

The performance indices are defined by Jj to Jig, where Jj = maxi. base shear of tower; ^2 = maxi. 

deck shear; Js = maxi. base moment of tower; 14 = maxi. deck moment; Js = maxi. cable force deviation; 

Jg = maxi. deck displacement; Jy = norm base shear of tower; Jg - norm deck shear; 19 = norm base 

moment of tower; JIQ = norm deck moment; Jn= norm cable force deviation; Ji2~ maxi. control force; 

Jn = maxi. actuator stroke; Ji4= maxi. actuator power; ]\s= total actuator power; Jig = number of 

actuators; Jn = number of sensors; and Jig = control resources. The norm Ml of a response quantity is 

-468' -1150' -468' 1870' 

I 

I 

Bent 1 Pier 2 

to 

I 

I 

Pier 3 Pier 4 

Fig. 1: Drawing of the Cape Girardeau Bridge [Drawing Courtesy of Dyke et al (2002)]. 

/Protective Devices 

Ill ( 2 ) 

S:S ( 4 )^— Number of Ac;tuators N ted (4) w mN V^l^: 

Illinois> Approach " ^ 

Bentl Pier 2 Pier 3 

Fig. 2: Locations of Protective Devices in the Cable-Stayed Bridge.

508 

defined as 114 - [ J* [•] 2 dt /t ] l ' 2 , where T is sufficiently large to allow the response of the structure 

to attenuate, and [•] is the quantity whose norm is to be calculated. Note that Ji to Ju are bridge 

response quantities normalized by the corresponding uncontrolled responses, and J^ to Jâre the 

required actuator capacities normalized by different quantities. Hence, the control performance is better 

if the value of indices J t (i =1,2,...,18) is smaller. 

Evaluation criteria Ji to Jis for each control strategies are shown in Table 1, in which the result 

denote the maximum value due to 3 earthquakes, i.e., El Centra, Gebze, and Mexico. Columns (2) and (6) 

of Table 1 show the results for the LQG sample controller considered in [Dyke et al (2002)], except that 

all the diagonal elements of the weighting matrix R are reduced from 1.0 to 0.62, so that the maximum 

actuator control force is increased to 996kN (see Table 1) to improve the control performance. 

For both H 2 B-EB and H 2 B-PB controllers, we use the same state-reduced order system and the 

Kalman-Bucy filter as that used hi the LQG sample controller. Our control output vector z^t) is 

selected as, zTz^z^z, where z T Qz was used in the objective function of the LQG sample 

controller. The constraint output vector z 2 (t) in Eq. (3) is equal to u(t), i.e., z 2 (t) = u(t) = [MI, u 2 ,..., 

ug] T where u l is the control force from one actuator at the ith location and it has a peak force 

limitation of 1000 kN. Hence, only the constraints on the actuator peak forces are considered in the 

constraint output vector z 2 (t), i.e., B l =1000 kN for i = 1, 2, ..., 8. Finally, measured output vector y(t) 

in Eq.(3) is identical to that used in the LQG controller, i.e., four accelerometers are installed on the top 

of the tower legs and one accelerometer on the deck at the mid-span, whereas two displacement sensors 

are located between the deck and pier 2 and other two displacement sensors are located between the deck 

and pier 3. Based on three earthquake records, the maximum total "energy" is computed as p 2 = 3.37 

m/sec 3/2 (El Centro earthquake), whereas the maximum peak ground acceleration (PGA) is p00= 3.42 

m/sec 2 (El Centro earthquake). With the input parameter 8j, P 2 and p^ above, the design parameters 

for the H 2 B-EB controller are y j = y = 4.76 for i = 1,2,..., 8, whereas that for the H 2 B-PB controller are 

y> =y = 68.95 for i = 1,2,..., 8 and cc=0.01. 

The resulting performance indices for these two controllers are presented in Table 1. A comparison 

of the results in Table 1 indicates that the performances of the proposed new controllers are superior to 

that of the LQG sample controller. Peak values of control forces for all controllers in Uj (i= 1,2,..., 8) 

Table 1: Evaluation Criteria and Peak Control Forces For LQG, H 2 B-EB and H 2 B-PB Controllers 

Criteria 

(1) 

Ji 

:-> 

J3 

J 4 

h 

J6 

J 

J 8 

J 9 

Force 

ui 

U2 

U 3 

U4 

LQG 

(2) 

0456 

1.340 

0.547 

1.073 

0.174 

3.043 

0.396 

1.348 

0.436 

LQG 

471 

471 

914 

914 

H 2 B-EB 

(3) 

0.454 

1.225 

0.494 

0.964 

0.162 

2.406 

0.394 

1.192 

0.398 

H 2 B-PB 

(4) 

0.453 

1.218 

0.493 

0.956 

0.161 

2.374 

0.394 

1.183 

0.397 

Criteria 

(5) 

JIG 

Jn(xlO- z ) 

Ji2(xlO' J ) 

Jl3 

Ji 4 (xi

509 

are presented in the lower part of Table 1. Peak control forces are maximum under the Gebze earthquake 

for the LQG controller, whereas these quantities are maximum under the El Centro earthquake for both 

H 2 B-EB and H 2 B-PB controllers as shown Table 1. It is observed that the peak control forces of all 

actuators for the proposed controllers are quite close to the upper limit 1 OQOkN. The main reason for the 

superior performance of our proposed controllers is that they are capable of pushing the control 

resources (i.e., peak control forces) of all actuators at all locations to the limit This is a significant 

advantage of the proposed new control strategies over other control methods. It is further observed from 

Table 1 that most of the response quantities of the bridge have been reduced significantly through the 

installation of control devices. 

CONCLUSIONS 

Two H- based control strategies with "energy-bounded" or peak-bounded excitations have been 

proposed and their applications to the seismic response control of a benchmark cable-stayed bridge have 

been presented. For design loads specified by a class of "energy-bounded" or peak-bounded excitations, 

both controllers are derived by minimizing the upper bound of the HI performance. The design syntheses 

of these control strategies are developed and formulated within the framework of linear matrix 

inequalities (LMIs), so that the LMI toolbox in MALAB can be used effectively and conveniently. The 

performance of these two control strategies in reducing the seismic response of structures has been 

demonstrated using a benchmark cable-stayed bridge. Simulation results indicate that: (i) the 

performance of the proposed control strategies is excellent in comparison with the LQG sample 

controller, and (ii) the seismic response of cable-stayed bridges can be reduced significantly through the 

application of control technology. Although actuators are considered in this paper as a possible control 

device, it has been shown that passive and semi-active control systems, including passive viscous 

dampers, MR dampers, resetting semi-active stiffness dampers and semi-active friction dampers, can be 

used to effectively reduce the seismic response of the cable-stayed bridge [ e.g., Moon et al (2002), 

Agrawal et al (2002) ]. 

References 

Achkire, Y. and Preumont, A. (1996), "Active Tendon Control of a Cable-Stayed Bridges", J. of 

Earthquake Engineering and Structural Dynamics, 25: 6, 585-597. 

Agrawal, A.K., Yang, IN., and He, W.L. (2002), "Performance Evaluation of Some Semi-Active 

Control Systems For A Benchmark Cable-Stayed Bridge", Proc 3 rd World Conference on Structural 

Control, Como, Italy. 

Dyke, S.J., Bergman,, L.A., Turan, G. and Caicedo, J.M. (2000), "Benchmark Control Problem For 

Seismic Response of Cable-Stayed Bridge", Proc. 3 rd International workshop on Structural Control, 

Paris, France, Website: http://wusceelcive.wustl.edu/quake/. 

E.A. Johnson, G.A. Baker, B.F. Spencer, Jr., and Y. Fujino (2002), "Semiactive Damping of Stay 

Cables," Journal of Engineering Mechanics, ASCE, in press. Web version online at http://rcf.usc.edu/ 

~j ohnsone/papers/smartdamping_tautcablej em.html. 

Housner, G.W., Bergman, LA., Caughey, T.K., Chassiakos, A.G. and others (1997), "Structural 

Control: Past, Present, and Future", J. ofEngrg Mechanics, ASCE, 123:9, 897-971. 

He, WJL, Agrawal, A.K., and Yang, J.N. (2002), "A Novel Semi-Active Friction Controller for Linear 

Structures Against Earthquakes", to appear in ASCE Journal of Structural Engineering. 

Moon, S.J., Bergman, LA, and Voulgaris, P.G« (2002), "Sliding Mode Control of a Cable-Stayed 

Bridge Subject to Seismic Excitation", Proc. 3 r World Conf on Structural Control, Como, Italy. 

Ohtori, Y., Christenson, R.E., Spencer, B.F. and Dyke, SJ. (2000), " Benchmark Control Problems for 

Seismically Excited Nonlinear Buildings", World Wide Website: http://www.nd.edu/ 

~quake/bench.html, to appear in the Special Issue of ASCE J. Engrg Mechanics, 2003.

Spencer, B.F., Jr. and Sain, M.K. (1997), "Controlling Buildings: A New Frontier in Feedback", IEEE 

Control Systems, 17:6,19-35. 

Spencer, Jr., B.F., Dyke, S.J., and Deoskar, H.S. (1998), "Benchmark Problems in Structural 

Control-Part 1: Active Mass Driver System, and Part 2: Active Tendon System." J. Earthquake Engrg. 

and Structural Dynamics, 27:11,1127-1147. 

Warnitchai, P., Fujino, Y., Pacheco, B.M. and Agret, R. (1993), "An Experimental Study on Active 

Tendon Control of Cable-Stayed Bridges", J. of Earthquake Engrg. & Structural Dynamics, 22:2, 

93-111. 

Yang, IN., and Giannopoulos, F. (1979a), "Active Control and Stability of Cable-Stayed Bridge", 

ASCEJ. Engineering Mechanics, 105, 677-694, 

Yang, J.N., and Giannopouios, F. (1979b), "Active Control of Two-Cable-Stayed Bridge", ASCE J. 

Engineering Mechanics, 105,795-810. 

Yang, J.N., Agrawal, A.K., Samali, B. and Wu, J.C. (2000a), "A Benchmark Problem For Response 

Control of Wind-Excited Tail Buildings", Proceedings of!4th Engineering Mechanics Conference, 

ASCE, May 21-24, Austin, Texas, CD-ROM, 8 pages, Website: http://www-ce.engr. 

ccny .cuny .edu/people/factdty/agrawal/benckmark.html, to appear in Special Issue of ASCE J. Engrg. 

Mechanics, 2003. 

Yang, J. N., and Agrawal, A. K., (2000b), "Protective Systems For High-Technology Facilities Against 

Microvibration and Earthquake", International Journal of Structural Engineering & Mechanics, 10:6, 

561-575. 

Yang, J. N., Kim, J.H., and Agrawal, A.K., (2000c), "A Resetting Semi-Active Stiffness Damper for 

Seismic Response Control",/, of Structural Engineering, ASCE, 126:12, 1427-1433. 

Yang, J.N., and Agrawal, A.K., (2002a), "Semi-Active Hybrid Control systems For Nonlinear Buildings 

Against Near-Field Earthquakes", InternationalJournal of Engineering Structures, 24, 271-280. 

Yang, IN., Lin, S. 9 Kim, J.H., and Agrawal, A.K. (2002b), "Optimal Design of passive Energy 

Dissipation Systems Based on H* and H^ Performances", Journal of Earthquake Engineering and 

Structural Dynamics, 31:4, 921-936. 

510

Proceedings of the International Conference on c-11 



EVALUATION OF SUPPLEMENTAL ENERGY DISSIPATION 

DEVICES IN PROTECTING HIGHWAY BRIDGES WITH 

SOIL-STRUCTURE INTERACTION 

Jian Zhang and Nicos Makris 

Department of Civil and Environmental Engineering, University of California, Berkeley 

Berkeley, CA 94710, USA 

ABSTRACT 

This paper presents the response analysis of a freeway overcrossing equipped with isolation bearings and 

fluid dampers in order to evaluate the efficiency of supplemental energy dissipation devices to suppress 

the seismic response of highway bridges. Recognizing that the soil-structure interaction affects 

appreciably the earthquake response of highway overcrossings, the study employs an elementary stick 

model enhanced with frequency-independent springs and dashpots that approximate the dynamic 

stiffnesses of approach embankments and pile groups. The nonlinear behavior of columns as well as 

hysteretic behavior of seismic protection devices (elastomeric bearings and viscous fluid dampers) are 

included in the study. The emphasis is placed to understand the effect of supplemental damping in 

association with the ability of the structure to dissipate energy through soil-structure interaction. It is 

concluded that the fluid dampers are effective in controlling displacement demands, however the 

flexibility of the approaching embankments is responsible for partially reducing their efficiency. 

INTRODUCTION 

Earthquake damage in most highway overcrossing is the result of excessive seismic displacements and 

large force demands that have been substantially under-estimated during past design. A direct 

consequence of the under-estimated seismic displacements, which were the combined result of poor 

representation of the kinematic characteristics of the ground, low lateral forces, and overestimated 

stiffnesses, was that the seating length at the deck supports was unrealistically short and the lateral 

separation between adjacent structures were typically inadequate, resulting in loss of support or pounding 

(Maragakis and Jennings 1987). These geometrical inconsistencies resulted in spectacular failures that 

have been witnessed during the recent 1989 Loma Prieta and the 1994 Northridge earthquakes in 

California and the 1995 Kobe earthquake in Japan. 

In view of these failures many research programs were launched after the 1971 San Fernando earthquake 

to study the seismic resistance of highway bridges. Improvements have been achieved in both design and 

analysis of bridge structures with the help of strong motion records. Extensive retrofit programs have 

been implemented in California, which include jacketing of columns and the use of composite materials 

(FHWA 1995). While retrofit programs are underway there is ongoing need for better understanding the 

behavior of each substructure elements of the bridge and to what extend its individual behavior affects the 

overall response of the bridge structure. In particular, the increasing need for safer bridges in association

512 

with the rapid success of seismic protection devices in buildings has accelerated the implementation of 

large-capacity damping devices in bridges (Delis et al. 1996; Smyth et al. 2000; Papanikalas 2002). The 

presence of the isolation system and other protective devices is expected to alter appreciably the response 

of bridge and therefore the role of its substructure elements in association with the role of soil-structure 

interaction needs to be reexamined for this class of "flexible" structures. 

The promises of modern seismic protection technologies about their ability to operate under strong 

shaking has directed most of the attention on the performance of bearings and dampers under large 

displacements and large velocities. The interaction of these devices with the remaining bridge structure is 

an issue that has been either incorporated in the response analysis indirectly via the finite element analysis 

of the entire bridge with large computer codes or has been neglected; partly because the transmitting 

forces are usually relatively small and the reaction structures are usually relatively stiff. 

Ebstorreric Bearing 

Figure 1. Elastomeric bearing and fluid damper layout of 91/5 Overcrossing (left); and photograph of 

four dampers installed at east abutment (right). 

In this paper the efficiency of modern seismic protection technologies is examined by analyzing the 

seismic response of a newly constructed highway overcrossing in southern California, which is the first 

seismic isolated overcrossing in the United States equipped with fluid dampers. The 91/5 overcrossing of 

interest is a continuous two-span, cast-in-place prestressed concrete box-girder bridge. It is supported at 

each end-abutments on four elastomeric pads while it is attached with four fluid dampers. Figure 1 shows 

the layout of elastomeric bearings and fluid dampers (left) and a view of four fluid dampers installed at 

the east abutment. The deck is supported near the center bent by a prestressed reinforced concrete 

outrigger which rests on two pile groups, each consisting of 49 driven concrete friction piles. Figure 2 

shows an idealization of seismic isolated overcrossing together with its approach embankments and pile 

foundations. The interesting characteristic of this structure is that its transverse and longitudinal modal 

periods lie in the range between 0.4sec and O.Ssec, that is a period range for which supplemental damping 

on a single-degree-of-freedom structure has beneficial effect. Furthermore, the bridge is approached from 

each side by earth embankments that have a tendency to amplify the free-field motion and increase the 

role of soil-structure interaction. Accordingly the assessment of the efficiency of the seismic protection 

devices is conducted by accounting in our analysis the effect of soil-structure interaction at the end 

abutments/approach embankments and at the foundations of the center columns. In principle lengthening 

the period of a structure with mechanical isolation reduces accelerations and increases displacements.

Nevertheless, a more flexible configuration offers to the deck additional mobility that may result to 

undesirable response. 

513 

Pile Foundation 

Approach 

E mbanlcmeni 

and Pile 

Foundation at 

Abutment 

Approach 

JSmbaiikment 

and Pile 

Foundation at 

Abutment 

Figure 2. Schematic of an idealized model of the 91/5 Overcrossing 

MACROSCOPIC CONSTITUTIVE LAWS OF SUBSTRUCTURE ELEMENTS 

The bridge is decomposed into its main substructure components including approach embankments, pile 

foundations, center bent and seismic protection system that consists of isolation bearings and fluid 

dampers. Simple macroscopic constitutive laws that capture satisfactorily the restoring and energy 

dissipation mechanisms of these substructural elements within the deformation levels of interest are 

adopted. 

Recognizing that soil-structure interaction affects appreciably the earthquake response of conventional 

highway overcrossings, a recent study by Zhang and Makris (2001, 2002a,b) developed a simple 

procedure to compute the kinematic response functions and dynamic stiffnesses of approach 

embankments that has been validated against recorded data from two well-instrumented bridges with 

integral abutments, the Meloland Road and the Painter Street overcrossings. That study also indicated 

that & although the dynamic stiffnesses of approach embankments and pile foundations are in general 

frequency dependent quantities, their low frequency values do not fluctuate appreciably with frequency 

and one can replace them with frequency independent springs and dashpots. 

For the center bent, the columns are expected to behave inelastically. Therefore, their behavior ^ is 

represented by the nonlinear moment-curvature curve, that was computed with the geometry and material 

properties of columns. 

The mechanical behavior of seismic protection systems is nonlinear since isolation bearings yield or slide, 

while fluid dampers deliver forces that depend on a fractional power of the piston velocity. One can use a 

bi-directional bilinear model to characterize their behavior. Under shear deformation the elastomeric pads 

deform nearly elastically (K eff * 5MN/m) until they develop a lateral force F = pJV, where p, = 0.3 is 

the friction coefficient of the pad-deck interface and N is the normal force on the pad. In this study the 

force at which sliding initiates is F« 0.3MN and the yield displacement is u y = F/K sff * 0.06m. Each 

fluid damper implemented is designed to produce 250kips at a piston velocity of 42in/s. Their behavior is 

nonlinear and the force output is proportional to a fractional power of the velocity, 

P(/) = C a |w(Ol a sgn[M(0], wk ere fractional power, CL Q IJ taken to be 0.35 in this application. 

Accordingly, the damping constant, C a is 1.09MN- (s/m) ' .

Figure 3 summarizes the static push-over curves that result from the constitutive models adopted in this 

study. Each push-over curve extends to the range of deformation that the corresponding element 

experiences during the strong (Case 1) and moderately strong (Case 2) earthquake loading. All 

substructure elements except the isolation bearings exhibit a nearly elastic behavior. This observation is in 

agreement with observations from the studies on highway overcrossings, such as the Meloland Road and 

the Painter Street Overcrossings that have been shaken by strong earthquakes (Werner et al. 1987, 

McCallen and Romstad 1994, Goel and Chopra 1997). Even the center bent that shares a large fraction of 

the horizontal inertial loading behaves nearly elastic. This finding is in agreement with recent design 

practice adopted by Caltrans. 

514 

Below East Abutment (.r or y) 

ilar for West Abutment) 

0.06 0.08 0.1 

Displacement (m) 

Figure 3. Summary of force-displacement (push-over) curves of various substructure elements of 

interest m this study. Each curve extends to the range of deformation that the corresponding element 

experiences. Case 1: strong earthquake shaking; Case 2: moderately strong earthquake. 

SEISMIC RESPONSE ANALYSIS 

Figure 4 presents a stick model idealization of the 91/5 Overcrossing. The bridge superstructure is 

modeled by beam elements with a massless rigid link at each end that preserves the skewed geometry of 

the bridge deck and serves as the connecting elements between bridge deck and the abutment. The 

elastomeric bearings are replaced by the bilinear models that will produce the same hysteretic loop as a 

sliding bearing. The fluid dampers are replaced by nonlinear dashpot elements and are arranged in a way 

exact as the layout during the construction. The elastomeric bearings serve as the connecting element 

between deck and abutment while the abutment is supported by "springs" and "dashpots" that will replace 

the presence of embankment and pile foundation under the abutment footing. The detailed information on 

the values of these spring and dashpot constants can be found in the thesis by Zhang (2002). 

Because of the proximity of the bridge to active faults, eleven strong ground motions that have been 

recorded in California relatively close to the fault of major earthquakes are selected for the simulation. 

The bridge is subjected to the recorded free-field accelerations at the center bent and the amplified 

accelerations at end abutments along the transverse and longitudinal directions simultaneously. In general 

the fault-normal component is applied to the transverse direction while the fault-parallel component is 

applied to the longitudinal direction.

515 

Embankment, Beanngs 

& Fluid Dampers 

Embankment, Bearings 

& Fluid Dampers 

Pile Foundation 

(Equivalent Beam & Dashpot 

Stick Model 

Figure 4. Numerical model of 91/5 Overcrossing 

In this paper, the dynamic response of three configurations of the 91/5 Overcrossing are compared: (a) the 

as built configuration where the deck is supported at both ends by elastomeric pads and is equipped with 

fluid dampers (Pads+Dampers); (b) same configuration as (a) but without fluid dampers (Pads only); (c) 

the bridge-deck is rigidly connected to integral abutments (Integral Abutments). 

Figure 5 summarizes the peak total accelerations and relative displacements of Point A (east end of deck) 

for various earthquakes. The results are sorted according to the peak ground acceleration of the faultnormal 

component of each earthquake record. The thin lines are the results for all three cases when soilstructure 

interaction is neglected while the think lines are the results when soil-structure interaction is 

included in the analysis. Along the longitudinal direction, the bridge with sitting abutment is more 

flexible than that with integral abutments, so accelerations are smaller and displacements are larger. 

Damping reduces both displacements and accelerations of the flexible configuration. However, along the 

transverse direction, the case of bridge with integral abutment not only yields smaller accelerations but 

also smaller relative displacements. This can be explained by looking at the transverse mode of the two 

configurations. The transverse mode of the case of integral abutment is primarily a flexural mode, 

whereas in the case of sitting abutment the transverse mode is primarily a translational mode where the 

entire deck translates sideways without flexing appreciably. This results to larger displacements at the 

deck ends and also larger accelerations. Supplemental damping reduces both displacements and 

accelerations but the response of the bridge with sitting abutments appears to under-perform the response 

of the bridge with integral abutments. Figure 5 also indicates that the soil-structure interaction increases 

both accelerations and displacements. Similarly, Figure 6 plots the peak total accelerations and relative 

displacements of Point B (middle of the deck). The trend of accelerations and displacements along the 

longitudinal directions resemble the trend that one observes at Point A. Along the transverse direction the 

results for accelerations and displacements of three configurations are mixed because the mid-span moves 

sideways approximately the same amount regardless of whether the transverse movement is the result of a 

primarily flexural mode or of a primarily translational mode. Again the bridge response is considerably 

underestimated when the effect of soil-structure interaction is neglected. 

Figure 7 plots the peak forces behind the end abutments. Clearly, the isolated configuration reduces the 

longitudinal forces but not the transverse forces. Interestingly, the presence of fluid dampers yields 

transverse forces that are higher than the forces when the bridge has integral abutments. Figure 8 plots the 

peak shear forces at the base of columns. Along the transverse direction the isolated bridge transmits 

approximately the same forces to the column base that the bridge with integral abutments transmits. 

Along the longitudinal direction, the differences are dramatic since in some earthquakes the column base 

shears of the isolated bridge are more than two times the column base shears of the bridge with integral 

abutments.

CONCLUSIONS 

516 

This paper presents a case study on the seismic response of the 91/5 Overcrossing that is equipped with 

isolation bearings and fluid dampers at its end-abutments. The study adopts the substructure method and a 

reduced-order stick model. The simple macroscopic constitutive laws of the main substructure elements 

are proposed. The study reveals that lengthening of the period of an overcrossing by introducing isolation 

bearings reduces the longitudinal accelerations of the deck but increases the translational accelerations. 

The supplemental damping is beneficial but in many cases the configuration with integral abutment is 

shown to yield the most favorable response. Soil-structure interaction is responsible for increasing 

displacements while having mixed effect on accelerations. The response of the isolated bridge is 

significantly underestimated when soil-structure interaction is neglected. 

REFERENCES 

Delis, E.A., Malla, R.B., Madani, M, and Thompson, K.J. (1996), "Energy dissipation devices in 

bridges using hydraulic dampers", Proc. Structures Congress XIV, Vol. 2, 1188-1196, Chicago, IL. 

FHWA (1995), "Seismic retrofitting manual for highway bridges", Publication No. FHWA-RD-94-052, 

McLeon, VA. 

Goel, R.K. and Chopra, A.K. (1997), "Evaluation of bridge abutment capacity and stiffness during 

earthquakes", Earthquake Spectra, 13(1): 1-23. 

Maragakis, E.A. and Jennings, RC. (1987), "Analytical models for the rigid body motions of skew 

bridges", Earthquake Engineering and Structural Dynamics., 15(8):923-44. 

McCallen, D.B., and Romstad, K.M. (1994), "Analysis of a skewed short-span, box-girder overpass", 

Earthquake Spectra, 10(4):729-55. 

Papanikolas, RK. (2002), "Deck superstructure and cable stays of the Rio-Antirion bridge", Proc. 4th 

National Conference on Steel Structures, Patros, Greece. 

Symth, A.W., Masri, A.F., Abdel-Ghaffar, A.M., and Nigbor, R.N. (2000), "Development of a nonlinear 

multi-input/multi-output model for the Vincent Thomas Bridge under earthquake excitations", Proc. 

12th World Conference on Earthquake Engineering, Paper No. 2211, Upper Hurt, New Zealand. 

Werner, S.D., Beck, J.L. and Levine, M.B. (1987), "Seismic response evaluation of Meloland Road 

Overpass using 1979 Imperial Valley earthquake records", Earthquake Engineering and Structural 

Dynamics, 15:249-74. 

Zhang, J. and Makris, N. (2001), "Seismic response analysis of highway overcrossings including soilstructure 

interaction", Report No. PEER-01/02, University of California, Berkeley, CA. 

Zhang, J. and Makris, N. (2002a), "Kinematic response functions and dynamic stiffnesses of bridge 

embankments", Earthquake Engineering and Structural Dynamics, 31, in press. 

Zhang, J. and Makris, N. (2002b), "Seismic response analysis of highway overcrossings including soilstructure 

interaction", Earthquake Engineering and Structural Dynamics, 31, in press. 

Zhang, J. (2002), "Seismic response analysis and protection of highway overcrossings including soilstructure 

interaction", Ph. D Dissertation, University of California, Berkeley, CA.

517 

Longitudinal 

-o- Integral Abutment (SSI) 

-o- W Pads (SSI) 

•*•• W Pads & Dampers (SSI) 

-e— Integral Abutment (No SSI) 

-a- w Pads (No SSI) 

•*• W. Pads & Dampers (No SSI) 

Peak Ground Acceleration (FN) 


Figure 5. Peak total accelerations (top) and peak relative displacements (bottom) near east end of deck (Point A) due to 

various earthquake motions ordered with increasing peak ground acceleration of the fault-normal component 

3 

2.5 

2 

f 

Transverse 

Integral Abutment (SSI) 

W Pads (SSI) 

W. Pads &. Dampers (SSI) 

Integral Abutment (No SSI) 

W Pads (No SSI) 

W Pads & Dampers (No SSI) 

f 

'

518 

1 

OS 

h 

: 0 6 

5 

5 04 

Transverse 

Integral Abutment (SSI) 

W Pads (SSI) 

• W Pads & Dampers (SSI) 

Integral Abutment (No SSI) 

W Pads (No SSI) 


Longitudinal 

02 

0 

Figure 



7. Peak forces behind end abutments due to vanous earthquake motions ordered with increasing 

peak ground acceleration of the fault-normal component 

0.8 

: 

h 

: 0.6 

J04 

" 0.2 

!> 

: 0 6 

08 

0 &f3a!,,i l l i f I a i I t * s 

1 I 11 H I i I 1 

Longitudinal 

o Integral Abutment (SSI) 

-Or- W Pads (SSI) 

*r W Pads & Dampers (SSI) 

-e- Integral Abutment (No SSI) 

-a- w Pads (No SSI) 



peak Ground Acceleration (FN) 

Figure 8. Peak forces at base of center columns due to vanous earthquake motions ordered with 

increasing peak ground acceleration of the fault-normal component


Advances and New Challenges in Earthquake 51 9 


SEISMIC RELIABILITY ANALYSIS OF 

MULTISTORY ISOLATED BRICK BUILDINGS 

Lingxin ZHANG Minzheng ZHANG Jinren JIANG and Jiepmg LIU 

Institute of Engineering Mechanics, China Seismological Bureau, Harbin, China 150080 

ABSTRACT 

This paper analyzes the seismic reliability of four multistory isolated brick buildings with different 

story numbers and design intensities by Latin Hypercube Sampling technique and nonlinear seismic 

time history response analysis. And the results of the isolated buildings and the corresponding 

buildings without isolation are compared. On the basis of the comparative results, the specific 

suggestions for the seismic design of the isolated brick buildings are proposed. 


With the development of structural isolation technique, people have recognized that using the isolation 

technique can mitigate effectively damage to the buildings subjected to earthquake. Butlt is not clear 

what the seismic safety of the isolated buildings can be increased, compared with the corresponding 

buildings without isolation. This paper studies the safety of the isolated buildings by taking four 

multistory brick buildings with laminated rubber bearings as examples and analyzing their seismic 

reliabilities. The seismic reliability is analyzed by Latin Hypercube Sampling technique and nonlinear 

seismic time history response analysis. In the analysis, the uncertainties of earthquake load and 

structural parameters are considered. 

2 FOUR MULTISTORY BRICK BUILDINGS WITH LAMINATED RUBBER BEARINGS 

In order to evaluate the seismic reliability of multistory isolated brick buildings, a set of representative structures 

needs to be established according to Latin Hypercube Sampling technique. For this reason, we regard multistory 

isolated dwelling brick buildings as examples, and take the maximum story number of the brick buildings for various 

intensities according to Chinese " L Code for seismic design of buildings" (GBJ11-89). Thus, four multistory dwelling 

brick buildings with laminated rubber bearings are chosen, which are an eight-story building for intensity VI, a 

seven-story building for intensity W, a six-story building for intensity Vffl, and a four-story building for intensity DC, 

respectively. Taking use of the typical common design parametric values of multistory dwelling bnck buildings, four 

representative multistory dwelling brick buildings, as shown in table 1, are established by Latin Hypercube Sampling 

technique. The seismic design of multistory dwelling brick buildings is conducted according to Chinese "Code for 

seismic design of buildings"(GBJll-89). The thickness of the inner wall is 24cm, but 37crn for the first and second 

floors of four-story building designed for intensity IX. The grade of mortar is usually M2.5, but it can be increased to 

M5.0 and M10.0 according to the requirement of seismic design. The transversal wall is taken as the 

load-bearing wall in the four buildings. 

The thickness of laminated rubber bearing is 50mm. 

Sponsored by National Natural Science Foundation of China, Grant No. 59978047

520 

TABLE 1 

REPRESENTATIVE MULTISTORY DWELLING BRICK BUILDINGS 

Design parameters 

Story 

Story height (m) 

Top story height (m) 

Bay (m) 

Building width (m) 

Building length (m) 

Outer wall thickness (cm) 

Load of roof (N/m 2 ) 

Load of floor (~N"/m 2 ) 

1 

4 

2.7 

2.8 

3.3 

10.3 

32.2 

37 

3310 

3790 

2 

6 

2.7 

2.7 

3.9 

11.2 

34.6 

37 

3310 

3790 

Buildings 

3 

7 

2.8 

2.8 

3.6 

10.3 

34.6 

37 

3310 

3790 

4 

8 

2.8 

2.8 

3.6 

10.3 

34.6 

37 

3310 

3790 

3 NONLINEAR SEISMIC RESPONSE ANALYSIS 

The brick building model is a shear stick model. The hysteretic model used is a tnlinear model, which 

is given through referring to a large number references [5], as shown in Fig.l. The details of the 

formulas of stiffness and strength in this model are given in Ref. [5]. This model includes a negative 

stiffness branch. In order to avoid the probable unstable phenomenon in the iteration method and 

probable non-definite abnormal matrix in the variable stiffness matrix method during dealing with the 

negative stiffness, we use the method of nonlinear dynamic response analysis based on pattern of 

self-equilibrating stresses [5, 6] to analyze the response of multistory brick buildings. 

Figure!: Hysteretic model of brick wall 

Figure 2: Elasto-plastic hysteretic model 

The hysteretic model used in laminated rubber bearing is an elasto-plastic hysteretic model, as shown 

in Fig.2. 

The earthquake ground motion input used in this analysis is the artificial ground motion [7]. The 

earthquake acceleration time history is expressed as: 

a(t) = \|/(t)a,(t) (1) 

where, a s (t) is a stationary Gaussian process with zero mean value. ¥(t) is an envelope function

521 

describing nonstationary process. 

The acceleration time history a(t) is normalized using its maximum amplitude a max . The normalized 

nonstationary acceleration time history is as fellows: 

V(t)a s (t) 

(2) 

So, the nonstationary acceleration time history is: 

where, A p is a given peak ground acceleration. 

= A p a m (t) (3) 

4 UNCERTAINTIES OF EARTHQUAKE LOAD AND STRUCTURAL MODEL 

The parameters of structural model considered in this analysis are the viscous damping ratio and the 

hysteretic model parameters. The viscous damping ratio of brick structures can be expressed as [2]: 

where, A is the sum of horizontal cross sectional area of transversal walls for each floor. F is the area 

of structure for each floor. 

The hysteretic model is characterized by five parameters: the initial stiffness KO, the hardening 

stiffness KI, the softening stiffness KI, the crack strength Q c , and the ultimate strength Q u . These 

parameters can be determined by the experimental data. In this analysis, KQS KI and Q c are treated as 

independent random variables. In terms of KI and a stain hardening ratio otj, KI can be expressed as K 

= -aiKi. The relationship between Q u and Q c is more or less fixed, and its variability is very small. So 

it is omitted. There is not pinching effect in inelastic reloading stiffness of the hysteretic model. The 

degree of inelastic unloading stiffness has some effect on energy-dissipation under cyclic loadings, but 

it hasn't effect on the maximum deformation of structures. So it is reasonable to assume the unloading 

stiffness K o as deterministic. 

The coefficients of variation of KO, a, Q c and ^ for the brick walls with and without constructional 

columns are listed in table 2, respectively [2]. 

TABLE 2 

PARAMETRIC UNCERTAINTIES OF STRUCTURAL MODEL 

(4) 

Coefficient of variation 

of model parameters 

& 

Pa, 

Po. 

P< 

Brick wall without 

constructional columns 

0.30 

0.33 

0.30 

0.30 

Brick wall with 

constructional columns 

0.30 

0.42 

0.20 

0.30 

The artificial wave is obtained by transforming the mean response spectrum into the power spectrum. 

So the parameters of earthquake model include the mean response spectrum, the duration of 

earthquake ground motion and the damping ratio. We take the standard response spectrum in Chinese 

"Code for seismic design of buildings"(GBJll-89) as the mean response spectrum. The former is

522 

obtained by simply averaging a large amount of strong earthquake acceleration response spectra and 

making smooth. The coefficient of variation of the normalized mean response spectrum changes with 

period. But for general site and the range of period of multistory dwelling brick buildings, according to 

Ref. [8], it can be taken as PR = 0.26. The duration of strong motion is taken as the 1/2 peak 

acceleration duration i. By calculating 24 strong earthquake records longer than 2.75s from Ref. [8], 

we obtain that its mean value is 8.25s, and its coefficient of variation (3 T = 0.678. The standard response 

spectrum in the code is the one with the damping ratio of 0.05. When the structural damping ratio isn't 

0.05, the response spectrum is revised by the damping revising coefficient formula (12) specified in 

Chinese "Design code for antiseismic of special structures" (GB50191-93). 

In this paper, we don't consider the parametric uncertainties of laminated rubber bearings. 

5 THE EARTHQUAKE LOAD-STRUCTURE SYSTEM 

The Latin Hypercube Sampling technique is utilized to establish the earthquake load -structure system 

used in nonlinear time history analysis. In the analysis, four parameters describing the structural model 

and four parameters describing the earthquake model are considered. The uncertainties for each 

parameter are expressed in terms of three representative values, i.e., mean, mean minus and plus one 

standard deviation. So, for the ensemble of structural model, from the combinations of three 

representative values of the four parameters, a total of 81 structural models can be established. For the 

ensemble of earthquake time histories, from the combinations of three representative values of 

response spectrum and damping ratio, nine response spectra are obtained. For each response spectrum, 

three stationary time histories are generated. Thus, 27 stationary time histories are produced. It is noted 

that 27 different sets of random phase angles are used to generate these time histories. And then three 

envelope functions represented by strong motion duration are applied to each stationary time history to 

generate three normalized nonstationary time histories. Three strong motion durations are taken as its 

mean, mean minus and plus 0.8 times standard deviation [2]. Thus, a total of 81 normalized earthquake 

acceleration time histories are generated. Finally, using the Latin Hypercube Sampling technique, these 

earthquake time histories are matched to the structural models so that 81 samples of the earthquake 

load-structure system are constructed for seismic response analysis. 

6 STRUCTURAL VULNERABILITY ANALYSIS 

6.1 Limit States and Structural Capacity 

In this study, the limit state of structure is defined in terms of structural ductility factor, and five limit 

states representing initial crack damage, slight damage, moderate damage, severe damage and collapse 

of structure are considered. For each limit state, a corresponding capacity in terms of the ductility 

factor can be established. The ductility factor of the brick wall is defined as the ratio of the maximum 

TABLE 3 

DUCTILITY FACTOR CAPACITY 

Limit states 

Initial crack damage 

Severe damage 

Collapse 

Brick wall without 

constructional column 

PR °R 

1.0 0.3 

1.6 0.3 

2.6 0.3 

Brick wall with 

constructional column 

PR °R 

1.0 0.3 

2.6 0.3 

4.8 0.3

523 

deformation to the crack deformation. The structural capacity can be usually modeled by a lognormal distribution. 

According to Ref. [2], on the basis of the crack feature of wall in each hysteretic skeleton curve stage and compared 

with the true earthquake damage degree of buildings, the median jI R and logarithmic standard deviation C R can be 

obtained, as shown in table 3. The collapse capacities of brick wall with and without constructional columns are taken 

as 90% and 85% ultimate strength, respectively. The capacity for the severe damage is related to the ultimate strength. 

The capacity for the initial crack damage is related to the crack strength. The capacities for the slight damage and the 

moderate damage are taken as the values at 1/5 and 3/5 points between the capacities for the initial crack damage and 

the severe damage, respectively. The median and logarithmic standard deviation of story capacity of the brick 

buildings are obtained by composing the mean and standard deviation of the brick walls with and without 

constructional columns according to their cross section ratio. 

6.2 Probabilistic Response of Structures 

For each earthquake load-structure system, the nonlinear seismic response analysis is carried out. The 

i-th story ductility ratio is: 

IT 

*-^ (5) 

where, U max ,i is the maximum absolute inter-story deformation of the i-th story; U c ,i is the crack 

deformation of the i-th story. 

The statistical analysis is utilized to determine mean, standard deviation and distribution function of 

sample. The maximum structural response can be modeled by a lognormal distribution [2]. 

6.3 Vulnerability Analysis 

The structural vulnerability with respect to a particular limit state is defined as the conditional 

probability that structural response E exceeds the structural capacity R. The i-th story limit state 

probability of the buildings can be written as: 

P fi =P n (R,

524 

levels of peak ground acceleration. 

7 SEISMIC RELIABILITY ANALYSIS 

The structural seismic reliability analysis is usually to determine the limit state probability PF with 

respect to a particular limit state during the structural service life. It can be expressed as: 

The structural reliability index with respect to a particular limit state is. 

1=6 

where, k, is the occurrence rate of earthquake with intensity I, during the structural service life. For a 

given earthquake risk curve FI (I,), it can be written as: 

P t (Ii) is the conditional limit state probability with respect to a particular limit state for a given 

intensity I,, i.e., the structural vulnerability with respect to a particular limit state given above. 

The peak ground accelerations with respect to various intensities are taken as 0.05g> 0.10g^ 0.20gN 

0.40g and O.SOg for intensity VI, VI, VI, IX, and X, respectively. 

The analytical multistory dwelling brick buildings are assumed to be located in Tianjing, Beijing and 

Xichang. Their basic intensities are VI, VI, and K, respectively, and their earthquake risk curves may 

refer to Ref. [10]. 

Table4 lists the seismic reliability indexes of four multistory brick buildings with and without isolation 

located in region of intensity VI (Tianjing), VIII (Beijing) and EX (Xichang) for various limit states 

and their ratios. From this table, we can see that the ratios of the seismic reliability indexes of the brick 

buildings with and without isolation for various limit states are more and more large from low intensity 

region to high intensity region. This illustrates that the isolated brick buildings built in higher intensity 

region have better seismic performance than in lower intensity region. The seismic reliability indexes 

of a six-story brick building without isolation designed for intensity VI in region of intensity VI and 

a seven-story brick building without isolation designed for intensity VI in region of intensity VI for 

various limit states are slightly smaller than the seismic reliability indexes of the corresponding 

isolated brick buildings designed for intensity VI in region of intensity K and designed for intensity 

W in region of intensity VI, respectively. Thus, compared with the brick buildings without isolation, 

the isolated brick buildings can be designed with 1 intensity lower than the basic intensity, and the 

story number of the isolated brick building can be broken through the limit of the code, such as a 

six-story isolated brick building may be designed in region of intensity DC and a seven-story isolated 

brick building in region of intensity VI. As seen in Table 1, the plane and elevation of the seven-story 

brick building designed for intensity VU are the same as those of the eight-story brick building 

designed for intensity VI. The seismic reliability indexes of the seven-story brick building without 

isolation designed for intensity VI in region of intensity VI for various limit states are far smaller 

than those of the eight-story isolated brick building designed for intensity VI in region of intensity Vfl. 

Thus, compared with the brick buildings without isolation, it is no problem that the isolated brick 

buildings can be designed with 1 intensity lower than the basic intensity and increasing 1 story, and it 

can assure that the isolated brick buildings designed in this way have higher seismic safety.

525 

TABLE 4 

SEISMIC RJELIABILITY INDEXES 

Story 

number 

Design 

intensity 

4 

IX 

6 

V1I1 

7 

vn 

8 

VI 

Site 

intensity 

vn 

vm 

K 

Vfl 

WI 

IX 

VD 

WI 

IX 

vn 

VII 

IX 

Isolation 

Without (1) 

With (2) 

(2)/(l) 

Without (1) 

With (2) 

(2)/(D 

Without (1) 

With (2) 

(2)/(l) 

Without (1) 

With (2) 

(2)/(D 

Without (1) 

With (2) 

(2)/(l) 

Without (1) 

With (2) 

(2)/(l) 

Without (1) 

With (2) 

(2)/(l) 

Without (1) 

With (2) 

(2)/(l) 

Without (1) 

With (2) 

(2V(1) 

Without (1) 

With (2) 

(2)/(l) 

Without (1) 

With (2) 

(2)/(l) 

Without (1) 

With (2) 

(2)/(D 

Initial crack 

damage 

3.858 

4.702 

1.22 

3.105 

3.955 

1.27 

2.391 

3.380 

1.41 

3.073 

4.090 

1.33 

2.280 

3.361 

1.47 

1.373 

2.700 

1.97 

2.747 

3.979 

1.45 

1.933 

3.242 

1.68 

0.931 

2.559 

2.75 

2.586 

3.693 

1.43 

1.755 

2.935 

1.67 

0.687 

2.181 

3.17 

Slight 

damage 

4160 

4.823 

1.16 

3.390 

4.084 

1.20 

2.725 

3.527 

1.29 

3.345 

4.349 

1.30 

2.550 

3.595 

1.41 

1.700 

2.968 

1.75 

3.016 

4.234 

1.40 

2.200 

3.468 

1.58 

1.248 

2.819 

2.26 

2.757 

3.819 

1.39 

1.932 

3,067 

1.59 

0.912 

2.345 

2.57 

Limit state of structure 

Moderate 

damage 

4,458 

4.983 

1.12 

3.664 

4.267 

1.16 

3.043 

3.734 

1.23 

3.664 

4.594 

1.25 

2.881 

3.821 

1.33 

2.108 

3.229 

1.53 

3.290 

4.465 

1.36 

2.463 

3.673 

1.49 

1.566 

3.051 

1.95 

2.996 

4.006 

1.34 

2.178 

3.274 

1.50 

1.226 

2.594 

2.12 

Severe 

damage 

4.612 

5.0 

1.08 

3.836 

4.401 

1.15 

3.241 

3.885 

1.20 

3.884 

4.723 

1.22 

3.110 

3.967 

1.28 

2.390 

3.393 

1.42 

3.430 

4.575 

1.33 

I 2.611 

3.789 

1.45 

1.757 

3.187 

1.81 

3.160 

4.152 

1.31 

2.346 

3.433 

1.46 

1.431 

2.782 

1.94 

Collapse 

4.999 

5.0 

LOO 

4.286 

4.718 

1.10 

3755 

4.236 

1.13 

4.364 

5.0 

1.15 

3.584 

L 4.311 

1.20 

2.952 

3.783 

1.28 

3.763 

4.813 

1.28 

3.012 

4.069 

1.35 

2.260 

3.509 

1.55 

3.534 

4.534 

1.28 

2.749 

3.839 

1.40 

1.945 

3.255 

1.67

526 

8 CONCLUSIONS 

This paper gives the seismic reliability indexes of four multistory dwelling brick buildings with and 

without laminated rubber bearings in given regions of intensity YE, VI and IX by Latin Hypercube 

Sampling technique and nonlinear seismic time history response analysis. Through the comparison of 

the seismic reliability indexes of the brick buildings with and without isolation, three results are 

obtained as follows: 

(1) The isolated brick buildings built in higher intensity region have better seismic performance than in 

lower intensity region. 

(2) Compared with the brick buildings without isolation, the isolated brick buildings can be designed 

with 1 intensity lower than the basic intensity, and the story number of the isolated brick building 

can be broken through the limit of the code, such as a six-story isolated brick building may be 

designed in region of intensity IX and a seven-story isolated brick building in region of intensity 

VHI. 

(3) Compared with the brick buildings without isolation, it is no problem that the isolated brick 

buildings can be designed with 1 intensity lower than the basic intensity and increasing 1 story, 

and it can assure that the isolated brick buildings designed in this way have higher seismic safety. 

These results provide basis for the design of multistory isolated brick buildings. This method can also 

be used in the seismic reliability analysis of other isolated building structures. 

References 

[1] Hwang, H.M. and Jing- Wen Jaw (1990). Probabilistic Damage Analysis of Structures. J Struct 

Engrg 116:7. 

[2] ZHANG Ling-xin, JIANG Jin-ren and LIU Jie-ping(2002). Seismic Vulnerability Analysis of 

Multistory Dwelling Brick Buildings. Earthquake Engineering and Engineering Vibration 22:1, 

49-55. (in Chinese). 

[3] ZHANG Lingxin and JIANG Jinren (1997). Latin Hypercube Sampling and Its Application to 

Structural Reliability Analysis. World Information on Earthquake Engineering 13:4, 1-6. (in 

Chinese). 

[4] National Standard of the People's Republic of China: Code for seismic design of buildings 

(GBJ11-89), Architectural Industry Press of China, (in Chinese). 

[5] ZHANG Lingxin and JIANG Jinren (1998). Nonlinear Seismic Response Analysis of Multistory 

Brick Buildings. Proceedings oftheSth National Conference on Earthquake Engineering, Beijing, 

418-423. (in Chinese). 

[6] ZHANG Lingxin and JIANG Jinren (1997). Method of Elasto-plastic Dynamic Response Analysis 

Based on Pattern of Self-equilibrating Stresses. Earthquake Engineering and Engineering 

Vibration 17:4, 9-17. (in Chinese). 

[7] JIANG Jinren and HONG Feng (1984). Conversion Between Power Spectrum and Response 

Spectrum and Artificial Earthquakes. Earthquake Engineering and Engineering Vibration 4:3, 1-11. 

(in Chinese). 

[8] JIANG Jinren, LU Qinnian and SUN Jingjiang (1995). Statistical Characteristics of Strong Ground 

Motion Specified by Response Spectrum and Power Spectral Density Function. Earthquake 

Research in China 9:4, 387-402. Allerton Press, INC. /NEW YORK. 

[9] National Standard of the People's Republic of China: Design code for antiseismic of special 

structures (GB50191—93), Project Press of China, (in Chinese). 

[10] TAO Xiaxin etc. (2000). Engineering Design Earthquake Level and Seismic Motion Parameter. 

Personal communication.

SYSTEM IDENTIFICATION, MONITORING 

SYSTEM AND DAMAGE DETECTION




r 

STRUCTURAL HEALTH MONITORING FOR LARGE 

STRUCTURES USING AMBIENT VIBRATIONS 

Juan M Caicedo 1 , Enk Clayton 1 *, Shirley J Dyke 1 and Masato Abe 2 

Department of Civil Engineering, Washington University in St. Louis 

St. Louis, Missouri, USA 

2 Department of Civil Engineering, University of Tokyo, 

Tokyo, JAPAN 

ABSTRACT 

With the advent of more sophisticated computing software and hardware, the practicality 

of effectively implementing ambient vibration based global structural health monitoring 

algorithms has been greatly enhanced. These improvements have facilitated the implementation 

of systems to measure the dynamic behavior of large structures such as sky 

scrapers and long span bridges. These flexible structures have features that introduce additional 

challenges to the field of structural health monitoring. This paper examines the effectiveness 

of the Natural Excitation Technique (NExT) combined with the Eigensystem 

Realization Algorithm (ERA) to identify modal parameters in such structures. Simulated 

data from a mathematical model of the Bill Emerson Memorial Bridge is used. Ambient 

vibrations, modeled as stationary broadband inputs, are used for exciting the model. 

INTRODUCTION 

Virtually every facet of our daily activities depends on the reliability of our civil infrastructure. 

Long span bridges and tall buildings are costly to construct and can be critical to 

the smooth operation of our nations' infrastructure. The loss or unnecessary closure of key 

structures in the transportation network has severe consequences on regional or national 

economies. There is a clear need to continuously maintain the function of these structures 

well into the future. 

Presently, simple yet time consuming and labor intensive methods such as visual inspection 

and hammer echo are utilized in the field to monitor the health of structures such as 

highway overpasses and tunnel linings. There are limitations concerning the implementation 

of the these approaches, the most prominent being: the safety of those inspecting the 

structures; the difficulty of inspecting large or inaccessible parts of structures; and, the inefficiency 

of such time consuming approaches. As a result of the shortcomings found in 

these methods, mathematically based global structural health monitoring (SHM) tech- 

* Currently an undergraduate civil engineering student, Univ. of Tennessee, KnoxviIIe, Tennessee USA.

530 

niques for monitoring the health of structures have been borrowed from the aerospace industry 

to be refined and applied to civil infrastructure. These techniques facilitate the 

monitoring of structural health in an automated mode, reducing the uncertainty in the integrity 

of the structure and providing valuable information to decision makers after a major 

event. 

One important issue in SHM is obtaining appropriate structural response data. For best results, 

one would typically excite the structure with an actuator or shaker, and measure the 

responses. However, for many structures, using forced vibration to induce response for 

SHM or system identification purposes may be infeasible or forbidden (due to concerns of 

the owners or danger that the structure may already be damaged). However, ambient excitation 

(e.g., microtremor, wind, traffic, etc.) can often provide sufficient response for system 

identification to be performed. Further, ambient input excitation is often unmeasured 

(or actually immeasurable) and methods that require knowledge of the input must be selected. 

Additional challenges arise from the nature of large, flexible structures. The characteristics 

of these structures include: they are continuous structures that are not wellrepresented 

with lumped masses; lateral and torsional motions can be highly coupled; using 

a limited number of sensors is feasible; and, they have closely-spaced modes. 

The focus of this paper is to examine the capabilities of the Natural Excitation Technique 

(NExT) combined with the Eigensystem Realization (ERA) algorithm to extract modal 

parameters from a structure with closely spaced modes. A numerical model of the Bill 

Emerson Memorial Bridge, developed for the benchmark problem on the control of cablestayed 

bridges, is used to perform simulations and obtain response records. For this initial 

study, the bridge is excited using a stationary broadband force as a rough simulation of 

ambient vibrations. Three sensor configurations are considered to investigate the dependence 

of the methodology on the sensor configurations. 

DESCRIPTION OF THE BRIDGE AND FINITE ELEMENT MODEL 

The cable-stayed bridge used for this benchmark study is the Bill Emerson Memorial 

Bridge spanning the Mississippi River (on Missouri 74—Illinois 146) near Cape Girardeau, 

Missouri, designed by the HNTB Corporation (Hague, 1997). The bridge is currently under 

construction and is to be completed in 2003. Instrumentation is being installed in the 

Emerson bridge and surrounding soil during the construction process to evaluate structural 

behavior and seismic risk (Qelebi, 1998). 

As shown in Fig. 1, the bridge is composed of two towers, 128 cables, and 12 additional 

piers in the approach bridge from the Illinois side. It has a total length of 1205.8 m (3956 

ft). The main span is 350.6 m (1150 ft) in length, the side spans are 142.7 m (468 ft) in 

length, and the approach on the Illinois side is 570 m (1870 ft). A cross section of the 

deck. The bridge has four lanes plus two narrower bicycle lanes, for a total width of 29.3 

m (96 ft). The deck is composed of steel beams and prestressed concrete slabs. The 128 

cables are made of high-strength, low-relaxation steel (ASTM A882 grade 270). The H- 

shaped towers have a height of 102.4 m (336 ft) at pier 2 and 108.5 m (356 ft) at pier 3. 

Each tower supports a total 64 cables. The cross section of each tower varies five times 

over the height of the tower. The approach bridge from the Illinois side is supported by 11 

piers and bent 15, The deck consists of a rigid steel diaphragm with a slab of concrete at 

the top. Sixteen 6.67 MN (1,500 kip) shock transmission devices are employed in the connection 

between the tower and the deck. These devices are installed longitudinally to allow 

for expansion of the deck due to temperature changes. Under dynamic loads these 

devices are extremely stiff and behave as rigid links. Further details are provided in the paper 

describing the benchmark control problem statement Dyke et al, (2002),

531 

350.6m 142.7m 

.0130') i (468') ,. 

Bent ! 

Pier i 

0 Cable Number 

Figure 1. Drawing of the Emerson Bridge. 

Based on the drawings of the Emerson bridge, a three-dimensional finite element model of 

the bridge was developed in MATLAB® (1997). A linear evaluation model is used in this 

benchmark study. However, the stiffness matrices used in this linear model are those of the 

structure determined through a nonlinear static analysis corresponding to the deformed 

state of the bridge with dead loads (Wilson and Gravelle, 1991; Dyke et al, 2002). Additionally, 

the bridge is assumed to be attached to bedrock, and the effects of soil-structure 

interaction are neglected. 

The finite element 

model employs beam 

elements, cable elements 

and rigid links. 

The nonlinear static 

analysis is performed 

in ABAQUS® (1998), 

and the element mass 

and stiffness matrices 

are output to MAT- 

LAB® for assembly. 

Subsequently, the constraints 

are applied, Bent 

and a reduction is performed 

to reduce the 

Figure 2: Finite Element Model. 

size of the model to 

something more manageable. The finite element model, shown in Fig. 2, has a total of 579 

nodes, 420 rigid links, 162 beam elements, 134 nodal masses and 128 cable elements. The 

towers are modeled using 50 nodes, 43 beam elements and 74 rigid links. Constraints are 

applied to restrict the deck from moving in the lateral direction at piers 2, 3 and 4. Boundary 

conditions restrict the motion at pier 1 to allow only longitudinal displacement (X) and 

rotations about the Y and Z axes. The cables are modeled with truss elements. In the finite 

element model the nominal tension is assigned to each cable. 

The model resulting from the finite element formulation has a large number of degrees-offreedom 

and high frequency dynamics. Thus, some assumptions are made regarding the 

behavior of the bridge to make the model more manageable for dynamic simulation while 

retaining the fundamental behavior of the bridge. Static condensation is performed by first 

partitioning the mass and stiffness matrices corresponding to the structure DOF into active 

and dependent DOF (Fig. 2). Application of this reduction scheme to the full model of the 

bridge resulted in a 419 DOF reduced order model. The first 100 natural frequencies of the 

reduced model (up to 3.5 Hz) were compared and are in good agreement with those of the 

909 DOF structure. The damping in the system is defined based on the assumption of 

modal damping. The damping matrix was developed by assigning 3% of critical damping 

to each mode. This value was selected to be consistent with assumptions made during the

532 

design of the bridge. A more detailed description of the finite element model of the bridge 

can be found in Dyke et al, (2002). 

PROBLEM FORMULATION 

The objective of this research is to investigate the capabilities of the NExT/ERA technique 

for identification of the modal parameters of this structure (Caicedo et al., 2002). Herein, 

the excitation is assumed to be ambient, and is considered to be unknown for the identification 

procedure. Ambient inputs are simulated by applying uncorrelated, stationary 

broadband random forces at all nodes of the structure. The modal parameters are determined 

through acquisition of free response data from ambient responses and then identification 

of modal parameters. The methodology is described in the following sections. 

Acquiring Free Response Data 

Here it is assumed that the excitation to the system consists of ambient vibrations, which 

are not measurable. James et al. (1993) showed that the matrix of cross-correlation functions 

between the responses of the system and a response selected to be the reference response 

are a solution to the homogenous equation of motion. We refer to this step as 

application of NExT. Consider the equation of motion for a N degree of freedom classically 

damped linear system 

Mjc(f) + Cjc(/) + Kx(0 = f(r) 

where x is the N x 1 vector of displacements, M , C , and K are the N x N mass, damping, 

and stiffness matrices, respectively, and f(t) is the vector offerees acting on the system. 

Assuming the excitation and responses are each stationary random processes, Eq. (1) 

is written 

MX(t) + CX(r) -r KX(r) = F(t) (2) 

where X(t) is a displacement stochastic vector process and F(t) is the stochastic excitation 

vector process. Assuming that the structural parameter matrices are deterministic, 

postmultiplying Eq. (2) by a reference scalar response process X t (s) , and taking the expected 

value of each side yields 

, s) + CR^t, s) + KJR^f, s) = R FX (t, s) (3) 

where R(-) denotes a vector of correlation functions. Assuming that X, X and X are 

weakly stationary processes, and the excitation is uncorrelated with the responses, Eq. (3) 

can be written as (Bendat and Pierson, 2000) 

(I) 

+ Ktf^Cc) = 0 (4) 

Thus, the vector of displacement process correlation functions, R^r(t) , satisfies the homogeneous 

differential equation of motion. 

To implement this method, one of the available responses is selected as the reference signal, 

jc r (r) , the cross spectral density functions between the reference signal and each of the 

response signals are obtained, and an inverse fast Fourier transform is performed to determine 

the cross correlation functions. The reference channel should be selected such that 

all of the modes are observed in the responses at that location. If the reference channel lo-

533 

cation corresponds to a node of one of the modes, that mode will not be observed. This 

method allows the cross spectral density (CSD) functions to be averaged over a number of 

samples to increase the accuracy in the CSDs. Windowing should be used to minimize the 

effects of leakage. 

Identification of Modal Parameters 

Once the time domain free response data is obtained, there are numerous techniques available 

for identifying the modal parameters Here the ERA (Juang and Pappa, 1985) is 

adopted because it is quite effective for identification of lightly damped structures and is 

applicable to multi-input/multi-output systems. In the ERA, the Hankel matrix is formed 

y(k) 

(5) 

where y(k) is the response vector at the £th time step. The parameters s and r correspond 

to the number of columns and rows (of response vectors) in the matrix. This matrix is 

evaluated for H(0) and a singular value decomposition is performed, 

H(0) = PDQ 7 (6) 

Relatively small singular values along the diagonal of D correspond to computational 

modes and the associated rows and columns are eliminated to form the condensed D^v, 

P jV , and Q N matrices. The state matrix for the resulting discrete time system and the associated 

matrix in the output equation are found using is found using 

_! _1 _! 

A = D^ 2P/H(l)QtfDtf 2 C - K T P N D N 

2 

where E r = [I 0. .0]. Because this is a discrete time system, it is then transformed to 

the corresponding continuous time system. The natural frequencies are found by determining 

the eigenvalues of the continuous time matrix A, and C is used to transform the computed 

eigenvectors of the state matrix corresponding to the non-physical states in the 

identified model, to the values of the mode shapes at the floors of the structure. The ERA 

method was implemented in MATLAB (1997). 

NUMERICAL RESULTS 

The approach described in the previous section was implemented to identify the modal parameters 

of the structure from simulated acceleration records. Uncorrelated stationary 

broadband forces were used as the excitation in the lateral, longitudinal and transverse directions. 

The excitations were applied along the deck of the bridge. Thirty minute acceleration 

record with a sampling rate of 12.5 Hz was used for the identification procedure. 

Three different sensor configurations were considered to study the capabilities of the 

methods when the number of sensors is varied. Figure 3 shows the identification models 

used and the location of the sensors on the bridge for each case. Three sensors in each tower 

measure accelerations in the transverse and longitudinal directions. All deck sensors 

measure acceleration in the transverse and vertical directions. Case 1 uses a total of 128 

(7)

534 

* Reference Channel 

128 Sensors 

Case 1 

72 Sensors 40 Sensors 

Case 2 Case 3 

Figure 3. Sensor Locations. 

sensors, case 2 uses 72 sensors and Case 3 uses 40 sensors. 

Cross correlation functions were calculated as the inverse Fourier transform of the cross 

spectral density function, and treated as free response data. For these calculations the reference 

channel shown in Fig. 3 was selected. Frames of 512 points with 50% overlapping 

were used to determine cross spectral density functions. A representative cross spectral 

density and cross correlation function are shown in Fig. 4. Note that this structure may 

have many closed spaced modes at very low frequencies, as is typical of large flexible 

structures'such as cable stayed or suspension bridges. 

-190 

£-2001 

3-210 

W 

•S -220 

5 

.r: -230 

If -240 

*3 

< -250 

A 

i 

• / 

; 'S 'i „ 

CD 1 

5 

H 

I- 1 -2 

! ' ' 

Mn'«V ' " ^" * " "" —- vw 

ill 

• 25 °0 2 4 6 ~ M Q 5 . 10 15 2 


Time (s) 

Figure 4. Representative Cross Spectral Density and Cross Correlation Functions. 

The ERA was used to identify the modal parameters 

from the cross correlation functions. 

Figure 5 shows a typical singular values plot 

obtained with the ERA. In most structures a 

sharp drop in the values will occur, indicating 

a finite number of singular values appropriate 

for effective modeling of the structure (Juang 

and Pappa, 1985: Caicedo et al, 2001). In this 

case, this sharp drop is not present due to the 

closely spaced modes. Here, the correct natural 

frequencies and mode shapes are selected 

by examining the identified damping ratio. 

Only modes with damping ratios between 2.5 

and 3.5% are accepted. 

a-" 

•I- 

Singular Value No. 

Figure 5. Singular Values. 

Table 1 provides the identified natural frequencies and the error in the frequency with respect 

to the corresponding analytical value. Negative values indicates that the identified 

natural frequency was higher than the analytical value. Vertical responses are analyzed independently 

of transverse responses for this initial study, and thus torsional modes are not

535 

listed separately. Vertical modes are expected to have lower natural frequencies than transverse 

modes because the deck is stiffer in the transverse direction. As a result more vertical 

natural frequencies were identified. 

Note that when the number of sensors is decreased, the error in the identified natural frequencies 

increased, and the number of natural frequencies identified decreases. For the 

first two cases, ten vertical natural modes were identified, and for the third this number 

was reduced to 8. Also note that the first mode in Cases 2 and 3 corresponds to the third 

mode in Case 1, indicating that the two lower modes are not identified with the reduced 

sensor configuration. In the transverse direction five modes were obtained for the first 

case, and only three modes were found for the other two cases. Two pairs of identified and 

analytical (actual) mode shapes were compared to each other to demonstrate the accuracy 

of the results. Figure 6 shows a comparison between the identified and analytical mode 

shapes. Good agreement was found. 

Casel 

CO, (Hz) Error, (%) 

1.062 

1.258 

1.454 

1.632 

2.191 

2.559 

2.788 

3.380 

3.770 

4.156 

U.66y 

1.825 

2.636 

2.897 

3.725 

CONCLUSIONS 

Table 1: IDENTIFIED NATURAL FREQUENCIES. 

U.4SU 

-0.894 

-0.197 

0.684 

-0.317 

0.478 

0.397 

0.782 

1.245 

1.558 

-U.M.5 

-1.012 

0.302 

-0.733 

2.014 

Case 2 

0),(Hz) Error, (%) 

1 453 

1 638 

1.901 

2.190 

2.788 

3.380 

4.154 

4.565 

5.616 

5.866 

Vertical modes 

-u.iiy 

0.318 

L_ " L515 

-0.272 

0.420 

0.803 

1.606 

3.718 

-1.085 

2.009 

Iransverse Modes 

U.668 

1.824 

2.870 

—- 

-U.4&U 

-0.925 

0.187 

— 

- 

Case 3 

CO, (Hz) Error, (%) 

1.43U 

1.637 

2.191 

2.780 

3.383 

3.784 

4.143 

4.461 

- 

- 

1. 823 

2.856 

3.728 

— 

— 

U 116 

0.380 

-0.333 

0.681 

0.714 

0.865 

1.870 

5.896 

- 

- 

-u.y/3 

0.691 

1.941 

- 

- 

The NExT/ERA technique shows promise as a means for detenruning modal parameters 

to be used for structural health monitoring of large, flexible structures with closely spaced 

modes and low natural frequencies. This technique will permit monitoring large scale 

structures where ambient excitations (unknown input) are present as an excitation. This 

g « Identified 

B 

. ° 

Analytical .-*-* 

° * * * * e I » o 

« towers . 

Mode 76 - Vertical (2.184 Hz) 

Mode 7 - Horizontal (0.665 Hz) 

Figure 6. Identified vs. Analytical Modes.

536 

approach was applied to simulated data from a numerical model of the Emerson bridge 

under ambient excitation that is modeled as a stationary broadband force. The technique 

was successfully used to identify modal parameters of the structure for the different sensor 

configurations. The results indicated that fewer sensors resulted in the identification of 

fewer modes. 

Future investigations will consider a broader class of ambient vibrations, and use the modal 

parameters to identify structural parameters for damage detection. More information 

can be found at: http://wusceeLcive.wustl.edii/. 


This research is funded in part by the National Science Foundation (CAREER Grant No. 

CMS 9733272). The second author also acknowledges support from the Research Experiences 

for Undergraduates in Japan in Advanced Technology (NSF Grant No. INT 

0202630). 

REFERENCES 

ABAQUS (1998). Hibbitt, Karlsson & Sorensen Inc. Pawtucket, RI. 

Beck, J.L., May, B.S., and Polidori, D.C. (1994). "Determination of Modal Parameters 

from Ambient Vibration Data for Structural Health Monitoring," Proceedings 

of the First World Conference on Structural Control, Pasadena, CA, June. 

Beck, J.L., Vanik, M.W., Polidori, D.C., and May, B.S. (1998). "Structural Health 

Monitoring Using Ambient Vibrations," Proc. Struc. Engrs World Cong., Til8-3, 

San Francisco. 

Caicedo, J.M., Dyke, S.J., Moon, S.J., Bergman, L.A., Turan, G., and Hague, S. 

(2003). "Phase II Benchmark Control Problem for Seismic Response of Cable- 

Stayed Bridges," J. of Struc. Control, Wiley (submitted). 

Caicedo, J.M., Marulanda, J., Thomson, P., and Dyke S.J., (2001) Monitoring of 

Bridges to Detect Changes in Structural Health, Proc. of the 2001 American Control 

Conf, Arlington, Virginia, June 25-27. 

(Jelebi, M., (1998), Final Proposal for Seismic Instrumentation of the Cable-Stayed 

Girardeau (MO) Bridge, U.S. Geological Survey. 

Dyke, S.J., Caicedo, J.M., Turan, G., Bergman, L.A., and Hague, S. (2002). "Phase 

I Benchmark Control Problem for Seismic Response of Cable-Stayed Bridges," J. 

of Struc. Engrg, ASCE, (in press). 

Farrar, C.R. and James, G.H. Ill (1997). "System Identification from Ambient Vibration 

Measurements on a Bridge," Journal of Sound and Vibration, 205:1, 1-18. 

Hague, S. (1997). "Composite Design for Long Span Bridges." Proc. of the XV 

ASCE Structures Congress, Portland, Oregon. 

James, G.H. Came, T.G, and Lauffer, J.R (1993). "The Natural Excitation Technique 

for Modal Parameter Extraction from Operating Wind Turbines," SAND92- 

1666, UC-261, Sandia National Laboratories. 

Juang, J.N. and Pappa, R.S. (1985). "An Eigensystem Realization Algorithm for 

Modal Parameter Identification and Model Reduction." J. Quid Control and Dvn 

8, pp. 620-627. '' 

MATLAB (1997). The Math Works, Inc. Natick, Massachusetts.




SIMULTANEOUS ESTIMATION OF STRUCTURAL PARAMETER 

AND EARTHQUAKE EXCITATION FROM MEASURED 

STRUCTURAL RESPONSE 

J.Chen 1 - 2 , J.Li 2 and Y. L. Xu 1 

'Department of Civil and Structural Engineering, The Hong Kong Polytechnic University 

Hung Horn, Kowloon, Hong Kong, China 

Department of Building Engineering, College of Civil Engineering, Tongji University 

Shanghai 200092, China 

ABSTRACT 

This paper presents an element level system identification method in the time domain to 

simultaneously estimate structural parameters and earthquake-induced ground motion from measured 

structural responses only. The method, named the dynamic compound inverse method, combines the 

least-squares technique and the statistical average method together, leading to a proper iterative 

procedure for a better solution of the problem. The earthquake-induced ground motion is first 

conjectured using the measured structural responses and the assumed initial values of the structural 

parameters to be identified. The conjectured ground motion is then forced to comply with the 

dynamics of the given structure using the statistical average method. The modified ground motion is 

further used to provide a new estimation of the structural parameters using the least-squares technique. 

Repeating the iterative procedure until a preset convergence criterion is reached then provides the 

simultaneous estimation of structural parameters and earthquake excitation. Numerical example using 

a shear building is carried out to demonstrate the feasibility of the dynamic compound inverse method. 

The results show that the method can estimate the structural parameters and the ground motion 

satisfactorily for the cases that the structural responses are not polluted or slightly contaminated by 

noise. 

INTRODUCTION 

The frequency-domain modal approach is commonly used to identify natural frequencies, modal 

damping ratios, and mode shapes of a structure from measured structural responses without requiring 

the information on external excitation but assuming the external excitation to be a white noise random 

process. Though the natural frequencies and mode shapes are representative of the stiffness of a 

structure to some extent, it may not be possible to use the identified natural frequencies and mode 

shapes to quantify even a severe level of structural damage because they are not sensitive to the 

damage of individual structural members. Furthermore, earthquake-induced ground motion cannot be

538 

assumed as a white noise random process and most of existing buildings have no sensors to measure 

the ground motion directly underneath the building. Therefore, the direct identification of structural 

parameters at the element level in the time domain attracts more and more attention from researchers 

and engineers. 

As far as identification of a structure under seismic excitation is concerned, several techniques have 

already been proposed to estimate the structural parameters and input excitation from the measured 

response alone. Toki (1989) and Hoshiya (1995) solved this problem by assuming that the coda of the 

measured response time history represented the free vibration response of the structural system and 

estimated the structural parameters using the extended Kalman filter technique from this part of the 

measurements. The input ground motion was then estimated using the obtained parameters. However, 

it is difficult to decide the exact starting time of the free response part without knowing time history of 

the ground motion. Moreover, the performance of Kalman filter is sensitive to the initial condition 

selected. Wang and Haldar (1994) proposed another approach to simultaneously identify the input 

excitation and structural parameters in the element level. In their method, the input was initially 

assumed to be zero at the first four sampling time points, structural parameters were estimated at this 

four points using the least-squares technique. Then, the input information on all the sampling time 

instants was computed using the estimated parameters and measured responses. Afterward, structural 

parameters could be identified again using the estimated input excitation and measured responses of 

the full length. The iteration procedure was stopped when the identified input excitation at the first 

four points converged to a predetermined tolerance level. Later on, by combining with the extended 

Kalman Filter, this approach was expanded (Wang and Haldar, 1997) for the case that the response 

measurements of a structure were available at only limited locations. More recently, Cho (2000) 

improved this approach into a more generalized style and ameliorated the convergence of least-squares 

method by using multiple model QR decomposition algorithm. 

In this paper, the element-level time-domain approach suggested by Wang and Haldar (1994) is 

improved for simultaneous identification of earthquake excitation and structural parameters from 

measured structural responses using a combination of the least-squares-technique and the statistical 

average method. Numerical examples using shear type buildings are carried out to demonstrate the 

feasibility of the improved method, named the dynamic compound inverse method. 

DYNAMIC COMPOUND INVERSE METHOD 

Basic Equations 

The response of a N-story shear building subjected to a base acceleration X g (t) is governed by the 

following equation 

MX(t) + CX(t) -f KX(t) = -MX g (2.1) 

where M is the diagonal mass matrix of the building with all the elements provided. C and K, each 

of order N by N, are the damping and stiffness matrices that consist of structural parameters k,, c, 

(i=l,N) to be identified X(t), X(t) and X are the relative displacement, velocity and acceleration 

response vectors of the building with respect to the ground. Eq.(l) can be rearranged into the following 

form for identification.

539 

in which 

H = [H(t l ), H(t 2 ), -.., H(t L )] T (2.3) 

0 = [c 13 c 2 , -.., C N , k p k 2 , .... k N f (2.4) 

P = [P(t l ), P(t 2 ), »., P(t L )f (2.5) 

where H is the response matrix of velocity and displacement of the building. H is a rectangular 

matrix of order (L x N) x J, where L is the number of sampling points in the measured structural 

response time history and J is the number of unknown parameters. For the present case, J =2*N. 

Vector 9 contains all the unknown parameters and P is a vector related to the ground acceleration 

and building inertia forces. 

At any sample time instant t,, one has 

H(t,) = 

"*, 

0 

X 1 -x 2 0 

x 2 — x, x, — x 3 

0 

0 

x, x,-x, 0 0 

0 x,-x t x 2 -x 3 0 

(2.6) 

0 

0 0 

x n -x 

0 0 0 x n »x Q 

P(t 1 ) = [-m l x 1 (t l )-m l x g (t t ), -m.jXjtt.J-mjX^t,), •••, -m N x N (t l )-m N x g (t l )] r (2J) 

The responses of the structure are supposed to be measured in terms of displacement, velocity and 

acceleration. If the ground acceleration X g (t) is available, parameter estimation is trivial using, for 

instance, the least-squares technique (Hsia, 1977) as 

0 = [H T H] 1 H T P (2>8) 

However, the problem here is the ground acceleration is unknown and it is expected to simultaneously 

estimate with the structural parameters. In this connection, a dynamic compound inverse method is 

suggested below. 

Dynamic Compound Inverse Method 

The dynamic compound inverse method is actually an iterative identification procedure that consists of 

the least-squares technique for parameter identification and the statistical average method for forcing 

the identified input excitation to comply with the dynamic equilibrium of the concerned building. The 

implementation of the dynamic compound inverse method can be divided into the following steps. 

Stepl: Since both the structural parameters 0 and the ground motion X g (t) are unknown, initial 

values for the structural parameters have to be assumed, for instance, let 9 Q = [l, 1, •••, l] T . It will 

be demonstrated later that the performance of the suggested method is not sensitive to this initial 

assumption. The symbol ' A ' stands for the estimated value and the subscript 4 0* of 0 stands for the 

number of iteration times. 

Step2i From Eqn.2.2, the vector P can be estimated using # 0 and the measured structural responses, 

resulting in P = H0 0 . Then, comparing with Eqn.2.7, one may have

540 

where r denotes the r th story. 

t,), P(t 2 ), .-, P(t L )] r (2.9) 

P(t,) = [P,(t,), P,(t,),..., P N (t,)f ( 2 - 10 ) 

p r (t,) = -m r x r (t,)-m r k< rt (t,) r = l,...,N (2.11) 

Step3: From Eqn.2.11, at any sample time instant t,, the ground acceleration x g (t,) can be estimated 

from p r (t,) for each story as 

^r) (t l ) = x r (t 1 )-p r (t,)/m r r = l,.",N, i = l,-,L (2.12) 

According to the structural dynamics, all the N ground acceleration Xg r) (t,) obtained from Eqn.2.12 

should be the same at the time t, no matter which story it is estimated from. However, they may not be 

the same because of the difference between the assumed initial values and the actual values of the 

structural parameters. Therefore, a statistical average process is introduced to compute the average of 

x g r) (t,) to force the N ground accelerations to be the same. 

*.(t,) = Z*(O i = V",L (2.13) 

r«l 

Step4: Now, reconstruct the vector P in Eq. (2,7) using the averaged value x g (t,)and the measured 

structural responses, leading to 

p r (t I ) = -m r x r (t I )-m r I g (t 1 ) r = l,-,N (2.14) 

where the symbol '-'denotes the modified value. 

PW-faW, P 2 (U -, P N (t,)F (2-15) 

p^lSco, f(t a ), »., p(tjf (2.16) 

StepS: Eqns. 2.14 to 2.16 are different from Eqns. 2.9 to 2.10 in that after Step 4 the estimated ground 

acceleration time histories at each story are forced to be identical in Eqns. 2.14 to 2.16. Therefore, the 

improved estimation of the unknown structural parameters can be obtained by 

^ = [H T H]" I H T P (2.1) 

Step6: Replace # 0 by 0 { , and repeat Step 2 to 6 until the following convergence criteria are satisfied. 

max (2-18) 

( 2 - 19 ) 

where the subscript tier stands for the current iterative step, / means the /th element of vector 0; s e 

and Eg are the predetermined index for the structural parameter and the input vector, respectively. 

They are generally a small number of the magnitude between 10" 4 to 10" 6 . Clearly, the first criterion is 

imposed on the estimated structural parameter while the second criterion is imposed on the estimated 

ground motion. Once the iterations converge, the vector 0 obtained in Step 5 gives the final

541 

identification results of structural parameters while the averaged value x (t) obtained in stepS gives 

the time history of the ground motion. 

The key point of this improved method is the statistical average process in Steps 3 and 4, which 

actually provides a measure to convert the dynamics of structure under earthquake excitation into a 

mathematical condition that ensures the convergency of the iterative procedure. Moreover, the 

statistical average process is independent from the parameter identification algonthm chosen. In this 

connection, for special cases different parameter identification methods can be adopted instead of the 

least-squares method. 

NUMERICAL EXAMPLES 

The applicability of the dynamic compound inverse method is 

demonstrated in this section through a numerical example of a 3- 

story shear building (see Fig.3.1) under earthquake excitation (see 

Fig. 3.5) for both noise-free and noise contamination cases. The 

properties of the building, as used by Yang and Agrawal (2000 j for 

a high tech building, are mi=350,25,m 2 =262.29 and m 3 =175.13/0rc; 

ci=4369, C2-291.3 and c 3 =145.6&Vs/m; ki=4728400, k 2 =315230 

and k3-157610AW/m. The three natural frequencies of the building 

"' 

are computed as 3.447, 7.372 and 19.155 Hz, with the 

corresponding modal damping ratios are 1%, 2.14% and 5.56%, 

respectively. The peak acceleration of the input ground motion is 

120cm/s 2 . The sampling time interval of the ground motion is 

" s 

0.02s and the time duration is 80s, resulting in a total of sampling Fig3.1 3-story shear building 

points N=4000. The dynamic responses of the building are 

computed by the Wilson-0 algorithm and are then taken as the measured responses in the following to 

identify both the structural parameters and the ground motion. The computed (measured) relative 

displacement, velocity and acceleration responses of the top floor are depicted in Fig. 3.2(a) to (c), 

respectively. 

Time (sec) 

Time (sec) 

(a) Displacement (b) Velocity (c) Acceleration 

Fig.3.2. Computed dynamic responses of the top floor of the building 

Noise-free Case 

The measured dynamic responses without noise contamination are first used to identify the structural 

parameters and the* input acceleration ground motion with the dynamic compound inverse method. The 

identification results are summarized in TableS.l, in which the conditions in Cases 1 to 4 are all the

542 

same except for the initial values for the structural parameters selected in order to examine the 

sensitivity of the method to these assumed values. The initial values of the structural parameters for 

Cases 1 to 4 are given in TableS.l in bracket together with the identified results. Cases A to D in Table 

3.1 are the cases using the different number of sampling points L but with the same initial value. For 

all the cases, the convergence indices are set as 10~ 6 . For Casel, the convergence curve, which is 

defined as the variation of the identified value normalized by the actual value with the iterative time, is 

depicted in Fig.3.3 and Fig.3.4 for building damping and stiffness parameters, respectively. 

It is seen from Table 3.1 that the stiffness parameters identified by the dynamic compound inverse 

method are almost the same as the true values for all the cases including Case D where the sample 

points of only 200 are used. For the damping parameters, the maximum relative estimation error is less 

than 0.1%. It can be also seen that even for very poor initial values selected as in Case 4, the true 

structural parameters can be still identified. Thus, one may conclude that the accuracy of the identified 

method is not sensitive to initial values of the structural parameters selected. From the convergence 

curves shown in Fig.3.3 and Fig.3.4, it is clearly demonstrated that with the increase of iteration times 

the estimated parameters converge rapidly to the true values. Moreover, the input ground motion 

identified is found to be nearly the same as the real input acceleration. 

Case 

L 

Cl 

Ci 

C3 

ki 

k 2 

k* 

TABLE3.1 

IDENTIFICATION RESULTS FROM NOISE-FREE BUILDING RESPONSES 

2 

3 4 A B C 

4000 4000 4000 2000 1000 500 

4370.9(1000) 4370.9(-1) 4370.9(-0.32) 4370.9 4370.9 4369.7 

291.37(1000) 291.37(-1) 291.37(1000) 291.37 291.37 291.35 

145.60(1000) 145.59(-1) 145.58(1.2) 145.59 145.59 145.48 

4728337(1000) 4728370(1) 4728308(0.03) 4728306 4728306 4728482 

315210(1000) 315208(1) 315208(-1234) 315208 315208 315216 

157581(1000) 157581(1) 157581(23232) 157581 157581 157584 

1 

4000 

4370.9(1) 

291.36(1) 

145.59(1) 

4728307(1) 

315208(1) 

157581(1) 

L: Number of sampling points used 

D 

200 

4369.5 

291.39 

145.57 

4728495 

315219 

157586 

\ 

-.r 

c, 

t 

80 100 120 140 

Iteration times 

Fig.3.3 Convergence curve for damping 

coefficients 

Q 20 40 60 80 100 120 140 

Iteration times 

Fig.3.4 Convergence curve for stiffness 

coefficients 

Noise-polluted Cases 

To assess the capacity of the dynamic compound inverse method against measurement noise in the 

responses, the numerically generated zero-mean Gaussian white noise is added to the measured 

responses in two ways. In the first way, which is denoted as NCI (noise easel), the identical noise 

process is added to the displacement, velocity and acceleration responses of the same floor with the 

noise level controlled by the root-mean-square ratio between the noise process and the response time-

543 

history to be polluted. In the second way named as NC2, different noise processes are added to the 

dynamic responses of the same floor. As a result, the three different noise processes are simulated and 

used in NC1 while the nine different noise processes are simulated and used in NC2. Moreover, three 

noise levels of 1%, 3% and 5% are considered in NCI whereas NC2 considers three noise levels of 1%, 

2%, and 3%. The identification results are listed in Table 3 2 and Table 3.3 for NCI and NC2, 

respectively. 

TABLE3.2 

IDENTIFICATION RESULTS FROM NOISE-POLLUTED BUILDING RESPONSES (NCI) 

Noise Level 

1% 

Estimated Error 

3% 


5% 


Cj 

C2 

C3 

ki 

k 2 

k 3 

PGA 

4409.6 

294.73 

147.52 

4727325 

315159 

157559 

1.2097 

0.93% 

1.18% 

1.32% 

0.02% 

0.02% 

0.03% 

0.8% 

4510.6 

302.29 

151.58 

4726698 

315047 

157483 

1.229 

3.24% 

3.78% 

4.11% 

0.04% 

0.06% 

0.08% 

2.41% 

4642.6 

310.97 

155.92 

4727935 

314917 

157365 

1.248 

6.26% 

6.75% 

7.09% 

0.01% 

0.10% 

0.16% 

4% 

PGA(m/s~): Peak ground acceleration identified 

TABLE3.3 

IDENTIFICATION RESULTS FROM NOISE-POLLUTED BUILDING RESPONSES (NC2) 

XT . T , 1% 2% 3% 

iNUIbC l_,CVCi 

C] 

C2 

C3 

k, 

k 2 

k 3 

PGA 

Estimated 

4276.9 

283.66 

144.21 

4673818 

312894 

156761 

1.1952 

Error 

2.11% 

2.62% 

0.96% 

1.15% 

0.74% 

0.54% 

0.4% 

Estimated 

4325.7 

274.55 

140.69 

4569663 

307930 

155069 

1.1796 

PGA(m/T): Peak ground acceleration identified 

Error 

0.99% 

5.75% 

3.37% 

3.36% 

2.32% 

1.61% 

1.7% 

Estimated 

4514.2 

264.94 

135.52 

4488512 

302899 

151148 

1.15 

Error 

3.32% 

9.05% 

6.93% 

5.07% 

3.91% 

4.09% 

4.16% 

From Table3.2, one may see that the dynamic compound inverse method can accurately identify the 

stiffness coefficients even though the noise contamination in the responses reaches a 5% level. For the 

damping coefficients and the peak acceleration of ground motion, the capacity of the method against 

noise contamination seems to be relatively weak. The maximum identification errors of damping 

coefficient are 1.32%, 4.11% and 7.09% for noise level of 1%, 3% and 5%, respectively. The 

identified time history of input ground motion is shown in Fig. 3.6 for a noise level of 5% in NCI. It is 

seen that the identified one is quite close to the true one, with the maximum error of the peak 

acceleration of 4%. For NC2 where different noise processes are added to different types of structural 

responses at the same floor, the identification errors increase moderately for damping coefficients and 

input peak acceleration but considerably for stiffness coefficients when compared with NCI. The 

maximum identification errors in damping coefficients are 2.62%, 5.75% and 9.05% for noise level of 

1%, 2% and 3%, respectively. The maximum identification errors in stiffness coefficients are 1.15%, 

3.36% and 5.07% for noise level of 1%, 2% and 3%, respectively. The estimated ground acceleration 

motion for a 3% noise level in NC2 is plotted in Fig.3.7. The peak acceleration of the estimated ground 

motion is found to be lower than the true peak acceleration by 4.2%.

544 

Time (sec) 

Time (sec) 

Time (sec) 

Fie.3.5 Input ground acceleration Fig.3.6. Estimated ground motion Fig.3.7 Estimated ground motion 

motion for 5% noise level in NC1 for 3% noise level in NC2 

CONCLUSIONS 

The dynamic compound inverse method has been introduced in this paper for simultaneously 

identifying the structural parameters and the ground motion using the measured structural .responses 

only. Numerical examples using a three-story shear building showed that the method could accurately 

identify the structural parameters and the ground motion simultaneously when the dynamic responses 

were noise free. For the dynamic responses at the same building floor polluted by the same noise 

process, the compound inverse method could still accurately identity the building stiffness coefficients 

but the damping coefficients and the peak acceleration of the ground motion were estimated with some 

acceptable errors. If the dynamic responses at the same building floor polluted by different noise 

processes, the maximum identification errors increase moderately for damping coefficients and input 

peak acceleration but considerably for stiffness coefficients when compared with the case with the 

same noise process concerned. The extension and the further application of the method are under way. 


The writers are grateful for the financial support from Tongji University through a PhD studentship to 

the first writer and The Hong Kong Polytechnic University through Young Professor Scheme to the 

third writer. 

REFERENCES 

Cho H. N. and Paik S. W (2000), Time domain system identification technique with unknown input 

and limited observation, Proceedings of 8 th ASCE specially conference on Probabilistic Mechanics 

and Structural Reliability, PMC2000-190 

dough R.W. and Penzien J. (1993), " Dynamics of Structures", 2nd Edition, McGraw-Hill, Inc. 

Hoshiya M. and Sutoh A. (1995), Identification of input and parameters of a MDOF system, 

Proceedings ofEASEC-5, Gold Coast, Australia, 1309-1314 

Hsia T. C. (1977), System identification: least-squares methods, Lexington, Mass 

Toki K., Sato T. and Kiyono J. (1989), Identification of structural parameters and input ground motion 

from response time histories, Structural Eng./Earthquake Eng, 6:2, 413-421 

Wang D. and Haldar A. (1994), Element-level system identification with unknown input, J. of 

Engineering Mechanics, ASCE, 120:1, 159-175 

Wang D. and Haldar A. (1997), System idetification with limited observation and without input, J of 

Engineering Mechanics, ASCE, 123:5, 504-511 

Yang J. N. and Agrawal A. K. (2000), Protective system for high-technology facilities against 

microvibration and earthquake, J. of Structural Engineering and Mechanics, 10:6, 561-575




CRUSTAL DEFORMATION MEASUREMENT WITH 

SATELLITE RADAR INTEFEROMETRY: A REVIEW 

XL Ding, YQ Chen, ZL Li, GX Liu and ZW Li 

Department of Land Surveying and Geo-Informatics 

Hong Kong Polytechnic University 

Hong Kong 

ABSTRACT 

Satellite radar interferometry uses synthetic aperture radar (SAR) images acquired with 

satellite-borne radars to accurately reveal tiny ground deformations or changes. Due to the 

unprecedented high resolution, accuracy and ground area coverage of the technology, it has 

fast become one of the most important technologies for the measurement of earthquake 

related crustal deformations. Results from such measurement can potentially provide 

information on stress changes in the crust that are crucial to understand the development of 

earthquakes. 

This paper briefly examines the capacity and limitations of the technology, its applications in 

earthquake related research, and its potential future development.




VISION-BASED SENSORS 

FOR MONITORING SEISMIC DEMANDS 

Tara C. Hutchinson 1 and Falko Kuester 

'Department of Civil & Environmental Engineering 

Department of Electrical and Computer Engineering 

University of California, Irvine USA 92697 

ABSTRACT 

A vision-based approach is evaluated for its applicability as a new sensing technology for measuring 

earthquake induced motions. Traditional motion sensors used in laboratory and field experiments must 

be physically attached to the structure and require cumbersome cabling, configurations and substantial 

time for set-up. Moreover, for reduced scale experiments, sensors generally add substantial mass or 

change the response characteristics of the system. The approach evaluated in this paper is 

advantageous since it requires very limited physical attachment to the structure of interest, is highspeed, 

high-resolution and does not introduce additional mass or otherwise modify the properties of the 

structure. In this exploratory phase, four digital high-speed, high-resolution cameras outfitted with redlight 

emitters are used to track reflective (nearly) mass less spherical elements discretely mounted on a 

scale 5-story steel frame structure. The structure is mounted on the University of California, Irvine's 

bi-axial shake table and subjected to different earthquake motions. The structure is also instrumented 

with a total of eleven conventional (wired) transducers (LVDTs and accelerometers) providing a 

unique comparison with the vision-based approach. Results from this exploratory study show that the 

non-intrusive vision-based approach is extremely promising in terms of its ability to capture inter-story 

drift, floor level velocities and accelerations, provided proper post-processing of the data occurs. 

INTRODUCTION 

Alternative methods of tracking seismic motions are desirable, particularly in the laboratory setting 

where scale models are often used to study earthquake response. Many fields, besides earthquake 

engineering, require precise high-speed motion tracking and thus new technologies are rapidly 

becoming available. Biomechanics, human gait analysis, robotics, virtual reality (VR), gaming and 

even entertainment have successfully employed a variety of new motion tracking techniques. Typically 

these applications require accurate six degree-of-freedom tracking, expressed by position (x, y, z) and 

orientation (yaw, pitch, roll). In earthquake engineering, generally the three positional degrees-offreedom 

are the most important motions to track and often only one of these dominates depending 

upon the problem. However, high spatial and temporal resolution (high sampling rates), large working 

volumes, limited instrumentation time and low-cost are also required. Depending on the particular

548 

environment constraints and the configuration of the system to be tracked, the required accuracy can 

range anywhere between millimeters to fractions thereof. Perhaps the most difficult problem faced in 

earthquake engineering experimental research is the cumbersome physical attachment and associated 

lengthy set-up times of most conventional motion sensors. For reduced scale experiments, these 

sensors generally add substantial mass or stiffness and therefore change the response characteristics of 

the system. 

In this paper, different motion tracking technologies are reviewed and a vision-based approach is 

selected and evaluated for its applicability as a new sensing technology for measuring earthquake 

induced motions. The approach is advantageous since very limited physical attachment to the structure 

of interest is needed, is high-speed, high-resolution and does not introduce additional mass or 

otherwise modify the properties of the structure. An exploratory study is conducted using four digital, 

high-speed, high-resolution cameras outfitted with red-light emitters to track reflective (nearly) mass 

less spherical elements discretely mounted on a scale 5-story steel frame structure. The structure is 

mounted on the University of California, Irvine's (UCIs) bi-axial shake table and subjected to different 

earthquake motions. The primarily uni-axial dominated motion of the frame provides a simplified 

reference expenment with reduced degrees of movement for evaluating this approach. The structure is 

also instrumented with a total of eleven conventional (wired) transducers [linear variable displacement 

transducers (LVDTs) and accelerometers]. A review of motion tracking technologies and specifically 

vision-based technologies is provided, followed by a description and select results from the exploratory 

studv conducted. 

TYPES OF MOTION TRACKING TECHNOLOGIES 

Specific types of motion tracking technologies include: (1) mechanical, (2) magnetic, (3) acoustic, (4) 

optical, (5) and hybrid solutions combining two or more of the previously mentioned techniques to 

achieve higher accuracy or sampling rates. Mechanical tracking techniques such as linear displacement 

transducers and linked boom structures (in VR) provide excellent results for linear and rotational 

movement measurements at the cost of limited working volumes, degrees-of-freedom, added mass and 

friction. They also generally require physical attachment or point-wise anchorage to elements of 

interest. Electro-magnetic techniques are less constraining and are based on generation of a magnetic 

field by an emitter and collection by a receiver. Used prevalently in VR, the transmitter system is 

generally statically mounted while the receivers can be freely moving within the field. While these 

systems can operate at nearly 150Hz sampling rates at a static accuracy of less than two millimeters 

and 0.5 degrees under ideal conditions, they are sensitive to metal components and other magnetic 

fields located in the workspace. This noise introduced by the environment can greatly distort the 

electromagnetic reference signal making it difficult to use in earthquake testing facilities. While these 

systems remove the mechanical constraint mentioned earlier and reduce sensor mass at the same time, 

currently a wired connection is still required. Acoustic or ultrasound techniques use time of flight 

information to triangulate the position of the target object, which is instrumented with active sensors. 

Generally, multiple emitters are mounted in a known pattern on the target and broadcast a signal in 

sequential order. The time of flight from the target to receivers mounted throughout the environment is 

then used to triangulate its position. Potential problems include ambient ultrasonic noise produced by a 

variety of electronic devices and slow update rates. 

Optical tracking techniques come in different forms and are often termed vision-based systems. Early 

and very extensive usage of optical tracking technologies can be traced to robotics and subsequently to 

the medical field for biornechanics and human motion studies with a strong focus on gait analysis [e.g.

549 

Cappazzo (1984), Heyn et al (1996), Kidder et al (1996), Sampath et al. (1998) and Hansen et al. 

(2002)]. The ability to carefully analyze and 

characterize a patient's manner and rate of 

movement (gait) has greatly aided in the 

development of treatment options for the physically 

handicapped. Requirements of very precise 

measurements, extended workspaces and limited 

physical constraints are similar to those faced in 

many civil engineering applications. An example of 

a human motion study (applied to dance) conducted 

using an optical (light)-based approach is shown in 

Figure 1. 

Vision-Based Systems 

Vision-based systems may be classified into either 

image- or light-based. Image-based systems in the 

context of motion tracking rely on feature detection 

between frames of a color texture map. In parallel 

with feature detection, image processing and 

reconstruction must be conducted, a challenging task 

for resolving the tracking to the level of resolution 

required for seismic motions. 

FIGURE 1. Example of human motion study 

(dance) using light-based approach (courtesy of 

Motion Analysis). 

Early forms of light-based tracking were based on emitter/receiver systems using arrays of light 

emitting diodes (LEDs) mounted statically at specific locations in a test space and a CCD camera 

(charge-coupled device) responsible for recording the created reference pattern. Given the exact 

spacing and position of the LEDs, the relative position of the CCD can then be triangulated. More 

recent techniques have removed the need for emitter/receiver pairs by directly analyzing image (video) 

data. The most popular approach currently uses reflective markers to identify points of interest within 

the environment that should be tracked, significantly reducing the required processing time. In this 

case, a range of wavelengths of light are filtered out, thus decreasing the required amount of 

information to be collection, allowing higher speeds and resolution to be captured. The CCD cameras 

are then used to measure the light intensity at each pixel. A strobe constructed of a cluster of highintensity 

LEDs is used on a per camera basis to illuminate the scene with red light. This strobe can 

easily illuminate reflective markers at distance between 2-25 meters. 

During the data acquisition stage only raw data (images showing the marker position) is acquired on a 

per camera basis. Each marker is represented by multiple pixels in the final image. Before the spatial 

position can be calculated, the marker "blobs" have to be approximated (by ellipses or circles) and 

their respective centroids determined. The position of the corresponding points can then be matched 

between image pairs obtained from two cameras and triangulated to obtain their 3D position. The 

corresponding markers between two images can be found based on the epipolar constraint, which 

states that a point in the first image must lie on the epipolar line in the second. (This reduces the 

matching problem from an area to a line segment). Given the projection of a marker into one image 

plane at p l and into another image plane at p 2 , the epipolar constraint is expressed by: 

•F- Pl =0 (1)

550 

where F is the fundamental matrix (Faugeras 1993). The 3D position of the markers can then be 

computed from the acquired marker sets using the intrinsic and extrinsic parameters of the camera 

system (e.g. position, focal length). Positional information (x,y,z) on a per marker basis and orientation 

(yaw, pitch, roll) is available through groups of markers aligned along two different axes. 

EXPLORATORY STUDY 

An exploratory study was conducted to investigate the applicability of a vision-system for capturing 

the response of a specimen mounted on UCIs bi-axial shake table. In this case, a scale 5-story steel 

moment frame structure with a dual damping system (Villaverde et al. 2002) subjected to uni-axial 

seismic motions provides a simple structure for evaluation of this new technology. The frame structure 

is 2.44 m in height and mounted with a roof isolation system. Figure 2(a) shows an elevation of the 

specimen mounted on the shake table with strategically placed spherical tracker elements (which show 

up as bright reflections in the photograph). The tracker elements (markers) are 12.5 mm diameter 

Styrofoam spheres wrapped in reflective tape. Compared with the mass of the specimen and the 

conventional transducers, the mass of the tracker elements is negligible. Four high-resolution 

(1280x1024 pixel) CCD cameras (Vicon, Oxford Metrics, Oxford, England) with maximum capture 

rates of 250 Hz were mounted on tripods and strategically placed facing the frame structure. The 

cameras were rigidly mounted with light emitting strobes. 

The model building is instrumented with a total of six LVDT displacement transducers, one attached at 

each fluid viscous damper (brace element) and one at the roof damping system. Five accelerometers 

were installed at the individual floor levels. In total, 21 spherical tracker elements were mounted at 

floor levels and at several locations on the base of the shake table for reference, as shown in Figure 

2(a). The use of the mass less spherical trackers allows multiple measurements for comparison and 

potential averaging. Figure 2(b) shows a detailed photograph of the roof level including selected 

placement between the roof and isolation system of the trackers. It took approximately 30 minutes to 

array the trackers in strategic positions on the frame. 

It is important to note that the LVDTs at the damper level provide an additional resistance to the 

damper element and overall system response, which is large relative to the scale of the model 

specimen. Although it may not be evident from the photograph, a roof isolation system is being studied 

in this work (Villaverde et al. 2002), thus the desired results are a comparison between two systems 

(with and without roof isolation), where each includes engaging the damping at the braces. Therefore, 

for these particular experiments, the additional resistance due to placement of the LVDT in-line with 

the brace is not important. 

The cameras were evenly spaced (at approximately 2 meters on center) along the line of action of 

lateral input seismic motion. They are aligned such that at least three of the cameras can observe any of 

the desired points at any given time. The closest camera was approximately 5.5 m, and the farthest 

camera was 6.0 m from the center of the specimen. 

Since positional information is computed by triangulation of the image data obtained by each camera, 

careful calibration is required. Calibration should ideally occur prior to running the tests but can be 

reapplied to the acquired raw data should a better calibration be required at a later time. Two phases of 

calibration were conducted: (1) static calibration and (2) dynamic calibration. Static calibration is 

achieved by placing a rigid L-frame within the static viewing volume. Markers are mounted on

551 

precisely defined distances on the rigid L-frame. Dynamic calibration is achieved by sweeping a wand 

with a marker mounted at each end through the desired viewing volume. For this exploratory study, the 

system was statically calibrated to be within 1 mm positional resolution and dynamically calibrated 

such that each of the camera's were within 0.5 mm precision of each other. 

Once the system is calibrated, the entire environment can be instrumented with an arbitrary number of 

passive markers. However, careful consideration must be given to the density of marker placement as 

related to the resolution of the camera system. If makers are placed too close to each other it will be 

difficult to distinguish them in the obtained image. In addition, if markers are placed behind an element 

in the scene, they will be occluded during motion capture. This can be challenging since elements 

within the scene move thus similar to conventional transducers, the magnitude and direction of 

movement must be anticipated. 

EXPERIMENTAL RESULTS 

Two earthquake motions were used and multiple runs were conducted (with and without the roof 

isolation system) to evaluate the accuracy and repeatability of the motion tracking system. In total, 

seven shake table simulations were conducted. The first input motion was a modified version of the 

SCT acceleration record from the 1985 Mexico City earthquake. The second motion was a modified 

form of the Takatori acceleration record from the 1995 Kobe earthquake. Each of these motions were 

tuned to the fundamental frequency of the model building (f = 2.375 Hz) (Villaverde et al. 2002) and 

applied as a base excitation to the structure. 

Figure 3 shows a comparison of inter-story drift ratios at the roof level of the model structure, 

measured using the vision-based approach and the LVDTs. In this case, the difference in maximum

552 

drift ratio was determined as Ymax = 

0.3% and 0.2%, for the structure with 

and without the roof isolation system, 

respectively. This corresponds to a 

difference in maximum drift of 1.3 

and 0.8 mm, respectively. Table 1 

summarizes the difference in 

maximum inter-story drifts obtained 

from each of the measurement 

technologies for each of the 

conditions considered by Villaverde et 

al. (2002). The largest difference 

between the vision-based system and 

the LVDTs was 2.8 mm. However, 

the average difference between 

maximum measured inter-story drifts 

was fairly reasonable at 1.4 mm and 

the standard deviation from these data 

was 0.8 mm. It is interesting to note 

that in all cases the vision-based 

system measurements were larger 

than that of the LVDTs. The visionbased 

measurements may have 

compounding errors, since inter-story 

drift is determined as the difference 

from several different markers, 

whereas the LVDTs were designed to 

directly measure inter-story drift. It 

should be noted that positional data 

obtained at the floor level was taken 

as the average from two spherical 

trackers mounted at each floor level 

[one mounted to the North, one 

mounted to the South, as shown in 

Figure 2(b)]. 

Acceleration and velocity time 

histories at the floor levels were also 

evaluated. Signal processing and 

numerical differentiation of the 

positional data was required for 

calculating acceleration and velocity 

time histories from the vision-based 

(positional) measurements. Two 

aspects of high frequency noise were 

observed in the vision-system 

positional data, requiring frequency 

domain filtering: (1) that imposed 

artificially from the reaction floor in 

Conventional LVDTs 

- Vision-based Measurem 

10 

Time (seconds) 

W 


(A) ~ 

FIGURE 3. Comparison of vision-based tracking system and 

conventional LVDT measurements for a 5-story steel frame 

model subjected to the Mexico City input motion: (a) with 

roof isolation system and (b) without roof isolation system. 

TABLE 1. Summary of difference in maximum measured interstory 

drift using the vision-based system and LVDTs. 

Earthquake Motion and 

Structure Condition 

Mexico City - toith roof isolation 

Mexico City - without roof 

isolation 

Takatori - without roof isolation 

Takatori - with roof isolation 

Floor 

Level 

5 

4 

3 

2 

1 

5 

4 

3 

2 

1 

5 

4 

3 

2 

1 

5 

4 

3 

2 

1 

Difference in 

Max. Drift (mm) 

1.3 

0.8 

0.5 

0.0 

1.4 

0.8 

1.2 

1.0 

1.3 

2.2 

0.0 

1.0 

1.8 

0.5 

2.3 

1.5 

2.7 

2.6 

1.7 

2.8

553 

the laboratory and (2) that imposed during numerical differentiation due to time stepping. 

High frequencies introduced during testing occurred because the cameras were mounted on the 

reaction floor surround the shake table and the shake table itself induces some motion into the 

surrounding floor during earthquake simulation. The magnitude of this noise was consistently observed 

to be small and outside of the frequencies of interest (> 50 Hz), thus it could easily be filtered in the 

frequency domain using a 4 th Order low-pass Butterworth filter. 

Double numerical differentiation also creates noise in acceleration traces. In this case, a 10 th Order 

low-pass Butterworth filter was applied after numerical differentiation to remove introduced high 

frequency noise. Low pass filters of comparable details were also applied to the conventional 

accelerometer measurements. 

Figure 4 shows a comparison of 

acceleration and velocity time 

Iristories (absolute) measured from 

the 5 th floor (roof level) of the 

structure during the Mexico City 

event with the roof isolation system 

engaged. Accelerations are 

compared between the double 

differentiated vision-system 

positional data and the conventional 

accelerometers. Velocities are 

compared between the single 

differentiated vision system data 

and integrated accelerometer 

measurements. These comparisons 

are very promising, with differences 

in maximum acceleration of 0.06g 

and differences in maximum 

velocity of 6 cm/second observed. 

Analysis of other floor levels and 

the different experiments showed 

that this was the largest difference 

between the two systems. However, 

_^ 

o 

to 


FIGURE 4. Comparison of velocity (a) and acceleration (b) time 

histories (absolute) at the roof level - Mexico City input motion 

with the roof isolation system engaged. 

it should be noted that selection of cut-off frequencies during the data processing was an important step 

in the evaluation of each case due to the different frequencies induced by the earthquake motions. 

SUMMARY 

A vision-based motion tracking system was evaluated for its accuracy in monitoring the seismic 

demands induced in a 5-story scale moment resisting frame structure. Reasonable comparison between 

time-varying positional responses was observed when comparing the vision system with conventional 

LVDT measurements. Although the system was calibrated statically to within a 1 mm displacement, 

the average difference in maximum displacements was observed as 1.4 mm. Signal processing and 

numerical differentiation of the positional data was required for 'calculating acceleration and velocity 

time histories from the vision-based (positional) measurements. Two aspects of high frequency noise

554 

were observed in the vision-system positional data requiring frequency domain filtering: (1) that 

imposed artificially from the reaction floor in the laboratory and (2) that imposed during numerical 

differentiation due to time stepping. Using low-pass filtering techniques, reasonable acceleration and 

velocity time histories were generated. The largest difference in maximum accelerations and velocities 

observed was fairly reasonable at 0.06g's and 6 cm/second, respectively. 

The initial comparison of measurements obtained from the vision-based system and conventional 

sensors in this study provides promising results for the use of such systems in future experimental 

earthquake research. Further study is required to evaluate the applicability of these types of systems for 

capturing seismic motions in 3-dimensions, including the potential effects of occlusions during the 

time-variant capture process. 


The experimental data from the study of Dr. Roberto Villaverde are greatly appreciated. The assistance 

of Mr. Charlie Hamilton and Mr. Bob Kajanzy of the Structural Engineering Test Hall (SETH) at UC 

Irvine is greatly appreciated. 

REFERENCES 

Cappozzo, A. (1984). "Gait analysis methodology." Human Movement Science. Vol. 3: 27-50. 

Faugeras, 0. (1993). Three Dimensional Computer Vision: a Geometric Viewpoint. MIT Press. 

Hansen, A.H., Childress, D.S., and Meier, M.R. (2002). "A simple method for determination of gait 

events." Journal ofBiomechanics. Vol 35: 135-138. 

Heyn, A., Mayagoitia, R.E., Nene, A.V., Veltink, P.H. (1996). "The kinematics of the swing phase 

obtained from accelerometer and gyroscope measurements." In the Proceedings of the 18 th Annual 

International Conference of the IEEE Engineering in Medicine and Biology Society. Amsterdam, pp. 

463-464. 

Kidder, S.M., Abuzzahab, F.S., Hams, G.F. (1996). "A system for the analysis of foot and ankle 

kinematics during gait." IEEE Transactions on Rehabilitation Engineering. Vol. 4(1): 25-32. 

Motion Analysis (2002). http-//www.motionanalysis.cnrn. 

Sampath, G., Abu-Faraj, Z.O., Smith, PA. and Harris, G.F. (1998). "Design and development of an 

active marker based system for analysis of 3-D pediatric foot and ankle motion." In the Proceedings of 

the 2V h Annual International Conference of the IEEE Engineering in Medicine and Bioloey Society 

Vol 20(5): 2415-2417. 

Vicon (2002). http://www.vicon.com. 

Villaverde, R., Aguirre, M. and Hamilton, C. (2002). "Roof isolation system implemented with steel 

oval elements: exploratory study." Proceedings of the 3 rd World Conference on Structural Control, 

Como, Italy. April 7-12.




555 

UNSCENTED PARTICLE FILTER FOR TIME DOMAIN 

IDENTIFICATION OF NONLINEAR STRUCTURAL DYNAMIC 

SYSTEMS 

K.Y. Koo and C.B. Yun 


Korea Advanced Institute of Science and Technology, Daejoen, Korea 

ABSTRACTS 

In this study, a recently developed unscented particle filter (UPF) technique is studied for 

identification of nonlinear structural dynamic systems. It is well-known that there is no general 

solution for parameter identification of nonlinear structural system. As an alternative in the sense 

of linear approximation solution, the Extended Kalman filter (EKF) has been frequently employed 

for identification of time-varying structural parameters. However, the EKF has several drawbacks 

such as biased estimations and erroneous estimations especially for highly nonlinear dynamic 

systems due to its crude linearization scheme. To overcome the weakness of the EKF, the UPF was 

recently developed. The UPF is a novel method for nonlinear, non-Gaussian, and on-line 

estimation. The algorithm is a particle filter that uses an unscented Kalman filter (UKF) to generate 

the importance proposal distribution. 

Numerical simulation studies have been carried out on SDOF and MDOF systems. The results on 

the linear and nonlinear SDOF systems show that the UPF gives more accurate and robust estimates 

under the existence of the system and measurement errors with rough initial guesses for the states 

and the error covariance matrices. The results on a five-story building structure subjected to 

nonlinear behavior at the bottoms of the columns show that the UPF has good estimation capability. 

The results from a series of numerical simulations indicate that the UPF is superior to the EKF for 

the system identification of nonlinear dynamic systems especially for highly nonlinear systems. 


Detection of structural change or damage is one of the most important and challenging issues in 

the structural health monitoring system. Statistical inference approaches are more appropriate in 

the detection of the structural change due to the complex nature of civil infrastructures and 

noise-polluted measurements. Especially Bayesian approach in statistical inference has useful 

structure in on-line estimation and has been applied to various fields, i.e., physics, control system 

and signal processing.

556 

The most well-known Bayesian solution is the Kalman filter which is complete and tractable 

Bayesian solution for linear dynamic system by making linear and Gaussian assumptions to 

simplify the optimal recursive Bayesian estimation Unfortunately, the problem of estimating 

the structural change often has nonlinear dynamics so the extended Kalman filter (EKF) was 

applied that is approximation solution for nonlinear dynamic system in a sense of piecewise 

linear approximation and has been reported on its successful applications in numerous state 

estimation problem But assumptions on extended Kalman filter may be violated in highly 

nonlinear systems. 

Particle filtering techniques are general approach in estimations of nonlinear dynamic system 

without assumptions on the form of the probability densities in question, that is, they employ 

full nonlinear and non-Gaussian estimation Recently the unscented particle filter has been 

developed by Wan and Merwe[5] that utilizes the unscented Kalman filter (UKF) to augment 

and improve the standard particle filter, specifically through generation of the importance 

proposal distribution In this study, the performance of the unscented particle filter was 

evaluated being a more general filtering technique for detection of structural change in civil 

infra-structures in comparison with the extended Kalman filter 

2. UNSCENTED PARTICLE FILTER 

Bayesian approach 

Given measurements of inputs u k and structural response y k , the goal is to estimate the 

parameters of the structural dynamic system. The optimal estimate in a sense of minimum 

mean-squared error (MMSE) is given by the conditional mean as follows 

1 = E[^|Y OA ] (2.1) 

where Y Q k is the sequence of observations up to time k. Evaluation of this expectation 

requires information of the a postenor density p(X k | Y OA ). The problems of determine 

posteriori density given a priori density p(X k | YQ^) is referred as the Bayesian approach 

and can be evaluated recursively according to the following relations 

where 

, x 

(22) 

P(Xt I YO*.!) = p(X k | Xî)p(X k^l | Xu^X**-! (2.3) 

and the normalizing constant p(Y k \ Y 0 A-l ) is given by 

P0t i Y 0 *..!) = J p(X k | Y 0 k _j) p(Y k | X k ) dX k (2 4) 

This recursion specifies the current state density as a function of the previous density and the 

most recent measurement data. The state-space model comes into play by specifying the state 

transition probability p(X k \ x^) and measurement probability or likelihood p(Y k I X k ) 

Unfortunately, the multidimensional integration indicated by Eqn (2 2) Make a closed-form 

solution intractable for most systems. The only general approach is to apply Monte Carlo 

sampling techniques that essentially convert integrals to finite sums, which converge to the true 

& 

solution in the limit.

557 

Unscented particle filter 

The choice of the importance proposal distribution q(X h \X Qk ,Y Qk ) is most important and 

critical design issue for filter performance The optimal importance proposal distribution which 

minimizes the variance on the importance weights is given by 

q(X k \ X 0k^Y Qk ) = p(X k I X Q^Y ok ) (2.5) 

That is, the true conditional state density given the previous state hasty and all observations 

Sampling form this distribution is impractical for arbitrary densities Consequently, the 

transition prior is the most popular choice of importance proposal distribution 

g(X k 1 X Q ,_!,r 0k ) = p(X k I XM) (2.6) 

The effectiveness of this approximation depends on how close the importance proposal 

distribution is to the true posterior distribution. If there is not sufficient overlap, only a few 

particles will have significant importance weights when their likelihood is evaluated 

An improvement in the choice of importance proposal distribution over the simple transition 

prior, which also address the problem of sample depletion, can be accomplished by moving the 

particles toward the regions of high likelihood, based on the most recent observations y k f An 

effective approach to accomplish this, is to use an EKF generated Gaussian approximation to 

the optimal importance proposal, that is, 

q(X k \X, k . l J, K }^q N (X k \Y^} (27) 

which is accomplished by using a separate EKF to generate and propagate a Gaussian 

importance proposal distribution for each particle, 

Ar(^l*W) = Al«i«), ^ = l^.. : N (28) 

That is, at time k one uses the EKF equations, with the new data, to compute the mean and 

covariance of the importance distribution for each particle from the previous time step k-1. 

Next, we redraw the rth particle (at time k) from this new updated distribution While still 

making a Gaussian assumption, the approach provides a better approximation to the optimal 

conditional importance proposal distribution and has been shown to improve performance on a 

number of applications. 

Prior 

Likelihood 

Fig. 2 1 Unscented Particle Filter Moving importance proposal distribution 

to regions of high likelihood using the most current observation 

By replacing the EKF with the UKF, we can more accurately propagate the mean and 

covariance of the Gaussian approximation to the state distribution. Distributions generated by 

the UKF will have a grater support overlap with the true posterior distribution than the overlap

558 

achieved by the EKF estimates In addition, scaling parameters used for sigma-point selection 

can be optimized to capture certain characteristic of the prior distribution if known, e g the 

algorithm can be modified to work with distributions that have heavier tail than Gaussian 

distributions such as Cauchy or Student-t distributions The new filter that results from using a 

UKF for importance proposal distribution generation within a particle filter framework is 

called the unscentedparticle filter (UPF) 

3. NUMERICAL SIMULATIONS 

Identification of nonlinear SDOF system: the EKF and the UKF 

In this numerical simulation, performance of the UKF and the EKF was evaluated on 

single-degree-of-freedom Bouc-Wen's hysteretic model as following 

u 

(3 1) 

(32) 

f (3 3) 

where u(t) is displacement, ® n is natural frequency, £ is damping ratio, $ is 

Bouc-Wen's nonlinear displacement related to nonlinear restoring force, u is ground 

acceleration and / & y are shape-control parameters to be identified X is^state vector 

used in the filtering techniques 

Estimation performance was compared for 13 by 13 cases of initial guesses of £(0 | 0) and 

(0 | 0) and estimation error (RMSE in percentile) was shown in Fig. 3 1 Performance 

between the EKF and the UKF is strongly distinct The EKF diverges almost region of 

simulation but the UKF shows robust estimation capability about the initial guesses of £(0 | 0) 

and P(0|0) 

%(0\0"> ieU ' 

a) The extended Kalman filter b) The unscented Kalman filter 

Fig. 3 1 Root mean square error (RJVISE) of estimation

559 

Application to 5-story building structure using unscented Kalman filter 

Numerical simulations are carried out to show the capability of parameter identification using 

unscented Kalman filter Identification performed for the five-degree-of-freedom 

shear-building model with Bouc-Wen's hysteretic nonlinear spring at the bottom of columns 

using the artificial responses data with 1% noise in RMS level In this example, the parameters 

to be identified are c t , k,, A and 7(1=!,2,3,4,5) The exact values of system parameters and 

estimated results are shown in Table 2 and estimation histories for linear and nonlinear 

parameters are shown in Fig 3 3 

'i I i H Bouc-Wen's 

VJ },. - Hysteretic 

Spring 

Fig 3 2 Five story building structure model 

TABLE 2.1 

THE EXACT VALUES FOR SIMULATIONS AND IDENTIFIED VALUES 

Story 

1 

2 

3 

4 

5 

Story 

1 

c true 

02 

02 

02 - 

02 

02 

A 

01 

C 

& guess 

01 

01 

01 

0.1 

01 

P 

Pguess 

00 

îdentified 

0.20011 

01982 

020131 

020015 

0 19839 

/îdentified 

0 099911 

Kfrue 

10 

10 

10 

10 

10 

Jirue 

02 

K 

Kguess 

20 

30 

20 

30 

20 

7 

"Jguess 

00 

îdentified 

99949 

10004 

99957 

10005 

99959 

îdentified 

02002

560 

0 2 4 5 3 10 12 14 16 18 20 0 2 4 8 8 10 12 14 16 IB 20 

_J I. t.- -J 1 L_ 

10 12 14 16 18 2D Q 2 4 B 8 10 12 14 

16 18 20 

6, 1 , , , , . 1 

0*2 

10 1'= 

0 

Q 2 4 5 8 10 12 14 16 18 20 Q 2 4 B 8 10 12 14 

6, , , , , , , ,_ 

4 i 

Li- 

0 2 4 5 8 10 12 14 16 18 20 Q 2 4 B 3 10 12 14 

2, , 1 r- 1 1 1 i— 

-J ' i ' 

4 5 8 10 12 14 16 18 20 0 2 4 S 8 10 12 14 

Time {sec} 

Time {sec} 

18 20 

Q 2 4 6 8 10 12 14 16 18 20 

0 2 4 6 8 10 12 14 1S 18 20 

Time {sec} 

Fig. 3.3 Estimation histories of system parameters 

(—: true values of system parameters)

561 

Identification of nonlinear SDOF system 

The performances of the UPF and the EKF are evaluated on nonlinear identification problem 

with Bouc-Wen's hysteretic spring mode mentioned in previous section (see Eqn. (3.2)). Fig. 

3.4 compares the estimates generated from a single run of EKF and UPF. The extend Kalman 

filter shows divergence phenomena whereas unscented particle filter (UPF) shows good 

estimation capability. 

8 

£' 

— 4 

F 7 

io 

J2 

"2 - 2 

1 ft 

-8 

-10 

fl 

D.erge ^ ji 

y/JV , i ! I ! ! 

•«' y» i / i i 

i 

— observed response 

— UPFestmaton 

— EKF estimation 

I l\ 

i f i 

f "hi 

'• ¥\ / \ / ^ 

\ i y v - 

r I i I i 

V II li ' * : 

V j 

Time (sec) 

i i \ 'i j. 

i I la A /\ 

15 "j 1 5 — EKF estmaton 

f ', 

^5 

1 5 

nllnearpar 

a 

z 

tn o 

^\ — 'Exact values 

S 

,| 

f^ 

i 

; 

-10 

^ 

Diverge 

1 

-IS 

05 1 IS 2 IS 3 3.5 4 45 5 

Time (sec) 

- 

Diverge 

— Exact values 

— UPF esUmaHon 

— EKF estmaton 

Time (sec) 

a) states and parameter estimation b) probability density functions estimated 

by unscented particle filter 

Fig. 3.4 Estimations by the extended Kalman filter and unscented particle filter

562 

CONCLUSION 

In this study, the recently developed unscented particle filter as an alternative to Extended 

Kalman filter has been applied to the identification of nonlinear structural dynamic systems. 

The results from a series of numerical simulation studies are summarized below 

1 Statistical inference for nonlinear dynamic system can be carried out by more generic 

approach using particle filter than linear approximated schemes such as the extended 

Kalman filter. 

2 Unscented particle filter may show better performance than generic particle filters by 

using importance proposal distribution estimated from the UKF. 

3 For real world applications of statistical filtering technique, the accurate mathematical 

model on system dynamics including damage phenomena is demanded. 

References 

1. Haykin, S (2001). Kalman filtering and Neural Networks, Wieley interscience 

2. Juiler, SJ. and Uhlmann, IK., (1994) A new approach for the nonlinear transform of 

means and covariance in linear filters, IEEE Transactions on Automatic Control, 

3. Koo, K.Y, (2001) Time Domain Identification of Nonlinear Structural Dynamic 

Systems using Unscented Kalman filter, M.S. Thesis, KAIST 

4 Maruyama, 0, Yun, C.B., Hoshiya, M. and Shinozuka, M., (1989). Program EXKAL2 

for identification of Structural Dynamic Systems, Technical Report NCEER, New 

York 

5. Rudolph, M., Doucet, A., Freitas, N. and Wan, E. (2000) The unscented particle fitler, 

Technical report Cambridge University Engineering Department 

6. Shinozuka, M., Yun, C.B., and Imai, H., (1982). Identification of linear structural 

dynamic systems, Journal of Engineering Mechanics Division, ASCE, 108:6, 

1371-1390 

7. Sorenson, H.W., (1985). Kalman filtering: Theory and Application IEEE process 

New York 

8. Yuri, C.B. and Shinozuka, M., (1980) Identification of Nonlinear Structural Dynamic 

Systems", Journal of Structural Mechanics, 8:2, 187-203 

9. Yun, OB., Lee. HJ. and Lee, C.G (1997) Sequential Prediction-Error Method for 

Structural Identification, Journal of Engineering Mechanics, 115-122




EMBEDDING ALGORITHMS IN A WIRELESS STRUCTURAL 

MONITORING SYSTEM 

Jerome Peter Lynch 1 , Arvmd Sundararajan 2 , Kincho H. Law 1 , and Anne S. Kiremidjian 1 

Department of Civil and Environmental Engineering, Stanford University 


Department of Electrical Engineering, Stanford University 


ABSTRACT 

A complex wireless sensing unit is designed for application in structural monitoring systems. 

Fabricated from advanced embedded system technologies, the prototype unit employs a spreadspectrum 

wireless modem for peer-to-peer communication between sensing units and a complex 32-bit 

computational core for local data interrogation. Utilizing the computational capabilities of the sensing 

unit design, a collection of structural health monitoring software applications can be implemented for 

execution by the units. Fast Fourier transforms and methods for fitting auto-regressive time-series 

models are employed as illustrative algorithms embedded in the wireless sensing unit. The research 

goal is a wireless sensing network supporting the collaborative processing of real-time measurement 

data for the identification of potential damage in a structural system. 

INTRODUCTION 

The field of structural engineering has historically derived tremendous benefit from the installation of 

structural monitoring systems. Data generated by structural monitoring systems provide insight to the 

performance of a structure over its operational life and during large external disturbances such as 

earthquakes. The current state-of-technology is characterized by centralized systems that employ 

sensors (such as accelerometers) wired to a centralized data acquisition system. The use of wires as 

the sole medium for data transfer from system sensors to data servers cause systems to have high 

installation and maintenance costs. As a result, the adoption of the technology is defined as sluggish. 

With installation and maintenance costs high for tethered systems, Straser (1998) proposed employing 

wireless communication technology to serve as a cost effective and reliable alternative for current 

cabled structural monitoring systems. 

The past decade has witnessed the explosive growth of wireless communications along with major 

advancements of integrated circuit and embedded system technologies. For wireless communication 

and embedded system technologies, the cost of hardware components continues to decrease while their 

functional capabilities broaden. As a result, a wireless sensing unit is designed and constructed from 

off-the-shelf hardware components to serve as the fundamental building block of wireless modular 

monitoring systems (WiMMS). Key features of the unit include wireless communications, high

564 

Figure 1 - Prototype wireless sensing unit 

resolution 16-bit digital conversion of interfaced sensors, and a powerful computational core that can 

perform various data interrogation techniques in near real-time (Lynch et al. 2002). The 

computational core's architecture is designed using two processors. The first is a low-power 8-bit 

Atmel microcontroller included for data acquisition functions such as reading measurements from 

interfaced sensors. The second, a 32-bit Motorola PowerPC microcontroller that supports hardware 

floating point calculations, is included in the design to perform the data interrogation tasks. The unit is 

compact at 20 in. 3 and the current total component cost is less than $500. A picture of the prototype 

wireless sensing unit is shown in Figure 1. 

With a flexible and capable hardware design, the wireless sensing units are to be implemented with the 

computational tasks required by a structural health monitoring system. Structural health monitoring 

algorithms can be embedded in the wireless sensing unit to assess changes in the system, and if 

appropriate, infer potential structural damage from time-history measurement data. To validate the 

sensing unit's role as a computational agent for structural health monitoring applications, two 

algorithms are embedded to locally process measurement data. The first is the fast Fourier transform 

that will be used to derive the frequency response function from time-history data; the frequency 

response function serves as the basis for many frequency-domain damage detection approaches. The 

second algorithm, representing the first step in a novel time-domain damage detection method, is the 

fitting of an auto-regressive time series model to time-history data. The functionality of both 

computational methods is to be illustrated on response measurements obtained from a laboratory test 

structure excited by a shaking table. 

OVERVIEW OF STRUCTURAL HEALTH MONITORING ALGORITHMS 

Structural health monitoring (SHM) is defined as the computational paradigm used to quantify a 

structure's health and well-being over its operational life. Application of SHM concepts to a diverse 

set of structures, ranging from aircrafts to traditional civil structures, has tremendous benefit in 

assessing safety and remaining operational life. While still in its infancy, researchers hi the SHM field 

have produced a variety of algorithms for the identification of damage in structures. To date, methods 

developed for damage detection can be classified as frequency-domain or time-domain approaches. 

The earliest damage detection methods correlated damage to changes hi structural stiffness. The 

methods use finite element models and linear modal parameters, such as natural frequencies and mode 

shapes, to identify the existence of damage and in some cases, even damage location and severity 

(Doebling et al 1996). Modal properties, like natural frequencies of a structure's modes, are observed

in the healthy structure. If major changes are observed in modal properties over a structure's 

operational life, the changes could be attributed to damage. The extraction of modal parameters from 

frequency response functions derived from vibration-based test data follows closely the developments 

in traditional modal testing (Ewms 1984). Classified as frequency-domain approaches, these methods 

had success in identifying large levels of damage in a structure but suffered from an inability to 

positively identify the onset of damage (Sohn and Farrar 2001). With respect to civil structures, 

environmental and operational variability can also cause changes in natural frequencies and mode 

shapes, rendering frequency-domain approaches difficult to apply except in cases of extreme damage 

(Sohn et al. 1999). 

In response to the difficulties associated with applying frequency-domain techniques to civil 

structures, new approaches to damage detection are being explored. In particular, approaches based 

upon the statistical pattern recognition paradigm appear promising (Sohn et al. 2001). This timedomain 

approach entails the use of statistical signal processing techniques applied on time-history 

measurement data to infer the existence of damage in the structural system. The approach extensively 

uses linear predictive time-series models. Assuming the structural response to be stationary, an autoregressive 

(AR) time-series model can be fit to time-history measurement data: 

565 

^=I>,V ( +^ (1) 

1=1 

The response of the structure at time t—kAt t denoted by x^ is a function of p previous observations of 

the response of the system, plus, a residual error term, rf. Weights on the previous observations of x k . t 

are denoted by the coefficient, &,. The residual error of the AR model is a damage sensitive parameter, 

but it is also influenced by the operational variability of the structure. To separate changes in the 

residual error resulting from damage and operational variability, an auto-regressive with exogenous 

input (ARX) time-series model is used to model the relationship between the AR model residual error, 

r/, and the measured response, Xk'. 

i=l y=0 

- + * (2) 

The residual error of the ARX model, e**, is the damage sensitive feature used to identify the existence 

of damage regardless of the structure's operational state. To implement the statistical pattern 

recognition approach, a structure is observed in its undamaged state under a variety of environmental 

and operational states to populate a database pairing AR(p) models of dimension p and ARX(a,b) 

models of dimension a and b. 

After measuring the response of a structure, denoted by 3;*, in an unknown state (damaged or 

undamaged), an AR(p) model is fit. The coefficients of the fitted AR model are compared to the 

database of AR-ARX model pairs previously calculated for the undamaged structure. A match is 

made by minimizing the difference (in a Euclidean sense) between the coefficients of the AR model 

calculated and an AR model from the database. If no structural damage is experienced and the 

operational conditions of the two models are close to one another, the selected AR database model 

should closely approximate the measured response. If damage has been sustained by the structure, 

even the closest AR model of the database will not approximate the measured structural response well. 

With an AR-ARX model pair selected from the database, the measured response, y/c, and residual 

error, r k y , of the fitted AR model are substituted in the ARX model of Equation (2) to determine the 

residual error, £/. If the structure is in a state of damage, the statistics of the ARX model residual, £/,

566 

will vary from that of the ARX model, e£, corresponding to the undamaged structure. The damage 

can then be identified when the ratio of the standard deviation of the model residuals exceeds a 

threshold value established from good engineering judgment: 

(3) 

Application of the linear prediction time-series model has been illustrated for a variety of structures 

from fast patrol boats to lumped mass models (Sohn and Farrar 2001, Sohn et al. 2001). 

EMBEDDED SOFTWARE DESIGN 

To illustrate the units' utility in an automated structural health monitoring system, two computational 

algorithms are selected and encoded for embedment in the prototype wireless sensing units. The first 

algorithm is the fast Founer transform (FFT) used to derive the frequency response function (FRF) 

from time-history data collected by the units. The second computational algorithm to be embedded is 

an auto-regressive (AR; time series fitting algorithm. 

Fast Fourier Transform 

The FRF of a structural system can be calculated directly from measurement data by using the 

computationally efficient fast Founer transform (FFT). The discretely measured response of a system 

in the time domain, h k , is converted to the response in the frequency domain, H n , by means of a 

discrete Fourier transform: 

H^ZV 1^ (4) 

JL*0 

where A T represents the total number of time-history samples. If the discrete Fourier transform of 

Equation (4) is directly calculated by the wireless sensing unit, the calculation is an 0(N 2 ) process. 

Utilization of the discrete fast Fourier transform reduces this calculation to an 0(Nlog2N) process. 

Although various forms of the fast Fourier transform are available for use, our implementation of the 

algorithm for embedment in the sensing unit's core follows closely the Cooley-Tukey method (Press et 

al. 1992). The approach reorders the initial time-history data and performs a series of two-point 

Founer transforms on adjacent data points. 

Auto-Regressive Time Series Modeling 

The wireless sensing units can individually perform most of the computations associated with the 

statistical pattern recognition algorithms. The role of the unit is well defined with the responsibility of 

recording the structural response, normalizing the response, and fitting AR models to the 

measurements. After the AR model is fit, the model's coefficients would be wirelessly communicated 

to a centralized data server housing a database populated by AR-ARX model pairs. Once a model 

match is made, the coefficients of the ARX model are transmitted to the wireless sensing unit where 

the standard deviation of the ARX model is calculated. The ratio of standard deviations of ARX 

model errors shown in Equation (3) is calculated and compared to a previously established threshold.

Software is written for the wireless sensing unit to determine the coefficients of an AR(p) model based 

on a segment of the recorded data. Multiplying both sides of Equation (1) by the current measurement 

sample, z/., and taking the expected value of both sides, the autocorrelation function of the autoregressive 

process is derived: 

567 

' cc v ' / i i T z> \ v "/ /c\ 

t=i 

PJ 

where cp^(k) are the discrete terms of the autocorrelation function of Xk. The autocorrelation function 

obeys the initial difference equation of the AR process. This yields a means of determining the 

coefficients of the AR process based on calculations of the autocorrelation of the measurement data. 

Resulting are the Yule-Walker equations (Gelb 1974): 

>-2) fc 

(6) 

Coefficients of the auto-regressive process are very sensitive to the way the autocorrelation of the 

process is determined. As a result, Burg's method has been proposed by Press et al. (1992) for 

determining the coefficients of the auto-regressive model directly from the measurement data. The 

method is recursive with its order increasing during each recursive call by estimating a new coefficient 

b t and re-estimating the previously calculated coefficients so as to minimize the residual error of the 

process. 

SOFTWARE VALIDATION 

To validate the performance of the wireless sensing unit in a structural health monitoring setting, 

validation tests upon a laboratory test structure are devised. A five-story shear frame structure, made 

from aluminum, is employed as shown in Figure 2. The lateral stiffness of each floor originates from 

the four vertical aluminum columns roughly 0.5 inches by 0.25 inches in cross sectional area. Each 

floor weighs 16 pounds. From log-decrement calculations of free vibration tests, the damping of the 

structure is approximated to be 0.5% of critical damping. The wireless sensing unit is securely 

fastened to the fourth story while the Analog Devices ADXL210 MEMS-based accelerometer is 

mounted upon the fifth story. The entire structure is fastened to the top of a one-directional lateral 

shaking table driven horizontally by an 11 kip actuator. Various excitations are applied at the base of 

the structure to dynamically excite the system. 

Determination of the Frequency Response Function 

A swept-frequency sine, also known as a chirping excitation, is applied to the base of the structure in 

order to excite the lower modes of response of the system. The chirping excitation has constant 

displacement amplitude of 0.075 inches with a linearly varying frequency of 0.25 to 3 Hz over 60 

seconds. During the excitation, the acceleration response of the fifth story is monitored. The 

measurement data is sampled at 30 Hz, well above the primary modes of response of the system 

analytically determined to be 2.96, 8.71, 13.70, 17.47, and 20.04 Hz.

568 

A 

11 25" 

11 25" 

1 

T 

1 

11 25* 

! 

5 _L 

11 25" 

-L s 

_, _ f 

•^ 

T 

Figure 2 - Five-story lumped-mass shear structure mounted upon a shaking table 

ADXL210 Measured Absolute Acceleration Response 

10 30 


Figure 3 - Absolute acceleration response at the fifth story from a sweep input excitation 

Figure 3 presents the absolute acceleration response of the shear structure to the input motion 

generated b"v the shaking table The absolute acceleration response measured using the accelerometer 

is in ver> good agreement with the theoretical response determined analytically 

The frequency response function of the recorded tune-history is locallv calculated from an embedded 

FFT algorithm The FFT is performed on 1024 consecutive tune points of the response between 10 

and 44 seconds The first three modes of response of the structure can be visually identified from the 

calculated response functions as shown in Figure 4 The modes are determined to be 2 87, 8 59, and 

13 54 Hz The frequencies of the calculated modes are within 3% of those calculated from the 

theoretical model 

Auto-Regressive Time Series Model Fitting 

The test structure is excited b> a white noise excitation (zero mean and 0 05 inch standard deviation) 

and the accelerometer on the fifth story measures the acceleration response as shown in Figure 5

569 

ADXL210 Frequency Response Function (30Hz) 

5 10 


15 

Figure 4 - Frequency response function denved from the structural acceleration response 

ADXL210 Measured Absolute Acceleration Response to White Noise Input 

30 

Time (sec) 

Figure 5 - Absolute acceleration response of fifth-story to a white noise input excitation 

The time history acceleration response of the structure as measured in this manner is relatively 

stationary with zero mean and a standard deviation of approximately 1 1 g Therefore, the record is 

suitable to fitting an auto-regressive time series model 

After recording the acceleration response of the test structure, an auto regressive model of 10 

coefficients is tit to the measurement data of Figure 5 After logging the coefficients an identical 

analysis is performed using Burg s auto-regressive function provided by MATLAB to compare the 

accuracy of the coefficients determined by the wireless sensing unit The coefficients determined by 

the wireless sensing unit and MATLAB, as presented in Table 1 are identical 

Table 1 - Coefficients of a fitted auto regressive model to stationary structural response 

bi 

1> 

b 3 

5 4 

b s 

be 

b 7 

b s 

b 9 

bio 

PowerPC 

166M) 

13009 

047D2 

01171 

06756 

15212 

12646 

0^989 

01498 

00891 

AR 10 

MATLAB 

16650 

13009 

04752 

0 1171 

06756 

15212 

12646 

05989 

01498 

00891

570 

CONCLUSION 

The realization of an automated structural health monitoring system has taken one step forward with 

the development of a wireless sensing unit constructed with advanced embedded system technologies. 

The result is a hardware design that is optimized for tasks specific to structural health monitoring 

applications. In particular, an advanced computational core is provided that is capable of locally 

processing measurement data to assess the state and possibly identify damage in a structure. A fast 

Founer transform algorithm and an auto-regressive time series algorithm have been embedded in the 

unit to illustrate the unit's computational capabilities. Both applications have been successfully 

applied to time-history measurement data taken from a five degree-of-freedom laboratory structure 

excited bv a shaking table. 


This research is partially sponsored by the National Science Foundation under Grant Numbers CMS- 

9988909 and CMS-0121842. The fruitful suggestions provided by Dr. Chuck Farrar and Dr. Hoon 

Sohn of the Los Alamos National Laboratory, have been invaluable to the progress of our research. 

REFERENCES 

Doebling, S.W., Farrar, C.R., Prime, M.B., and Shevitz, D.W. (1996). Damage identification and health 

monitoring of structural and mechanical systems from changes in their vibration characteristics: a literature 

review. Report LA-13070-MS, Los Alamos National Laboratory, Los Alamos, NM. 

Ewins D. J. (1984). Modal testing: theory and practice. Research Studies Press, Letchworth, England. 

Gelb, A. (1974). Applied optimal estimation. MIT Press, Cambridge, MA. 

Lynch, J. P., Law, K.H., Kiremidjian, A.S., Kenny, T.W., and Carryer, E. (2002). A wireless modular 

monitoring system for civil structures. Proceedings of2& h International Modal Analysis Conference (IMAC), 

Los Angeles, CA, February 4-7. 

Press, W. H., Teukolsky, S. A,, Vetterling, W. T., and Flannery, B. P. (1992). Numerical recipes in C: the art 

of scientific computing. Cambridge University Press, Cambridge, England. 

Sohn, H., Dzwonczyk, M., Straser, E.G., Kkemidjian, A.S., Law, K.H., and Meng, T, (1999). Experimental 

study of temperature effect on modal parameters of the Alamosa Canyon Bridge. Earthquake Engineering and 

Structural Dynamics, 28:9, 879-897. 

Sohn, H., and Farrar, C.R. (2001). "Damage diagnosis using time series analysis of vibration signals." Smart 

Materials and Structures, Institute of Physics, 10:4, 446-451. 

Sohn, H., Farrar, C.R., Hunter, N., and Worden, K. (2001). Applying the LANL statistical pattern 

recognition paradigm for structural health monitoring to data from a surface-effect fast patrol boat. 

Report LA-13761-MS, Los Alamos National Laboratory, Los Alamos, NM. 

Straser, E.G. {1998). A modular, wireless damage monitoring system for structures. Ph.D. Thesis, Department 

of Civil and Environmental Engineering, Stanford University, Stanford, CA.




MULTIFACETED SEISMIC EVALUATION AND RETROFIT 

STUDIES OF A MAJOR VIADUCT 

M. Saiidi 1 , A. Itani 1 , Q. Yang 2 , and S. Ladkany 3 

'Civil Engineering Department, University of Nevada, 

Reno, USA 

2 Civil Engineering Department, Northern Jiaotong University, 

Beijing, China 

3 Civil Engineering Department, University of Nevada, 

Las Vegas, USA 

ABSTRACT 

A major viaduct in downtown Las Vegas is slated for seismic retrofit. The structure has 22 spans and 

110 columns of various shapes. The performance and seismic vulnerability of the bridge and its 

adjacent rarnps are being evaluated. The focus of this article is on the effect of incoherent ground 

motion on the viaduct and an explanation of a pilot study on the use of superelastic shape memory 

alloy (SMA) to upgrade the seismic performance of the bridge elements. It is shown that combination 

geometric incoherency, wave passage, and site characteristics effects can increase the base shears 

considerably. The improvement in the capacity and ductility of concrete cubes confined by SMA rods 

has pointed out the potential of this new material for application in seismic retrofit. 

INTRODUCTION 

Several new concepts and issues have emerged in the past few years in seismic evaluation and retrofit 

of structures. The underlying principle in utilizing new concepts is to ensure a reliable, economical, 

and intelligent design that makes optimal use of innovative materials and information technology tools. 

Some of the new tools, while may be well developed in the arena of their origin require further 

research and development before they can be adopted in civil engineering projects. It may be hard to 

justify the development work for specific projects of relatively small magnitude. However, the 

potential benefit of such research can be significant for large projects and hence undertaking the socalled 

"high risk" studies may be warranted. 

In planning the seismic retrofit of major bridges, it is inevitable that the solution will involve both 

conventional and non-conventional strategies. This article presents some of the results of an on-going 

study to develop retrofit plans for a major freeway viaduct in Las Vegas, Nevada. Several aspects of

572 

the structural response and retrofit methods are being studied. The article will present a summary of 

the study of a selected number of the critical issues. 

DESCRIPTION OF THE VIADUCT 

The viaduct consists of two main 22-span continuous structures for the westbound and eastbound 

traffic on Interstate 95 in North Las Vegas (Fig. 1). In addition a 7-span on ramp and two off ramp 

badges, one with two spans and the other with three spans are connected to the viaduct. The 

superstructure is cast-in-place muiticell box girder in all the structures and has seven in-span hinges in 

the main structure. The main bridges are supported on multi-column piers with two to four diamond 

shape columns and the ramp bridges are supported on irregular octagonal single-column piers some 

with a pedestal. Figures 2 and 3 show the cross section of the columns. Single columns are detailed 

vuth a one-way hinge at connection to the footing and columns of the multi-column bents are detailed 

with two-way hinges at the base. 

The viaduct was constructed in 1969 and was widened in 1984 by adding one row of columns and a 

new box cell on the north side with little seismic load and detailing consideration. The columns are 

supported on spread footings. Based on the soil blow counts at the site of the bridge, approximately 

the west third of the bridge is on a relatively soft soil while the rest of the structure is on medium firm 

soil. 

Seismic codes place Las Vegas in areas of moderate seismicity category. To determine how critical a 

bridge is with respect to the need to upgrade its seismic performance, however, the average daily 

traffic (ADT) is also factored in. As a result the Nevada Department of Transportation has identified 

the viaduct as the most critical bridge in Nevada with respect to priority for seismic retrofit because of 

its high ADT in addition to having seismic deficiencies. Seismological studies of the Las Vegas area 

have pointed out the possibility of strong ground motion due to soil amplification despite the fact that 

the area is categorized as having only a moderate probability of experiencing high seismic forces (Su 

etal. 1998). 

CRITICAL ISSUES 

The original structure and the subsequent expansion were not designed to resist significant seismic 

forces. As a result there are many deficiencies in the bridge with respect to seismic detailing. These 

include inadequate lateral steel in the columns (Fig. 2 and 3), short anchorage length for longitudinal 

bars, insufficient shear steel in the beams, very low amount of bottom steel in the beams at 

connections to columns, a lack of top mat of steel in footings, short seat widths at hinges, and a lack of 

shear steel at beam column connections. Considering the length of the viaduct and variation in soil 

properties, there is also uncertainty about the performance of the bridge under incoherent ground 

motions at different supports. 

The study to address these deficiencies include (1) a three dimensional detailed nonlinear modeling of 

the viaduct subjected to earthquakes to determine the performance level under different loading 

scenarios, (2) a reduced linear modeling of the bridge to study the effect of incoherent ground motions, 

(3) shake table studies of the as-built and retrofitted models of single-column bents, (4) shake table

573 

studies of the as-built and retrofitted models of multi-column bents, and (5) an exploratory study of 

the application of innovative materials for retrofit. Due to space limitation only a summary of the 

research in Tasks 2 and 5 is presented in this article. 

INCOHERENT GROUND MOTION EFFECTS 

Although commonly assumed to be uniform, ground motion at different supports of a structure may 

vary from one support to the next. Variation in support excitation may be due to three causes: (1) 

geometric incoherency, (2) wave-passage, and (3) local site geotechnical characteristics. The first 

effect is due to randomness of the earthquake motion even at a given site with practically the same 

distance from the earthquake source. The second effect is caused by delay in the motion from one 

support to the next as the earthquake wave passes through the site. Finally the third source of 

variation is that, for the same bedrock motion, soils of different properties and depth transfer the 

motion to the surface differently. The first two factors can be particularly important for relatively long 

structures, whereas the third factor can be important even for short structures with supports located on 

soils layers that are different from one another. 

The total length of the bridge is 552.7 m. Furthermore soil investigation of the site prior to 

construction have shown that the soil type over the western third of the bridge may be categorized as 

soft, while the soil for the rest of the bridge is in the category of medium firm. Considering the length 

and the variation in the soil type it was decided to include all three sources of deficiencies in the study. 

No site-specific seismic studies and detailed soil investigations have been conducted for the viaduct. 

As a result it was felt that the study of the incoherency effects would have to be approximate. A 

reduced linear model of the viaduct was used to determine the trends in the response relative to the 

response for uniform ground motions. The structure was treated as a planar system subjected to inplane 

ground motion loading in the longitudinal direction of the bridge. Details of the study are 

presented by Yang et al. 2002. 

Figure 4 shows the elevation of the reduced model of the viaduct. To develop the reduced model, 

each segment of the structure between adjacent hinges was modeled by a single mass supported on a 

column that had the same lateral stiffness as that of the bridge segment. Bridge hinges were modeled 

as pin connections. The coherency function was adopted from a previous study that was based on 

random processes principles developed by Yang and Chen (2000). Eight acceleration records were 

generated at bedrock with slightly different peaks in a range of 0,38 to 0.41g. These motions were 

used to determine the displacement and acceleration histories at the ground level of the soil column 

under each support of the reduced model. To study the effect of wave passage, a wave velocity of 600 

m/sec. was used for the entire viaduct even though the soil varied from the west to the east. This was 

done because blow counts of the western and eastern parts, although placed the soils in different 

categories, were not drastically different. 

The effect of different combinations of incoherency parameters were studied by focusing on the 

"column" base shears in the reduced model. An artificial acceleration record generated for a 

magnitude 7.4 at 40 km distance from the epicenter was used. To establish a benchmark, the model 

was analyzed for two uniform ground motions obtained by combining the bedrock motion near the 

middle of the viaduct one with amplification due to soft soil and the other with amplification with 

medium firm soil. The average of the two responses was used as the representative response for 

uniform ground motion.

574 

The ratios of base shear at different column base nodes for the non-uniform and uniform motions are 

shown in Fig. 5. In case 3 only the effect of site characteristics was included. Cases 4 and 5 included 

the wave passage effect in addition to the local site effect, one for waves traveling from east to west 

and the other traveling in the opposite direction. In case 6 the combination of local site effect and 

geometric mcoherency was included. Finally cases 7 and 8 accounted for all three parameters, for 

waves traveling from east to west and west to east, respectively. A ratio exceeding one on the vertical 

axis indicates an increase in force due to incoherent ground motion. It can be seen in Fig. 5 that cases 

3,4 and 5 were not generally critical. However, for cases 6 to 8 the base shear ratios exceeded 1 in 

man}* piers. The maximum ratio occurred in Pier 7 with ratios being 1.59, 1.72, and 1.84 for case 6, 7, 

and 8, respectively. It was concluded that the inclusion of all three mcoherency effects can be the 

most critical combination and even eliminating the wave passage effect (case 6) led to force ratios that 

exceeded one in one-half of the piers. 

STUDY OF SHAPE MEMORY ALLOYS 

Superelastic shape memory alloys (SMAs) have found their applications in medical equipment, 

aerospace, and dentistry. Their ability to deform well into the yielded state and recover their entire 

deformation upon unloading makes them potentially attractive for earthquake resistant design. The 

application of SMA rods for bridge seismic restrainers has been explored by DesRoches (2001). As 

part of the study of the retrofit alternative for the Las Vegas Downtown Viaducts, the performance of 

SMA rods made from Nickel and Titanium (NiTi) used in combination with concrete is being 

investigated. Basic characteristics of SMAs in providing concrete confinement and in acting as 

longitudinal reinforcement are determined in concrete cubes and beams subjected to cyclic loads. 

Figure 6 shows a flag-shape stress strain relationship that is characteristic response of superelastic 

SMA materials. A series of 150x150x150 mm concrete cubes have been cast with central pipes in two 

directions for the passage of SMA rods. Figure 7 shows the compression load setup for one of the 

cubes. The SMA rods are anchored by nuts and stiff steel bearing plates to confine the concrete. The 

rods in the setup shown are of 13 mm diameter with a nominal upper yield plateau of 420 MPa and a 

lower yield plateau of 140 MPa. 

Only a limited number of cubes have been tested thus far. Figure 8 shows a sample result for cyclic 

loading of a cube confined with two 13-mm SMA rods. The unconfmed cube strength was 35.9 MPa. 

It can be seen that confinement increased the compressive strength to 52.2 MPa, an increase of 45 

percent. Using Mander's confined concrete model and the measured confinement stress, the estimated 

compresshe strength would be 55.6 MPa. The ultimate strain capacity of the confined concrete was 

near 0.029, which is approximately 10 times the nominal failure strain of unconfmed concrete. More 

tests are being conducted to compare the permanent damage experienced by SMA-rod confined cubes 

and cubes confined by steel rods and to develop confinement models as necessary. 

CONCLUSIONS 

Several aspects of the behavior and retrofit of a major viaduct are being studied. The results of the 

study of the incoherent ground motion effects suggest that internal forces can be significantly higher

575 

than those based on uniform ground motion when all three parameters, namely, geometric incoherency, 

soil characteristics under different supports, and the wave passage effects are included. The increase 

in the force even for earthquakes with peak acceleration of approximately 0.4g exceeded 80 percent. 

The exploratory study of superelastic shape memory alloys as a possible means to improve 

confinement of the substandard bridge elements has led to promising results. Both the compressive 

strength and ductility of concrete considerably increased as a result of the confinement. 


The study reported in this article is funded by the US Federal Highway Administration and the 

Nevada Department of Transportation. Several graduate students have been working on different 

aspects of the project and they are thanked for their dedication. The students are: Chadi Ayoub, Dong 

Gang, Nathan Johnson, Suresh Kandasamy, and Hasan Mohammad. 

REFERENCES 

Su, F., J. Anderson, S. Ni, and Y. Zeng, (1998) "Effect of Site Amplification and Basin Response on 

Strong Motion in Las Vegas, Nevada," Earthquake Spectra, V. 14, N. 2, pp. 357-376. 

Yang, Q., M. Saiidi, H. Wang, and A. Itani, (2002) "Influence of Ground Motion Incoherency on 

Earthquake Response of Multi-Support Structures," Civil Engineering Department, University of 

Nevada, Reno, Report No. CCEER 02-2. 

Yang Q., and Y. Chen, (2000) "A Practical Coherency Model for Spatially Varying Ground Motions" 

Structural Engineering and Mechanics, 9(2), pp. 141-152. 

DesRoches, R., (2001) "Application of Shape Memory Alloys in Seismic Rehabilitation of Bridges," 

NCHRP-IDEA 29, Annual Progress Report, Transportation Research Board, Washington, DC. 

Elevation View 

Fig. 1 Plan and elevation views of the viaduct

576 

SECTION A-A 

COLUMN DETAIL AT 

TOP OF PEDESTAL 

Fig. 2 Typical column cross section in single-column piers (1 in. =25.4 mm) 

Fig. 3 Typical column cross section in multi-column piers (1 in.= 25.4 mm)

577 

1-* 

67.06m ^^ 7757m ^ ^ 73.15m ^ 73.15m 71.17m7401m 

«•—•" •— —" ^" •"• *• -« *• • *•• » «• I*- 

1^1 A, V M2 b M 3 A. M4i A. MS ±, MA A. M L. Mgj 

C 

9 

» ~ • 

10 11 12 

• 

"13 1"4 

9 

715 

~ 

IB 

m 

IT 

~ • 

18 79 20 

* 

21 

- 

22 

m 

23 

il Pj H 2 |Pj H 2 iP 2 H 3 P 3 H 3 Pj HsjPj H 3^j H 4 !F 

1 

L — 

2 ' 3 . " 4 5 1 6. 7. 8i 

i 

Soft soil (Type 

1 . « 

M 

j 

I I 3 

1 

C 

- 1 J - 1 - ' 'i ' • 

1 • 1 " ! • ! . * "' j . 

"] . - a. 1 • - ' 1 1 ' 

Medium soil (Type ) ^ - 

'; j J - . " i •" 

: I I J "I *! 

i - .M - " ; 

j i • . 

^-c^^-. i . * •' i r - a ,. j^^, __ 

Bed rock 

J 

. 

Fig. 4 Reduced model of the viaduct 

F/F t 

5 

• Case3 • Case4 ^ CaseS 

T Case6 * Case + CaseS 

0.8 - 

2.0- 

1.6- 

1.2- 

0.4- 

+ 

¥ 

+ 

• 1 A 

t . . . . , 

* . A 

0 1 2 3 4 5 6 7 

Node number 

Fig. 5 Effect of incoherent ground motion on base shears

578 

Stress 

Strain 

Fig. 6 Superelastic behavior of SMA rods 

Fig. 7 Cubes confined by SMA rods 

0.01 0.015 0.02 

Vertical Strain 

Fig. 8 Venical stress strain relationship for SMA confined cube




NEW SYSTEM IDENTIFICATION ALGORITHHMS COMBINING 

MONTE CALRO FILTER AND GENETIC ALGORITHM 

T. Sato ', Sakanoue 2 and I. Yoshida 3 

1

580 

Genetic Algorithm(GA) and the MCF by Higuchi[3], we developed an algorithm implementing an GA 

operation into the MCF to speed up convergence of the MCF to identify the non-stationary structural 

parameters. 

STATISTICAL FILTER THEORIES 

State Space Model 

The general form of state space model, with non-linear and non-Gaussian characteristics, can be 

described by the following state transfer equation and observation equation. 

X* = F(*/i-l>Wj (1) 

in which \ ;J is the state vector, w, 7 the system noise, z /7 the observation vector, H the observation 

matnx, v the observation noise vector. 

Kalman Filter(KF) 

The KF is easily derived if the state transfer and observation equation are linear, and noises are 

Gaussian. The linear state transfer and observation equation are described as follows: 

in which 

„_, is the state transfer matrix, r n _, the state transfer matnx for system noise. 

The KF algorithm is defined by the following steps: 

1. Define an initial value of the state vector x 0 and its covanance matrix P 0 as well as the 

co variance matrix of the observation noise R,, . 

2. Calculate the pre-estimation value of the state vector x /7 and its covanance matrix M ;J from 

*„= „-!*„-! +r n _,w, 7 _ 1 (5) 

M fl =„_,„_,

581 

K /7 = P n H n R- 1 = MX(H, 7 MX+R, 7 )~' (8) 

5. Calculate the most likely-hood(or post-) estimation value of the state vector x by 

** =x lf +K ll (z /f -H fl xJ " (9) 

6. Return to 2. after renewal of the time step. 

Monte Carlo Filter(MCF) 

The MCF can be applied even if the state space model is non-linear and non-Gaussian. In the MCF, the 

probabilistic nature of the state vector is described by many realizations instead of first and second 

moments as defined in the KF. However, in the MCF, the observation noise vector v is assumed to be 

uniquely determined by a function G differentiate with respect to the observation vector z n as 

follows: 

\ n =//"'(z /7 ,xJ = G(z /7 ,x /7 ) (10) 

The MCF is an algorithm to identify the conditional distribution function p(\ n |Z,,)of the state 

variable x n when observation values Z = {r, •• = n }up to the present time n are given. Then, we define 

P(\ n |Z,,_,) as the predictor distribution and p(\ n \ Z /7 ) as the filter distribution. In this paper, each 

probability density function of each distnbution is approximated by finite realizations composed of 

m particles as follows: 

The MCF algorithm is defined by the following steps. 

1. Generate a initial set (j = l~m) of random number f^ obeying an arbitrary probability 

density function p Q (f 0 ) 

2. Repeat the following steps until the end of time steps. 

(a) Generate a set of random number w|/' obeying probability density function q(w n ) 

(b) Compute a state transfer of particles by 

(c) Compute the likelihood value of each particle by 

a ( n j] = r(G(z /7 ,b J/')) —— (14) 

n 

(d) Generate f ( n ]) by resampling of b (/) (A) 

fu) = b a» with probability -^— (15) 

(e) Return to (a) Yor (n

582 

IDENTIFICATION ALGORITHM 

State Transfer and Observation Equations 

For simplicity we formulate the system transfer and observation equations for a single degree of 

freedom system. The equation of motion for a structural system is 

in which, m is the mass, h the damping ratio, 

y 4- 2hcoy + co 2 y = -X g (16) 

co the natural circular frequency , and y is the 

relative displacement to the ground. X^ is the ground acceleration. And the state vector is chosen as 

follows: 

x= [y y h co} T (17) 

The state transfer equation is expressed by a nonlinear equation as follows: 

in which, g is a vector composed of four components 

gCx^Xf+W, (18) 

g={-2h(ay-(D 2 y-X 9 y 0 0) r (19) 

When the relative velocity and displacement are observed, the observation equation is defined as 

follows: 

n f f n (2Q) 

in which, H is given by 

H 

o 

O i o o 

(21) 

Identification Algorithm Using Hybrid Filter 

In Hybrid Filter, the state vector in the MCF is divided into two parts , in which one represents the 

responses of the structural system, \ a and the other represents the structural dynamic parameters, 

\p as follows: 

**={^v} r (22) 

Xp = [h,a)\ T (23) 

The distribution of x a is assumed to be Gaussian and identified by the KF algorithm. The 

distrubution of \ fl is expressed by many particles and identified by the MCF algorithm. 

The Hybrid filter algorithm is defined by the following steps: 

1 . Define an initial value of the state vector x ff 0 and its covariance matrix P 0 .

583 

2. Generate an initial set of random number \ ( f = [h ( Q'\a ( Q n \ obeying probability density function 


(a) Generate a set of random number wj/ 1 obeying probability density function 

(b) Compute the pre-estimation value of state vector \ p by 

(c) Calculate the pre-estimation value of the state vector x a by 

(24) 

= 

(25) 

y 

(d) Lineanzmg Eq.(25) to apply the KF algorithm. The state transfer matrix 

~ Jl-2A B ( -"S' J) A/ -co ( f~\ 

*•-{ A, , i 

584 

the simulated structural response. The observed data of all nodes are used for the identification. In this 

case, we asuume Gaussian distrubution for the dynamic parameters whose mean is true value and 

standard deviation is 0.01. It is because that the aim of this example is to detect damages of structures. 

The number of particles is 5000. 

Figure 1. The sketch of the non-stationary structural system 

The identified time histories of the damping ratio and natural circular frequency at node 1 are shown in 

Fig.2. In these algonthm, probability density functions of dynamic parameters are identified as a set of 

particles. We therefore expressed the identified results as the time series-histogram of particles. Fig.2 

shows that the identified value of the natural circular frequency converge to the true value for both 

cases using MCF and Hybnd filter. But that of the damping ratio tend to converge to the true value 

only for the case using Hybrid filter. 

Time step 

Time step 

0005 

0015 

damping ratio 

002 0025 0.03 

(a) normal MCF 

natural circular frequency 

Time step 

Time step 

001 

0.015 

damping ratio 

002 

0.025 Q,03- 

(b) Hybrid Filter 

2 4 


Figure 2. Comparison of the time histories of the identified dynamic 

parameters at node 1 by normal MCF and that by Hybrid Filter

585 

A FILTER COMBINIG GA AND MCF 

GA-MCF 

A study to investigate the relationships between the Genetic Algorithm and the Monte Cairo Filter has been 

conducted. Both the MCF and the GA is the algorithm to reconstruct a set of realizations representing the state 

variables from an initial set of random numbers. 

In the MCF, the particles representing the state vector are always generated obeying the conditional probability 

density function that is influenced by past-information of the structural response. The MCF has, therefore, a 

function to decrease the influence of past observation noises. On the contrary, it can not perform high tracking 

ability for an abrupt change of dynamic characteristics of structural systems. 

We developed an algorithm introducing the mutation procedure of Genetic Algorithm into the MCF to speed up 

convergence for identifying non-stationary structural parameters and called GA-MCF. The proposed method is 

defined by the following steps: 

1. Generate a initial set (j=l^/77)of the state vector f^1 

function p Q (\) 

obeying an arbitrary probability density 


(a) Generate a set of random number wj/ 1 obeying probability density function q(w,,) 

(b) Compute prediction partcles using the state transfer eaution 

bJ/UFff^X/') (32) 

(c) Compute the likelihood value of each particles by 

m (^, , 11 K ^ 5G 

(33) 

(d) Choose k particles from a set of particles obtained by Eq.(32) and construct a new set of the 

prediction particles B = (b! 7 ' J --b^} in which each particles has a large likelihood value. 

(e) Generate new filter particles by mutating the filter particles which generate a set of the 

prediction particles B and compute again prediction particles B' using Eq.(32) 

(f) Comparing the likelihood value of each component of B' with that of B When the 

likelihood value of a component in B' is larger than that of a component in B, the particle 

in B is replaced by the particle in B'. 

(g) Generate f,| ;) 

by the resampling of b (y) 

(h) Return to (a) until the end of observation data. 

Numerical examples 

To demonstrate the efficiency of the proposed GA-MCF filter to identify the non-stationary structural 

parameters. We simulate a set of observed structural responses of which dynamic parameters varies at ten 

seconds(1000 steps) after the earthquake motion input to the structure. The number of particles is 5000.In

586 

GA-MCF, the number of particles set used for mutation is 100 Fig 3 shows comparison of the time histories of 

the identified circular frequency of the first layer obtained by the normal MCF and that by GA-MCF The 

identified result obtained by the GA-MCF has higher tracking ability for the non-stationary change of structural 

parameters than that by normal MCF Fig 4 shows the number of particles that are replaced by new particles 

generated in step (e) of GA-MCF algorithm This means that mutations in GA work well to track abrupt change 

of structural parameters The number of renewal particles is large especially when the structural system becomes 

non-stationarv 

2 4 s 


{a) normal MCF 

^ GA-MCF 

Figure 3 Companson of the time histories of the identified natural circular 

frequency at node 1 by normal MCF and that by GA-MCF 

Figure 4 The number of the renewal particles m GA-MCF 

CONCLUSIONS 

We developed two new system identification algorithms, one is a hybrid algorithm combining the Kalman Filter 

and the Monte Carlo Filter The other is an algorithm taking into account the Genetic Algorithm to speed up 

convergence of the Monte Carlo Filter to identify non-stationary structural systems 

REFERENCE 

[ 1 ] Kitagawa, Genshiro {1996) Monte Carlo filter and smoother for non-Gaussian nonlinear state space models Journal of 

Computational and Graphical Statistics Vol 5, No I 1 -25 

[2] Yoshida, Ikumasa (2001) Damage Detection Using Monte Carlo Filter Based on Non-Gaussian Noises 8 th 

International Conference on Structural Safety and Reliability June 

[3] Higuchu Tomoyuki. (1997) Monte Carlo filter using the Genetic algorithm operators Journal of Statistical 

Computation and Simula!ion, Vol 59, No 1 1-23




REAL-TIME STRUCTURAL MONITORING SYSTEM 

A. M. Sereci 1 , D. Radulescu 1 and C. Radulescu 1 

îgitexx Data Systems, Inc. 

Pasadena, California 

ABSTRACT 

This paper presents the state-of-the-art technology already implemented in real-time structural 

monitoring and initial damage assessment approach. The short history of earthquake data recording is 

intended to provide a general view of the main possibilities of different recorders in time. The core of 

the paper focuses on how today's modern systems can provide the tools to do much more than simple 

event recording. The real examples are provided using already installed systems to show that today's 

technical advances combined with powerful data acquisition, analysis and broadcasting software is a 

perfect match for real-time structural monitoring studies. Today, affordable fast communication (DSL, 

Tl, T3), well established protocols (TCP/IP, UDP, FTP) combined with COTS (Commercially Of The 

Shelf) data acquisition hardware, and powerful software packages make this domain accessible. 

1. HISTORICAL OVERVIEW 

Earthquakes were studied for a long time. Due to their complexity, three major sciences emerged. 

Seismology, the science that studies the characteristics of the earthquake as natural phenomena; 

Engineering Seismology, the science that studies the wave characteristics at a given site; and 

Earthquake Engineering, the science that studies the response of a structure during an earthquake. 

Different recording equipment has been developed for each of these sciences to provide accurate data 

required as an input for different analysis methods. Sometime, an instrument has been developed to 

cover more than one science. The time proved that this approach had a cost/benefit flaw for at least 

one science. Also, the location of the instrument requires site selection parameters, which are different 

from one science to another. For example, a seismological station requires a quiet site, direct contact 

with the bedrock or as close as possible, grid type distribution, high resolution etc. Engineering 

Seismology requires grid type installation in high-populated areas combined with geological survey to 

describe the near surface stratification, local amplification phenomena, propagation, attenuation and 

frequency content at a given site. Finally, the Earthquake Engineering science requires structural

588 

instrumentation to monitor and determine the structural response during an earthquake and also to 

determine the damage level based on advanced data analysis. This application requires especially: 

High number of channel count 

Different type of sensors (accelerometers, pressure, temperature, relative displacement, etc) 

Relatively high dynamic range (min. 96dB) 

Compact system hardware 

Redundancy 

Easy configurable real-time data analysis 

Full-remote control 

Powerful data broadcasting 

Real-time design limits exceedance analysis 

A short historical evolution of the instrumentation dedicated to structural monitoring is presented 

below (see Table 1.1) 

TABLE 1.1 

EVOLUTION OF STRUCTURAL RESPONSE MONITORING EQUIPMENT 

Type 

Analog 

Analog 

Analog 

Analog 

Digital 

Digital 

Digital 

Digital 

Digital 

Digital 

Digital 

Digital 

Digital 

Digital 

Recording 

Media 

Photographic Film 

Photographic Film 

FMTape 

FMTape 

Memory 

Memory 

Memory 

Memory 

Sundisk 

Sundisk 

Sundisk 

Sundisk 

PC-Based 

PC-based 

Number of 

Channels 

3 

13 

3 

12 

3 

18 

24 

3 

3 

24 

3, 6, or 12 

18 

32, 64, or 128 

32, or 64 

Dynamic range 

fdBl 

-60 

-60 

589 

This paper focuses on the last two systems presented in Table 1.1. These systems are considered stateof-the-art 

and here are some of their specific characteristics that make them distinguish from the 

others. 

2. DIGITAL SYSTEMS FOR STRUCTURAL MONITORING 

Today, the structural monitoring can be achieved using only digital systems. These systems have 

some specific technical characteristics that must cover the needs of such monitoring/analysis. Next we 

will analyze those characteristics that are important for this application. 

2.1 Resolution 

The resolution of the AID converter is an important parameter for digital systems. With respect to this 

parameter the question is: 

Is 16-bit (120dB 

In this case, the peak voltage generated by the sensor for the above theoretical acceleration is: 

a=1256uV

590 

If we compare this peak value with the LSB for the 16-bit A/D with gain 100 and full-scale +/-0.1V 

we can reach the conclusion that there is enough resolution for this type of measurements. 

Figure 2.1.1 shows the time series and the corresponding FFT for a signal generated by a Force 

Balance type accelerometer with a dynamic range >120dB and a full-scale of +/-lg, recorded with 16- 

bit resolution, gain 100, full-scale 4-/-0.1V. The sensor was located in Pasadena, on a quiet street, at 

the ground level, and oriented horizontal. 

Figure 2.1.2 shows a detail of the time series to underline the resolution of the signal 

In both figures time series traces from top down are: 

Ground acceleration in um/s" 

Integrated Velocity in um/s 

Reference velocity in um/s provided by a seismometer with T n = 3 sec and S= lOOV/m/s. 

2OOO OOOOOO -" 

TOOO OOOOOO -j 

O OOOOOO 

-1OOO OOOOOQ 

sooooo 

OOOOOO 

sooooo 

OOOOOO 

sooooo 

OOOOOO 

Time series Acceleration Velocity Reference Velocity 

V' 

Mtyb^^ 

I J I I 

V^v^^^ 

3 OOOOOOJ 

FFTforRef VotandVtf 

SOO OOOOOO - 

3T5 OOOOOO - 

FFT for Acceleration 

•A OOOOOO -j 

3 OOOOOO 4 

a oooooo 

2 SO OOOOOO - 

125000000- 

0000000-," 

oo rs iso aas soo yr 3 

0 OOOOOO _pxjv "w« 

OO TS TSO ZSO 35 O 

Noise signal recorded at ground level 

Figure 2.1.1 City noise recorded at ground level

591 

Time series Acceleration, Velocity, Reference Velocity 

2OOO OOOOOO - 

1000 OOOOOO - 

o oooooo - 

-1OOO OOOOOO - 

ooo 

25 

12 

O OOOQOO • 

-12 SOOOOO - 

25 OOOOOO 

OOOOOO 

SOOOOO - 

•4 OOOOOO - 

FFTfbrRef VdandVel 

SOO OOOOOO - 

FFT for Acceleration 

2 OOOQOO _ 

1 000000- 

0 OOOOOO - 

4 OOOOOO - 

3 OOOOOO - 

2 OOOOOO - 

1 OOOOOO - 

O OOOOOO- 

375 OOOOOO- 

25O OOOOOO- 

12S OOOOOO - 

O OOOOOO- 

Noise signal recorded at ground level 

Figure 2. 1, 2 Detailed time series to underline the resolution of the signal. 

The conclusion of this test is that a 16-bit A/D resolution with gain 1 00 and full-scale set to +/-0.1V is 

suitable for ambient vibration measurements. If the gain is set to 2, which correspond to +/-5V fullscale, 

then the system is suitable for strong motion type measurements. The difference between this 

settings and the use of a 2 1 -bit or greater resolution A/D is the overall dynamic range. In other words, 

the 16-bit system requires different settings for different type of measurements while the 21 -bit or 

greater system covers both applications with one setting. Of course the difference in price is 

significant and can become a cost/benefit factor. 

2.2 Number of Channels 

Detailed structural monitoring requires a high number of channels. The last two systems (See Table 

1.1) have standard 64 channels and optional 128

592 

2.3 Sampling rate 

Structural monitoring requires a relatively high sampling rate. A minimum of 100 sps (samples per 

second) per channel is required. Again, the last two systems (See Table 1.1) provide up to 500 sps per 

channel 

2.4 Data broadcasting 

For real-time monitoring, the data needs to be broadcasted to multiple remote stations. Each station 

should have software capable of recording, analyzing, and comparing data with different userspecified 

design limits and take different actions based on exceedance criteria. 

3. ENGINEERING NEEDS 

From engineers point of view the structural monitoring has different aspects. This is because different 

specialties are interested in: 

• Overall structural response 

• Structural elements behavior 

• Structural elements fatigue (long term monitoring combined with statistical analysis) 

• Soil-structure interaction 

This information is provided by different type of sensors. The system must be able to accept signals 

from all sensors, group them on categories, distribute remotely the signals to each user based on the 

field of interest, and provide the specific analysis tools for each user. All these functions are to be 

handled by the Central recording system (called server) and distributed in real-time to remote stations 

(called clients). 

Again, the last two systems (See Table 1.1) cover all these requirements. 

4. TURN-KEY SYSTEM 

Digitexx Data Systems, Inc. developed two systems (one with 16-bit and the second with 24-bit 

resolution) capable of performing ail functions described above. The system has been developed to 

use COTS components. Digitexx developed the software package following the concept presented in 

Figure 4.1.

593 

Server 

Data 

Acquisition 

t 

Sensors 

Client Client Client 

System Controlled Full 

Status Access Access 

Figure 4.1 System concept 

The left side in this figure shows the 'server'. The server consists of the hardware (Data acquisition 

and sensors) and the server software. This package is responsible for acquiring data, recording events 

based on trigger threshold exceedance, and distribute the data to multiple users 'clients' in -real-time 

over the internet using TCP/IP protocol and publish subscribe mechanism. 

Once you have the right to connect to the system, you will activate the 'client software'. This software 

package is highly customized to meet end-user requirements. It can perform real-time data analysis, 

data recording using event-driven triggers, on demand, and user-defined scheduled recording. 

These two pictures show a 

building in San Francisco (left) 

instrumented with Digitexx 

Real Time Monitoring System 

Model RTMS-2001RM (right). 

The system has 31 channels 

connected to uniaxial 

accelerometers distributed at 

different floors. The server has 

a DSL connection to the 

Internet.

594 

Any remote station that has the subscribe permission can connect to the server using the client software. 

This software is highly customized and it can perform intensive real-time data analysis. Below is an 

example of client software. 

e 

Picture 4.1 Client software screen shot 

This client software computes in real-time acceleration, velocity, and displacement. It also computes the 

response spectrum for a user-selected channel. Functions as on-demand and scheduled data recording can 

be seen on the same picture. 

5. FINAL REMARKS 

The characteristics which make the systems/concept presented in this paper state-of-the-art solution are: 

• High number of channel count 

• Reliable data acquisition and real-time data broadcasting 

• Simultaneous real-time connection of multiple clients 

• Real-time intensive data analysis focused on initial and cumulative damage assessment 

• Long term structural health monitoring 

• Exports data directly to standard commercial analysis software packages 

The authors acknowledge the scientific input received from Prof. W.D. Iwan (Caltech), Prof. C. E. 

Ventura (UBC), Prof. M. D. Trifunac (USC), and Dr. M. Celebi (USGS).




APPLYING E-MONITORING INFORMATION SYSTEM FOR 

HIGHWAY BRIDGES POST-EARTHQUAKE DAMAGE 

EVALUATIONS AND WARNINGS 

Guan-Chyun Shiah 1 , Sheng-Fuh Perng 2 

'Department of Civil Engineering, National Kaoshiung University of Applied Science 

Department of Civil Engineering, National Kaoshiung University of Applied Science 

ABSTRACT 

An e-Monitonng information system is Remote Monitoring using Wire-less technology on the Internet 

in which it utilizes the modern technology such as global positioning system (GPS), geographical 

information system (CIS), wireless communication technology (WCT), and mobile displaying system 

(MDS). This integrated system requires Pre-earthquake installations on the expansion joints, bridge 

piers, and bridge footings. When certain scale earthquake motion are detected by strong-motion 

instruments, using wireless or internet transmitting network to send out the signal to transportation 

controlling system for feasible signal control actions. During the earthquake loads period, the bridge 

structural crack monitoring system should according to its preset crack signal warning thresholds to 

decide on emitting warning signals to bridge authorities or disaster rescue teams in the nearby. 

INTRODUCTION 

The objective of a remote earthquake supervisory control system for highway bridges is to let highway 

transportation authority staff can immediately obtain the bridges' post-earthquake damage information 

and handle a remote system in a easier way. A remote E-monitoring control system, in general, consists 

of supervisory units and a central control unit. Traditional systems use private network to guarantee the 

quality and performance for the communication among all units i it always results in too many cables 

and great costs. ( Ref. 1 ) Also, the maintenance quality of the whole system affect the performance and 

the reliability of the system. To overcome the above drawbacks, an e-Monitoring earthquake information 

system in which combined GPS, GIS, WCT and MDS is proposed in this study. The proposed e- 

monitoring system consists of a central control unit, a supervisory controller with a gateway to internet 

or PSTN (Public Switched Telephone Networks). The central control unit issues controlling message to

596 

or receives environmental information from the supervisory controller gateway by internet, wireless or 

PSTN. The supervisory controller gateway gathers post-earthquake bridge damage information on 

which came from the supervisory controller unit through blue-tooth communication module, optical 

fibers or cables, and sends them to the central control units. The systematic framework of the proposed 

e-monitoring system for bridge post-earthquake evaluations and warnings is shown in Fig. 2. 

Bridge Damage Lessons Learned From 921 Ji-Ji Earthquake 

It is helpful to gather the historical bridges post-earthquake damage information for constructing a 

remote earthquake supervisory control system for future highway bridges post-earthquake damage 

evaluations and warnings. There are several bridges serious damages were reported after the Ji-Ji 

earthquake that took place in Taiwan on Sep. 21, 1999. After conducting field investigations on about 

990 bridges in Taichung, Changhua, Yunlin and Nantou counties since the 921 earthquake, the research 

team found that nearly 200 bridges had sustained damage. "At least 25 bridges have been classified as 

'extensively damaged' and nearly 70 percent of them were built prior to 1987, when older regulations 

governing earthquake-resistant construction was in force. Bridges within 10km of the Chelungpu fault 

were more vulnerable than elsewhere during 921 earthquake. Severe ground motion accompanied by a 

sudden slip of the fault on Sept. 21 led to a number of different types of damage, including bridge 

collapses, abutment tilt, pier breaks and crossbeam displacement. Scientists said that since it is hard for 

engineers to avoid building bridges right over fault lines, which are scattered over across the country, 

and that a better way to prevent disasters would be to build stronger bridges meeting stricter regulations. 

Conclusions drawn from scientists and engineers suggest reclassifying the existing four earthquakeresistant 

construction categories into two stricter categories. 

In 2000, the construction regulations were revised, the highest earthquake-resistant coefficients would 

be higher than the past threshold - which is 0.33G horizontal ground acceleration. Areas near the 

Chelungpu fault certainly would be included in a new "Category 1," which would stand for the highest of 

the two new proposed categories. Investigators found that different bridge designs resulted in different 

kinds of damage, and therefore adopting innovative ways to build bridges should be considered in the 

future. In the past, engineers tended to widen existing bridges without strengthening supporting piers 

first, and me method resulted in disasters during 921 quake. The reason for the collapse of bridges built 

by the Highway Bureau during the quake is because it adopted the simply supported beam system of 

construction. This type of design is vulnerable during quakes. If engineers were to adopt a better system, 

the continuous-beam system would be best for future bridge construction projects.

597 

The Necessity of Highway Bridge Instrumented With Earthquake Early Warning And Monitoring 

System 

Early detection of highway bridges' earthquake damage or degradation prior to local failure can prevent 

"runaway" catastrophic failure of the bridge system. In the event of an earthquake, the rapid detection of 

catastrophic failure of a strategic bridge can allow early and effective response by highway officials in 

rerouting traffic and avoiding an otherwise difficult situation. The large physical sizes of the bridges 

necessitate an extensive array of different 

sensors and appropriate technologies for data 

acquisition/reduction for rational health monitoring applications. The bridge earthquake damage 

monitoring systems should also be designed for a long life under severe environmental conditions. The 

system should be able to automatically detect, locate and assess structural damage anywhere within the 

bridge system, and to communicate the status (alerting) to responsible authorities. At its simplest 

application, the threshold exceed anywhere in the bridge is monitored and an alarm is provided if pre-set 

threshold levels have been exceeded. This alarm may be used to automatically divert the traffic if it 

indicates damage due to earthquake, explosion or impact. 

THE PROPOSED AUTOMATED EARTHQUAKE E-MONITORING SYSTEM IN TAIWAN 

Taiwan Central Weather Bureau (CWB) can now routinely obtain earthquake information in one minute 

after the occurrence of an earthquake using this new algorithm. In order to take full advantage of this 

capability, four automatic information broadcast media, namely, E-mail, World Wide Web, FTP server 

and pager system, have been configured to receive and transmit automatically the earthquake 

information from the CWB seismic monitoring system.(Ref 1). See Fig. 2. This new automatic 

earthquake information Warning and Monitoring system will enable the CWB to disseminate 

information about felt earthquakes even more quickly and widely than its current practice through fax 

and paper reports. 

Lessons Learned from Early Earthquake Warning And Health Monitoring Systems 

In all types of bridge health monitoring system, the early earthquake warning and monitoring system has 

a basic framework consisting of (1) a broad spectrum of monitoring techniques, (2) a mechanism for 

collating and evaluating this disparate data, (3) a routine for converting the relevant data into a 

prediction or a forecast of an impending hazard, and (4) a system for issuing the relevant warning to 

those at risk.

598 

Comparisons of Proposed E-Monitoring System with The Other Bridge Monitoring Tech 

Currently, there are various bridge health monitoring systems named "Traditional 

Wired & Client Server Type (TWCS ) , Metal Time Domain Reflectometry (MTDR ) , Optical Time 

Domain Reflectometry (OTDR) , and Bluetooth Supervisory Wired type.(Ref 2) Each of the above 

mentioned bridge monitoring system has its limitations and drawbacks which can be summarized as the 

followmgs: 

Twcs 

• Many wires are need for various sensors, simple client-server type connections. 

* Costly installation and maintenance. 

Mtdr 

Signal decay increased with the sensors' distance. 

Vulnerable for corrosion. And, Electro-Magnetic interference 

Otdr 

* Costly installation and maintenance. Also, Poor mechanics behavior 

• Signal loss hardly count. And, Not yet reach practical application stage. 

The proposed E-monitoring system for highway bridge expansion joint have the following Merit 

features for Many Taiwan existing simple supported bridges 

•Expansion joint movements 

•Measured Concrete and Ambient Temperatures, increasing the accuracy of data 

•Battery power system 

•Preset time interval measurements and transfers the data daily using Wire-less technology from any 

remote site to monitoring station 

•Simple to install and starts data logger within 1 hour. Fast installation response 

•Developed for Tropical Climate in which meet the industrial standards.

599 

Communication 

Information Manioulating 

• Satellite / \ • Decision 

• Wireless / Earthquake \ • GIS 

• Wired / Induced \ * Cnsis Management 

/ Bridge V Public Relation 

( Damage ) 

Affilate facilities 

\ 

\ 

Response 

Ppntor 

/ 

Image Procegîng 

• UPS \ 

• LCD display 

• Automatic Power \ 

/ Image Transmitting 

Supply 

• Net-Meeting 

• Seismic Isolator 

FIG 1 AN INTEGRATED EARTHQUAKE INDUCED BRIDGE DAMAGE RESPONSE SYSTEM

600 

LCD display Earthquake Warning 

load 

Bndgpspctor 

I "~* - " - /- L 

- Ti F ' 

ôad 

lu -r^ 

Rich s Scale 3U J- 

a 

Acceierometer 

y 

-—- .—, Electrolytic 

Vibrating Wire Sensors Tilt sensor 

tL 

v Data loggen with 

0 *- W reless Modem 

^ 0120-400 m 

i y , 

trasm tt nq Distance 

> " Server 

Crack: sensor 

Inclir ometer 

Supervisory Modem 

i /) gateway 

Smart Me^nitonng W i r e l e s s ^ 

RTQ 

communication S m s Alarm 

Tech GPRSM essa ge for 

(fatjre) 

Aijthoring 

*-*-^. — ^ 

/ 1 

Y GSM(current] ( 

^., ṙ.. - 

j^ ^< 

^SPRS 

Bridge Health 

Monitoring System 

WWW 

PDA 

Central Weathe 

Bareau(CWB) 

r 

Internet 

\ 

Early Earthquake 

Warning System for — 

Bridge Damage 

Microwave 

Internet 

5Ar.eiv«r 

r,._ 

•**== 

J server 

f W reless 

t microwave 

^ 

Taiwan strong 

Radio 

Monon Net 

0 

_L Serial Communication 

I3H — 

Wireless 

Modem 

L qjmcaîon 3 axis 

sensors seismic Accelerator

601 

TABLE 1 

THE COMPARISONS OF VARIOUS BRIDGE MONITORING SYSTEMS 

Technology 

Types 

Traditional 

Wired & 

Client Server 

type 

MTDR 

OTDR 

Blue-tooth 

Supervisory 

Wired sys 

E-momtonng 

With SMS 

alarm 

Economic Maintenance Reliability 

3 

4 

4 

3 

2 

4 

3 

4 

2 

2 

3 

3 

4 

2 

2 

Data 

Completeness 

Remark Weighting scores from 1-5 represents the best to the worst 

3 

2 

2 

2 

2 

Accuracy Easy Total Rank 

Installed 

3 

2 

2 

2 

2 

3 

4 

4 

2 

1 

19 

18 

20 

13 

10 

4 

3 

5 

2 

1 

Reference 

1 Yih-Mm Wu 1 Tzay Chyn Shin 2 Yi-Ben Tsar 3 Kai-Wen Kuo 4 On the establishment of anautomatic earthquake 

information broadcast system in Taiwan 

2 Yea Chy Chang Kao-Shing Hwang el The Implementation of Bluetooth Technology on Supervisory Control 

and Data Acquisition System MS Thesis National Chung Cheng University June 2001 

3 Dowding, C H andFC Huang 1994 Early Detection of Rock Movement with Time Domain 

RQflectQmetty, Journal of Geotechnical Engineering, ASCE, v 120, No 8, pp 1413-1427 

4 James C I Dooge, Lessons learned from early warning systems What do they have in common 

5 Peter L FuhrDryrerR Huston (1993) Stress Monitoring of Concrete Using Embedded optical Fiber Sensors 

Journal of Structural Engineering Vol 119 No 7 pp 22622269




SMART HEALTH MONITORING SYSTEM 

OF A PRESTRESSED BOX GIRDER BRIDGE 

Xuan Wang, 

Ming L. Wang 

Dept. of Civil and Material Engineering 

University of Illinois at Chicago, 842 W. Taylor St 

Chicago, Illinois 60607, USA 

ABSTRACT 

With increasing technology, many non-destructive testing methods have been developed for the health 

assessment of bridges. These methods range from the simplest ones that cover regular inspections and 

visual controlling to highly sophisticated methods that require expensive equipment and highly skilled 

personnel. This paper presents experiences gained from the health assessment of a pre-stressed boxgirder 

bridge. Dynamic and static methods, including two load tests, were used in evaluating the 

bridge's health. From these experiences, a new Smart Health Monitoring System was developed and 

applied in the real-time health monitoring of the Box-girder Bridge. Because this new system can preprocess 

all the raw data from each sensor, it improves largely the efficiency of data analysis. Also, it's 

able to supply the real-time health status of the bridge and all the statistical data on-line. 


The rating of damage severity should be based on how much the damage influences safety of the 

structure. Besides defects that directly threaten structural safety, there are damages that do not hinder 

proper performance of the bridge under normal traffic conditions but they decrease its safety margin, 

which means that the structure does not have required reliability. In such cases, possible overloading 

like traffic congestion may lead to the collapse. There are also some damages associated with 

durability that accelerate the deterioration processes in structural materials and finally may cause 

serious failure. An example is cracking in prestressed concrete structures exposed to severe weather 

conditions that may lead to the corrosion of prestressing reinforcement, reduction of steel area and 

finally rupture, e.g. sudden collapse of Bickton Meadows footbridge in Hampshire, UK in 1967 caused 

by corrosion of prestressing tendons. Traditional monitoring method is to measure the responses of a 

bridge periodly. It can't find and respond to any emergency happened on the bridge. And the raw data 

collected from various kinds of sensors is too much to be analyzed easily. Hence, engineers need a 

kind of new smart health monitoring system that can show the real-time health status of bridges. Also, 

the new system can preprocess all the raw data and calculate daily, weekly, monthly or annual 

statistical data.

604 

The Smart Monitoring System (SMS) was developed by the Bridge Research Center at the University 

of Illinois for IDOT under the title Development and Installation of a Permanent Health Monitoring 

Station for the South-Bound Kishwaukee Bridge. This paper describes the following aspects of the 

monitoring system that were developed as a part of this project: capabilities, sensors and locations, and 

analysis. The analysis of data obtained from the installed system reveals the health condition of the 

bridge between December 2001 to March 2002. The system consists of a rack-mount UNIX-based PC 

(top rack) which controls a multi-function data acquisition card and modem. Signal conditioning 

modules and anti-aliasing filters are contained in a separate enclosure (lower rack). The diagram of 

SMS was shown in Figure 1. 

Bridge Computer Station 

Software Functions 

1 Process 

J Diagnostic 

Algorithms 

Hardware Functions 

, Acquisition 

•decimated data stream 

updated hourly 

•inferred bridge status 

updated daily 

UIC Computer Station 

Hardware Functions 

Software Functions 

temperature 

iocal strains 

crack displacements 

acceleration (freq.) 

video snapshots 

IDOT 

Computer 

Station 

• system status 

• video frames 

• daily monitoring status 

• low-bandwidth data 

available in spreadsheet 

format 

Wired 

Ethernet 

Standard 

Windows 

PC 

Fig. 1 Schematic Diagram of Smart Monitoring System 

The objectives of Smart Monitoring System (SMS) are to monitor and evaluate the structural health 

status for Kishwaukee River Bridge, and to provide information for facilitating the inspection and 

maintenance activities. The SMS can collect and preprocess the raw data of structural responses such 

as strain, crack opening displacement and acceleration, in order to: 

1) Characterize the temperature influence on strain, crack opening displacement, and acceleration. 

2) Characterize the traffic-related inputs and responses. 

3) Establish the global frequencies of the structure. 

4) Provide the information on presence and extent of crack growth and state of shear reinforcement in 

the web. 

5) Provide the global information useful for FEM model updating (important for damage sensitivity). 

6) Locate regions that require more in-depth scrutiny through diagnostic tests and determine 

additional locations that require continuous monitoring. 

7) Provide information that will aid the owner and maintenance authority to make rational decisions 

when assigning the maintenance and repair budget. 

2 DESCRIPTION OF THE BRIDGE AND SENSOR LOCATIONS

605 

Kishwaukee River Bridge consists of two independent structures: the southbound bridge and 

northbound bridge. These two bridges were made by post-tensioning of single-cell precast segments. 

The deck of the bridge has five spans. The overall length of the deck is 334 m (1096 ft). The deck was 

built by the balanced cantilever method. Each cantilever consisted of seventeen segments 2150 mm 

(7'- 3 / 5 ") long and one pier segment 1067 mm (3'-6") long. Cast-in-place closures have a length of 984 

mm (3'-2 3 / 4 "). Figure 2 shows the longitudinal layout of the Kishwaukee Southbound Bridge with four 

piers. There are four kinds of sensors installed in the Kishwaukee Bridge, as shown in Figure 3: 

1) Electrical Resistance Strain Gages; 2) LVDT sensors (Linear Vibration Displacement Transducer); 

3) Thermocouples; 4) Accelerometers. 

76.2 76.2 76.2 51.82 

"1 

orth 

mtme 

nt 

n 

1-1 

Pier SB4 

Pier SB3 

n 

Pier SB2 Jl Soutl 

abutn 

Pier SB 1 

Fig.2 Longitudinal layout of the Kishwaukee Bridge (Units: foot) 

North Pier 4 

Abutment 

' 

Pier 3 

SMS 

Workstation 

1 

0 

Strain gage 

•»- 

• LVDT 03 gage - 

..;...: 

J 

i«tn 

0 Accelerometer 

Pier 2 

• Thermocouple 

, 

Fieri 

Fig. 3 Sensor Location in Kishwaukee Southbound Bridge (Plain view from top) 

u South 

Abutment 

3 DATA ANALYSIS OF STRUCTURAL MONITORING SYSTEM 

Figure 4 shows the flow chart of data analysis of Structural Monitoring System (SMS). As presented in 

the chart, the raw data will be divided into four parts according to the fixed channels: 1) Strain; 2) 

Crack opening displacement; 3) Acceleration; 4) Temperature. For every part, the real-time data can be

606 

shown on-line by the software DACview of SMS. SMS can store the data from the four kinds of 

sensors for one year. And based on the real-time data, the preprocessing program installed in the 

workstation of the bridge can separate the strain and displacement data into two parts as either the 

traffic effect or the temperature effect. The acceleration and temperature data also will be preprocessed 

for the purpose of efficiency. Moreover, SMS can automatically analyze and calculate the daily, 

weekly, monthly and annual statistical data. 

i 

j 

Bending Stttfness 

Max/Mm strains 

Strain | 

FEM Updating 

Number ot trucks 

Real-time Display 

/* 

Young s Modulus of concrete Truck Speed 

Data stored tor tracking (one vear) 

Compare to preset values 

Truck Weight 

Data pre processed for efficiency 

Warning for abrupt change 

v earlv monthly weekiv and daily summary Traffic Effect 

Temperature Effect 

Max/Mm strains 

Strain/degree change( "C ) 

Real time Display 

Stress in Shear Reinforcement 

Data stored for tracking (one year) 

Max/Mm crack opening displacement 

Shear Stiffness 

Number of trucks 

Data pre processed tor efficiency 

Compare to preset values 

Yearly monthly, weekly and daily summary 

Truck Weight 

Acceleration 

Real-time Display 

Data stored r or tracking (one year) 

Traffic Effect 

Temperature Effect 

Warning for abrupt change 

Max/Mm crack opening displacement 

Displacement/degree change 

Data pre processed tor efficiency 

Max/Mm G level 

Yearly monthly weekly and daily summary 

Frequency Analysis (the first 5 frequencies) 

'—i*j Temperature j 

Compare to preset frequency at prescribed temperature 

Red-time Display 

Warning tor unusual change of acceleration (Traffic lam or accident) 

Data stored for tracking (one year) 

Data pre processed for efficiency 

Yearly monthly weekly and daily summary 

Max/Mm temperature 

Inside/outside difference 

Outside/concrete difference 

Warning tor unusual change of temperature 

Fig. 4 Flow Chart of Smart Monitoring System 

3.1 Strain Data Analysis 

Due to the traffic and temperature effect, strain data were preprocessed and separated into two parts, as 

shown in Figure 5 and Figure 6. Based on the relationship between strain and truck weight, we found 

that a strain record over 4 microstrain represents a truck passing through the location of the strain gage. 

In order to verify the threshold of strain, the actual number of trucks and the time when the trucks 

passed the bridge were recorded on March 30, 2002 from 2:00 pm to 3.00 pm. Figure 7 shows the 

comparison between these two records of truck numbers. In addition, it can be observed that the onthe-spot 

record matches the result from the strain analysis. Since strain gages are located on SB2-N16 

and SB2-S16 and the distance between these two locations is known to be 225.6 ft, the speeds of 

trucks can be obtained by calculating the time difference between two strain peaks at these two 

locations. Figure 8 presented the record of truck speed in March 2002. As observed in the figure, one 

truck's speed is 51.3 m/h. According to the load test in 2000, the maximum positive strain in the web 

of the segment SB2-S16 of Kishwaukee Bridge, under two three-axles trucks each weighing 96,000 

Ibs, turned out to be 24 microstrain. And the maximum was 16.92 at the same location. Tablet shows 

the comparison of monthly maximum strain and curvature with the result of Load Test. As shown in

607 

the table, the monthly maximum values are close to the result of Load Test But none of them exceeds 

the value of Load Test. 

Maximum Strain is 43 7 microstram 

Minimum Strain is - 40 8 microstram 

""Maximum SfrairTis 24 microstrairTforlhelbad test' 

The temperature difference is 18 °C in this month 

"ZJ^r" 

2 3 4 5 6 7 8\ 9 10 11 12 13. 14 15 16 17/18 T-EL.20 21/22 23 24 25 26 27 28 29 

JIhe-Cban.ge-of stcain-iS-5 min 

temperature change of 1 °C 

Time (day) 

Fig. 5 Monthly Record of Strain on the East Web of SB2-S16 

due to Temperature Effect (March, 2002) 

Maximum positive strain was 20 4 mtcrostram during the month 

The positive strain of over 18 microstram was recorded 20 times 

Maximum positive strain was 24 microstram for the load test 

Fig. 6 Monthly Record of Maximum Positive Strain on the East Web of 

SB2-S16 due to Traffic Loads (recorded every 5 minutes in March, 2002)

608 

One truck count is based on the recorded strain value 

of over 4 microstram 

Q Actual Counting 

Record - 80 trucks 

D Monitoring System 

Record - 82 trucks 

Time (minutes) 

Fig. 7 Comparison of Number of Trucks based on Actual & Monitoring Records 

StratnofSB2-S16 

Maximum 102 

Minimum-15 2 

Strain on SB2-N16 

Maximum 162 

Minimum -20 9 

(Beginning Time Mar 4, 2002 10 40 00 pm) 

] 

Distance 225 6 ft, Time 3 s, Truck Speed 51 3 m/h 

Time (second) 

Fig. 8 Time Record of a Truck Passing SB2-N16 & SB2-S16 on Mar 4 

Dec 

Jan 

Feb 

Mar 

Table 1 Comparison of MontMy Max. Strain and Curvature with Load Test Result 

Maximum Strain Maximum Strain Maximum Curvature Maximum Curvature 

Monthly 

In Load Test Monthly in Load Test 

(microstrain) (microstrain) 

20.5 

24.0 

14.39 

16.92 

18.3 

24.0 

12.84 

16.92 

19.6 

24.0 

13.75 

16.92 

20.4 

24.0 

14.32 

16.92

609 

3.2 L VDTData Analysis 

LVDT sensor installed on the web of segment SB2-N4 can measure crack opening displacement. 

Based on the record, we can analyze if the growth of shear crack on the web has been controlled. 

Traffic loads and temperature change both have a significant effect on the change of crack opening 

displacement. Hence, the original LVDT data was separated into two parts. According to the load test 

hi 2000, the maximum crack opening displacement on the web of SB2-S16 was 0.0020 inch. Table 2 

shows the comparison between monthly maximum values and the load test result on the web of 

Segment SB2-S16. 

Dec 

Jan 

Feb 

Mar 

Table 2 Comparison of Monthly Max. Crack Opening Displacement with Load Test 

Maximum Change of Maximum Change of Crack 

Crack Width Monthly Width due to Load Test 

(inch) 

(inch) 

0.0021 

0.0020 

0.0022 

0.0020 

0.0023 

0.0020 

0.0022 

0.0020 

Maximum Truck 

Weight in Load Test 

(LBS) 

96000 ( x2 ) 

96000 ( x2 ) i 

96000 ( x2 ) 

96000 ( x2 ) 

The above data shows that since December 2001, the monthly maximum values of crack opening 

displacement on the web of SB2-N4 have been over the change of crack width measured in the static 

load test of 2000. It seems that the shear cracks on the webs are still the main problem for the health of 

this bridge. The vibration caused by high-speed heavy trucks has an effect on the growth of crack 

width. 

3.3 Acceleration Data Analysis 

Based on the record, it's clear that the vertical acceleration is the largest and the longitudinal 

acceleration is the smallest. This is due to the fact that heavy trucks excite the bridge mainly in the 

vertical direction. Because the longitudinal motion excited by the traffic load was too small, the signal 

corresponding to this direction was very difficult to capture. The comparison of vertical natural 

frequencies from 1986 to 2002 was shown in Table 3. Higher frequencies were measured in the winter 

while the lowest were measured in May. The modal frequencies of this bridge didn't seem to change a 

lot over these years. Those measurements are consistent with the effect of reduced temperature on the 

modulus of elasticity of concrete. 

Table 3 Comparison of vertical natural frequencies 

Frequency 

'Measured by CTL 

*'Measured by UIC 

Smart Monitoring System 

FEM result 

l.mode 

1.61-1.65 

1.60-1.65 

1.62 

1.650 

2.mode 

2.06-2.08 

2.05-2.10 

2.07 

2.132 

3 .mode 

2.64-2.66 

2.65-2.70 

2.68 

2.767 

4.mode 

2.95 

2.94 

3.010 

1} Measurements made by Construction Technology Laboratory in 1986 

* } Measurements made by Bridge Research Center of UIC in 2000 

5. mode 

3.90-3.98 

3.95-4.00 

3.93 

4.032 

5 CONCLUSIONS

610 

1) According to the SMS records due to traffic effects, the maximum values of positive strain on the 

webs of SB2-N16 and SB2-S16 are close to the maximum strain (24 microstrain) in the load test of 

2000. Moreover, the monthly maximum crack opening displacements have exceeded the maximum 

value (0.002 in) measured in the load test since December 2001. Although the flexural stiffness of the 

bridge is still unchanged, some measures should be taken to control the growth of shear cracks on the 

web within a reasonable limit. It is recommended to impose a restriction to the speed and weight of 

trucks going across the bridge. An automated camera system could be added in our Structural 

Monitoring System to photograph any vehicle surpassing the threshold of speed or weight that would 

cause harm to the bridge. 

2} ) Temperature change has a significant effect on the change of strain and crack opening 

displacement. Based on the records of Structural Monitoring System, the maximum change of strain 

due to temperature effect is from - 40.8 to 43.7. The corresponding temperature change is 18 °C. So 

the average change of strain is about 5 microstrain with the temperature change of 1 °C. Similarly, the 

change of crack opening displacement is about 0.0005 in for every temperature change of 1°C. 

3) The dynamic properties are almost the same as the results measured in 2000. The modal frequencies 

of this bridge seem to have stayed relatively constant over these years. 

4) The advanced monitoring system can automatically pick up and analyze the data and transfer the 

processed data from the bridge through a specified telephone line. It can count the number of trucks, 

calculate their speed and record the exact time upon which they pass the sensor location. It's quite 

useful for monitoring the traffic on the bridge in real time. Using this system as a basis, the next step 

will be to build a new traffic warning system. By embedding a decision-making system in the sensory 

system, this new system will be able to automatically react to the strain or acceleration signal 

exceeding the predefined threshold. For example, in the case of a heavy traffic jam or any severe 

accident happening on the bridge, it can send an on-line warning to the Department of Transportation 

immediately. 

References 

1. D. J. Ewins, (1985) Modal Testing: Theory and Practice, John Wiley, New York. 

2. R. Shepherd and A. W. Charleson, (1971) "Experimental Determination of the dynamic Properties 

of a Bridge Substructure", Bulletin of the Seismological Society of America, 61,1529-1548. 

3. R. Kohoutek, (1993) "Tests on Bridge over Talbragar River at Dubbo", Proceedings of the 11 th 

International Modal Analysis Conference, Kissimmee, FL., 1168-1174. 

4. P. K. Lee, D. Ho and H. W. Chung, (1987) "Static and Dynamic Tests of concrete Bridge", ASCE 

Journal of Structural Engineering, 113, 61-73. 

5. C.R. Farrar, W. E. Baker and T .M. Bell, (June 1994) "Dynamic Characterization and Damage 

Detection in the 1-40 Bridge over Rio Grande", Research report Los Alamos National Laboratory, 

LA-12767-MS. 

6. Wang, M. L, F. Xu, and George M. Lloyd, (July 2000) Results and Implications of the Damage 

Index Method Applied to a Multi-Span Continuous Segmental Prestressed Concrete Bridge, 

International Journal of Structural Engineering and Mechanics, Vol. 10, No. 1, 37-52. 

7. Wang M. L., and D. Satpathi, (Sept. 1997) Damage Detection of a Model Bridge Using Modal 

Testing, Structural Health Monitoring—Current Status and Perspective, Technomic Publishing Co. 

Edited by F. K. Chang, pp. 589-602. 

8. Wang, MX., F. Xu and G. Lloyd, (Aug. 2000) Health Assessment of a Post-tensioned Concrete 

Bridge, In proceedings of ASCE's 2000 Structure Congress, Philadelphia, PA.




*'' 

SIGNATURE RECOGNITION OF STRUCTURAL DAMAGE: DATA 

ANALYSIS AND MODEL-BASED VALIDATION FOR 

DESTRUCTIVE FIELD TESTS 

Ray Ruichong Zhang 1 , Matt MacRostie*, and Y.L. Xu 2 

1 Division of Engineering, Colorado School of Mines, Golden, CO 80401, USA 

2 Department of Structural Engineering, Hong Kong Polytechnic University, Hong Kong 

ABSTRACT 

This study uses the Hilbert-Huang transform (HHT), a method for nonlinear, nonstationary data 

processing, to analyze recordings of destructive vibration tests of substructures in Trinity River Relief 

(TRR) Bridge in Texas in its intact, minor- and severe-damage stages. It shows that the HHT method 

can identify the natural frequencies of the structure from the mixed frequency content in a recording 

that also contains the time-dependent excitation and noise frequencies, and then to quantify the 

downshift of frequencies for damaged structure relative to frequencies in the non-damaged structure. 

The above assertion is also validated by an ANSYS model-based analysis. 

INTRODUCTION 

The most common damage of bridge piers and abutments is caused by scour from floods, which cannot 

be visualized nor calculated by normal hydraulic and geotechnical analysis procedures. Using 

vibration measurements might aid in the detection of the damage. Without appropriate data analysis 

methods, however, it will not identify minor damage. For example, conventional (e.g., Fourier-based) 

data processing/analysis techniques may yield distorted, indirect, or incomplete information about 

vibration motion that is inherently nonstationary and also likely the result of a nonlinear dynamic 

process. This will mislead the consequent use of the data for damage signature recognition. 

Because of its ability to faithfully characterize nonlinear (time-varying amplitude and frequency) and 

nonstationary data, the HHT (Huang et al. 1998) can provide an alternative tool for data analysis and 

subsequent applications to damage-signature recognition. The HHT method builds on Empirical Mode 

Decomposition (EMD) and Hilbert Spectral Analysis (HSA). Based on the local characteristic time 

scale of any complicated time series, the EMD decomposes the data set into a finite, often small 

number of intrinsic mode functions (IMF) that admit a well-behaved Hilbert transform. The HSA of 

each IMF then defines instantaneous or time-dependent frequencies of the data that have physical 

meaning, unlike the HSA of the original data (Huang et al., 1998). The EMD, HSA, or their 

combination referred to as the HHT typically helps recover useful information from the data sets under

investigation and subsequently improve understanding of the underlying physical process. In this 

study, we use the HHT to characterize the structural damage from the recordings and model-based 

simulated data of controlled field vibration tests of two substructures in Trinity River Relief (TRR) 

Bridge in Texas in its intact, minor- and severe-damage stages. 

612 

HHT ANALYSIS OF DESTRUCTIVE VIBRATION TESTING DATA 

The bndge for the case study is the Trinity River Relief (TRR) Bridge, located on old US Hwy 90 on 

the west side of Liberty, Texas. The three stages of the bent shown in Fig. 1 is in order: 

(1) Intact stage: the column with sensor 15 was not broken and soil around the pile had the same 

height as the soil around the pile with sensor 13. 

(2) Minor-damage stage: the pile with sensor 15 was not broken, but soil around the pile was 

excavated, which was simulated for flood-induced minor scour with partial loss of load capacity. 

(3) Severe-damage stage: the pile with sensor 15 was broken with the steel bars left only and soil 

around the pile was excavated, which could be regarded as equivalent damage of the column due to 

earthquake-induced liquefaction or severe scour with complete loss of load capacity. 

Figs. 2a and 2b show respectively the time histories of vertical excitation force and its corresponding 

vertical vibration response at sensor 15 with the structure in intact stage. Both time histories are 

highly nonstationary, with the frequency sweeping almost linearly from 4 Hz at 0.3 s to 72 Hz at 5.7 s, 

referred to as chirp frequency for later use. Therefore, their Fourier spectra are unable to provide 

faithful time-dependent frequency content of the data. 

Fig. 3 a shows that the excitation contains a dominant energy with the chirp frequency increasing with 

time from 4 Hz at 0.3 s to 72 Hz at 5.7 s. The excitation also has energy at other frequencies such as 

high frequencies ranging from 20 to 50 Hz between 0.4 and l.ls shown in Fig. 3a, which can be 

visually verified from the excitation time history shown in Fig. 2a. 

It is known that the frequency content of the vibration response should contain primarily both the 

driving frequencies and the natural frequencies of the structure. This is confirmed by Fig. 3b. 

Specifically, Fig. 3b shows the energy with a primary frequency content linearly increasing with time, 

which is the signature of excitation with the chirp frequency. By comparing with Fig. 3 a, the signature 

of excitation in Fig. 3b can also be found at frequency band of 20-30 Hz between 0.4-0.8 s and of 0-30 

Hz between 4.7-5.7 s, among others. In addition to the above excitation-inherited energy, Fig. 3b also 

illustrates energy concentration in the frequency range of 10-20 Hz (others in 20-75 Hz) between 1-2 s. 

This is not inherent from the excitation in Fig. 3 a but rather the energy contributed from the structural 

vibration modes at a couple of lower (higher) natural frequencies, or the signature of the structure. 

This, together with the subsequent interpretation, will be verified later by a model-based analysis. The 

aforementioned vibration with 10-20 Hz at 1-2 s is likely excited by the force with a low-frequency 

band in 4-15 Hz at 0.4-1 s (see Figs. 2a and 3a,b). As time goes on (say 2-3 s), the vibration at the 

low-frequency band dies down quickly due to damping, or is too small to be shown in the figure in 

comparison with strong excitation-inherited energy. Since the first mode of vertical vibration motion 

should be dominant in the low-frequency band, the above observation and assertion suggests that the 

fundamental natural frequency can be found from the Hilbert amplitude spectrum by identifying the 

corresponding dominant structural energy, particularly around 1-2 s. 

Fig. 4a enlarges the Hilbert amplitude spectrum in Fig. 3b. Except the excitation energy at chirp 

frequency and at high-frequency band of 15-30 Hz between 0.5-0.7 s, all the other energy 

concentration shown in Fig. 4a is attributed primarily by the structure itself. We now focus on the

structural response energy with frequency up to 15 Hz during 1-2 s. The structural response energy 

below roughly 12 Hz (light lines or dots) is much less than that above 12 Hz (dark lines or dots). The 

response energy is generated likely by the force with the chirp frequency during 0.4-1 s and with the 

frequency band in 5-15 Hz at around 1+ s in Fig. 3a. If the fundamental natural frequency falls in the 

driving frequency range of 5-15 Hz, the response energy at the fundamental natural frequency should 

be much stronger than that at frequency lower, but not necessarily stronger than that at frequency 

higher since there exist second and higher natural frequencies and mixed frequencies between driving 

and natural frequencies. Fig. 4a suggests that the fundamental natural frequency is around 12 Hz. 

Fig. 4b shows the Hilbert amplitude spectrum of vibration response at sensor 15 with the bent in 

minor-damage stage. Besides the energy inherited from the excitation force such as at chirpy 

frequency, the response energy during 1-2 s has lowered its dominant frequency to 5 Hz at about 1.4 s. 

The observed 5 Hz in the minor-damage stage could be related to the mixed natural frequencies of the 

bent and pile, since the vibration at sensor 15 should reflect the dynamic characteristics of both the 

whole bent and the local member. The bent with excavated soil, i.e., the bent in the minor-damage 

stage, reduces the stiffness of the pile and thus the fundamental natural frequencies of the bent and 

pile. This is the signature of local damage in the recordings, which is not sensitively picked up by the 

Fourier data analysis. 

Such an explanation can be further strengthened from Fig. 4c, which shows the Hilbert amplitude 

spectrum of vibration response at sensor 15 with the bent in severe-damage stage. As seen in the 

figure, the dominant frequency corresponding to the structural energy has decreased to 3 Hz at 1.7 s. 

As a comparison for damage-signature recognition from recordings that are collected at sensor 15 close 

to the damage location, Figs. 5a-c show the Hilbert amplitude spectra of vibration data at sensor 13 

away from the damage location for the bent in intact, minor- and severe-damage stages respectively. 

Fig. 5 a shows that the dominant frequency to structural vibration energy is around 13 Hz at 2 s, slightly 

higher than 12 Hz identified from Fig. 4a, which could be caused by noise and various uncertainties 

introduced when recorded at sensors 13 and 15. Figs. 5b and 5c show the dominant structural 

vibration energy with lowest frequency of 10 Hz at 2 and 2.2 s respectively. Since sensor 13 is 

removed from the damage location, the vibration characteristics at sensor 13 will retain primarily the 

dynamic features of whole bent, which likely do not change significantly due to the local damage at the 

column with sensor 15. Accordingly, it is understood for the small change observed in the identified 

frequency associated with the dominant structural vibration energy for the bent in three-different stages 

in Figs. 5a-c. 

The above analysis suggests that signature of structural vibration in terms of natural frequencies of the 

whole bent and/or local members can be well distinguished from the driving frequency content by the 

HHT analysis of recordings, which is sensitive to the local damage if the recordings are collected near 

the damage location. 

613 

MODEL-BASED VALIDATION 

To verify the observation and assertion in the last chapter and thus to improve our understanding of the 

HHT-based damage characterization, we performed the HHT analysis of simulated vibration data with 

ANSYS-based two-dimensional (2D) finite-element-method (FEM) models for the bent 12 hi its intact 

and severe-damage stages. Fig. 6 shows the ANSYS-based 2D FEM model for the bent 12 in intact 

stage, which was built on design specifications and pertinent information by Sanayei and Santini

614 

(1998). In this study, we also extended the 2D model from the bent in the intact stage to that hi the 

severe-damage stage by releasing boundary restriction at the ground end of pile 15 in Fig. 6. 

Table 1 shows the first ten undamped natural frequencies of the bent in its intact and severe-damage 

stages, in which the first to the third natural frequencies in vertical directions and their reduction 

percentage in its severe damage are also indicated. To simulate the damping effects, Rayleigh 

damping was used in modeling, i.e., [C] = a 0 [M] + a : [K] where M, C and K are the matrices of mass, 

damping and stiffness, 0 0 and a } are the constant coefficients. We used the same force in Fig. 2a to 

excite the 2D model hi Fig. 6 and obtained vibration response at nodes 53 and 37 in Fig. 6 that 

correspond to the same locations in sensors 15 and 13 in Fig. 1, respectively. 

Fig. 7a shows the Hilbert amplitude spectrum of the response at node 37 of the bent in intact stage with 

no damping, i.e., a 0 = a } = d = 0. The response energy focuses at frequency band with its center 

primarily around 12.0, 19.7, and 69.2 Hz that are respectively the fust three natural frequencies in 

vertical direction in Table 1. This can be further clarified in marginal amplitude spectrum in Fig. 7b. 

Without damping, the bent is fully excited at the natural frequencies. Therefore, the vibration energy 

at natural frequencies is typically higher than the excitation-inherited energy at chirp frequency. 

Consequently, the energy inherited from the excitation at chirp frequency is not clearly shown in Fig. 

7a, except the strong portion with frequency band from 40-70 Hz at 3.5-5 s. With the damping, the 

relath e intensity of the energy from the structure and excitation should be changed, which can be seen 

from the Hilbert amplitude spectra in Figs. 7c,d with a Q = a : = d =0.00198 and 0.05305, respectively. 

Comparing Fig. 7c with Figs. 3a and noticing the fundamental natural frequency of the bent being 

11.98 Hz, we can conclude that the excitation-inherited energy in Fig. 7c is associated with chirp 

frequency and other frequencies such as high-frequency band in 10-30 Hz during 0.3-0.8 s, and that the 

vibration energy is associated with frequency band around 10-13 Hz during 1-1.6 s. The damped 

fundamental natural frequency in vertical direction is identified in Fig. 7c around 11.5 Hz, smaller than 

the undamped one with 11.98 Hz in Table 1 which is consistent with the vibration theories. Due to the 

large damping used and strong excitation at chirp frequency, the vibration energy at natural 

frequencies are greatly suppressed and the excitation energy is inherited, in comparison with the 

undamping case (see Figs. 7a,c). This phenomenon can be further clarified by Fig. 7d with a large 

damping. Fig. 7d shows that the vibration energy at 10 Hz primarily at 1.4 s. Since the damping here 

is extremely high, the damped fundamental natural frequency is greatly downshifted from undamped 

one at 11.98 Hz to the current one at 10 Hz. It should be noted here that it is with the damping 

d=0,05305 that the ANSYS model in Fig. 6 generated time history responses that are similar to those 

in recordings such as Fig. 2b. This implies that real damping of bent 12 in the vertical direction is very 

high. 

Depicted in Figs. 8a,b are respectively the Hilbert amplitude spectra of the response at nodes 37 and 

53 of the bent in severe-damage stage with a Q =a l =d =003979. Fig. 8b shows the vibration energy 

concentrates at 7 Hz from 0.5 to 1.2 s, i.e., the fundamental natural frequency of severe-damage bent 

in Table 1. In contrast, the dominant vibration energy in Fig. 8a is at around 12 Hz between 0.6-1 Hz 

and minor at 7 Hz at 1+ s. The difference of dominant response energy at frequency 7 Hz in Figs. 8a,b, 

i.e., damage-signature from recordings, helps identify qualitatively the damage column and severity.

615 

Table 1: Fundamental natural frequencies of Bent 12 in intact and severe-damage stages, modeled by 

ANSYS software in terms of design specifications 

Natural frequencies in Hz (comments) 

1st 

2nd (1st vertical natural frequency) 

3rd 

4th (2nd vertical natural frequency) 

5th 

6th (3rd vertical natural frequency) 

7th 

8th 

9th 

10th 

Intact 

3 99 

11 98 

14 84 

19 68 

44 01 

69 22 

93 88 

104 45 

106 37 

108 49 

Severe damage (decrease w r t intact) 

3 29 

6 95 (41 98%) 

12 47 

18 08 (8 13%) 

39 19 

52 50 (24 15%) 

72 50 

L_ 95 00 

105 82 

108 17 

Figure 1 (left): Vibration tests of Bent 12 in the Trinity Relief River Bridge under excitations at the 

middle of the deck. 

Figure 2 (right): Bent 12 in intact stage, (a, top) Forcing function and (b, bottom) Recorded 

vibration at sensor 15. 

Figure 3a (left): Hilbert amplitude spectra of forcing function for vibration at sensor 15 of Bent 12 in 

intact stage. 

Figure 3b (right): Hilbert amplitude spectra of vibration at sensor 15 of Bent 12 in intact stage.

616 

Htibert Spectrum or Sensor 15 Intacs 

Hilbert Spectrum of Vibration at Sensor 13 Intact 

Hilbwt Spectrum tor Sensor 15 No Pile 

T7I 

1 15 

Timtistc) 

Figure 4 (left): Hilbert amplitude spectra of vibration at sensor 15 of Bent 12 in (a, top) intact stage, 

(b, middle) minor-damage stage and (c, bottom) severe-damage stage. 

Figure 5 (right): Hilbert amplitude spectra of vibration at sensor 13 of Bent 12 in (a, top) intact stage, 

(b, middle) minor-damage stage and (c, bottom) severe-damage stage.

617 

JUSTS rs s." 

WUUR. 4 20O2 

IS 44.36 

E-BT 

Power Si:ia.ptia.as 

Trj_ixa.cy Bra-d-cye 20/ Bent: 12, Center 

Figure tf: ANSYS models for Bent 12 in the Trinity Relief River Bridge m intact stage, in which piles 

15 and 13 are respectively corresponding to the columns with sensors 15 and 13 in Fig. Ib, and nodes 

53 and 37 to the locations of sensors 15 and 13 in Fig. Ib. 

0 5 10 15 2D 35 30 : 

Frequmcy Hz) a** (0 00075)s 

Tim. I8«0 

a«»»|O.OOOTS)» 

Figure 7: For vibration at node 37 of Bent 12 in intact stage, (a, upper left) Hilbert amplitude 

spectrum with d=0, (b, upper right) marginal amplitude spectrum, (c, lower left) Hilbert amplitude 

spectrum with d=0.00198, and (d, lower right) Hilbert amplitude spectrum with d=O.Q5305.

618 

m ipcdfum-(Bent "2Mode S7brohen-ace0=00399t 

Hubert SOectasn- (Bert 13 Node S& broken-ace d* 003379) 

Figure 8: Hilbert amplitude spectra of vibration at (a, left) node 37 and (b, right) node 53 of Bent 12 

in severe-damage stage with d= 0.003979. 


This study presents the use of HHT to characterize damage of recordings and model-based simulated 

data of bridge structures in their intact, minor- and severe-damage stages. It reveals that the HHTbased 

characterization is able to identify the local dynamic properties from a limited number of 

recordings. In addition, comparison of the HHT analysis of two recordings from damaged and health 

structural elements in one condition stage of the structure can tell the signature difference of local 

structural members. This eliminates the need of a priori data required in traditional damage diagnosis, 

and thus significantly improves the efficiency in data collection. It should be emphasized here that the 

above observations are, at this point, qualitative and suggestive, not quantitative and conclusive. 

Validation requires further examination of other recordings and model-based simulation. 

AKNOWLEGMENTS 

The authors 'would like to express their appreciation for the financial supports provided by the 

National Science Foundation with Grant No. 0085272 and by the Federal Highway Administration 

DTFH61-96-00030. The financial supports are also acknowledged from the strategic research 

program in the Hong Kong Polytechnic University (HKPU) and US-PRC Researcher Exchange 

Program administered by Multidisciplinary Center for Earthquake Engineering Research, which makes 

it possible for the first author to conduct the study in HKPU during the summer of 2001. Thanks are 

also extended to Larry Olson who provided the destructive testing data. 

REFERENCES 

Huang, N.E., S. Zheng, S.R. Long, M.C. Wu, H,H. Shin, Q. Zheng, N-C. Yen, C.C. Tung, and M.H. 

Liu (1998a). "The Empirical Mode Decomposition And Hilbert Spectrum For Nonlinear 

And Nonstationary Time Series Analysis." Proc. Roy. SocLond., A 454 903-995. 

Sanayei, M. and E.M. Santini (1998) "Dynamic bridge substructure evaluation and monitoring 

system," Report of subcontract to Tufts University from FEW A Grant to Olson Engineering.

1-1 

INDEX OF CONTRIBUTORS 

Volumes III 

Abe,M. 195,529 

Abrams, D. 139 

Ahmad, S. 471 

Azuhata,T. 405 

Beavers,! 139 

Caffrey,J. 247 

Caicedo, J.M. 529 

Carlson, ID. 205 

Chandler, A.M. 257 

Chang, P.C. 9 

Chang, T-S. 495 

Chau,K.T. 17 

Chen,CR 381 

Chen,J. 537 

Chen,Y.Q. 545 

Cho, D.Y. 309 

Cho,H. 397 

Cho, S.W. 207 

Choi,K.M. 223 

Chung, L.L. 389 

Clayton, E. 529 

Dai, J.W. 265 

Dargush, G.R 397 

Ding,X.L. 545 

Dohi,H. 489 

Dyke, SJ. 529 

Elnashai, A.S. 139,273 

Elvidge,C.D. 149 

Fuchs, A. 239 

Fujitani,H. 405 

Furukawa,A. 317 

Furata,H. 413 

Geng,S.W. 85 

Goel,S.C. 285,463 

Gong,X.N. 381 

Gordaninejad, F. 239 

Gulkan,P. 33 

Guo, A.X. 421 

Han,W. 293 

Hasegawa,K. 149 

Hata, K. 405 

Hayashi, H. 149, 157 

Hibino, T. 437 

Higashida, M. 149,157 

Higuchi, S. 489 

Hitchcock, G.H. 239 

Hiwatashi,T. 405 

Hobson,V.R. 149 

Hsu, T.T.C. 301 

Huang, Z.L. 453 

Hutchinson, T.C. 3,547 

Irtaza, H. 471 

Itam,A. 571 

Jabbari,F. 503 

Jeng, C.H. 301 

Jiang, J.R. 519 

Jing,L.P. 165 

Juhn, G.H. 309 

Jung, H J. 207,223 

Kalkan,E. 33 

Kaneto, T. 131 

Kang,Y. 123 

Kasai, K. 51 

Kim, B.W. 207 

Kim,I.H. 309 

Kim, J.K. 309 

Kirernidjian, A.S. 563 

Kiyono,J. 195,317 

Ko,J.M. 17,453 

Kohiyama,M. 149 

Koo, K.Y. 555 

Korol, G. 239 

Krawinkler, H. 173 

Kroehl,H.W. 149 

Kuchma, D.A. 323 

Kuester,F. 547 

Ladkany,S. 571 

Lam,E.S.S. 17 

Laogan,B» 273 

Law, K.H. 429,563 

Lee, CM. 17 

Lee,G.C. 3 

Lee, I.W. 207, 223 

Lee,S.S. 285 

Lee, V.W. 99 

Leung, Y.K. 123 

Li, H. 421 

Li,J. 537 

Li,Z.L. 545 

Li,Z.W. 545 

Liao,T. 115 

Liao, Z.P. 293 

Lin,G. 331 

Lin, S. 503 

Liu, G.X. 545 

Liu,J. 123 

Liu,J.P. 519 

Lynch, J.P. 429,563 

MacRosite,M. 611 

Maki, N. 149, 157 

Makris,N. 511 

Mao, Y.Q. 215 

Masri, S.F. 247 

Matsui, T. 437 

Mayne,P.W. 115 

Minowa, C. 405 

Mo, Y.L. 301 

Moehle,J.P. 67 

Morita, K. 405 

Nagarajaiah, S. 215 

Nagel, O. 445 

Narahashi,H. 181 

Ni,Y.Q. 17,453 

Nishimura, S. 183 

Nomura, Y. 413 

Okuta, K. 489 

Ozawa, T. 489 

Park, K.S. 223 

Parra-Montesinos, G. 463 

Peng,J.Y. 421 

Perng, S.F. 595 

Qamaruddin, M. 471

SEP r:] 

1-2 

Qiao, F. 437 

Radhaknshnan, R. 397 

Radulescu, C. 587 

Radulescu,D. 587 

Sahasrabudhe, S. 215 

Saiidi,M. 571 

Sakanoue,T. 579 

Santo, N. 481 

Sato,T. 579 

Savage, S. 463 

Sereci,A.M. 587 

Sheikh, M.N. 257 

Shiah,G.C. 595 

Shiozaki, Y. 405 

Singh, M.P. 495 

Song,G. 231 

Spencer, Jr. B F. 69 

SU.R.K.L. 257 

Sugiyama, E. 437 

Sunakoda, K. 405 

Sundararajan, A. 563 

Suzuki, Y. 421 

Takao,M. 131 

Tarn, CM. 123 

Tao,X.X. 85 

Tjhin,T.N. 323 

Tsujii, Y. 489 

Tu, W.G. 339, 369 

Wang, J.Y. 453 

Wang, M.L. 603 

Wang,X. 239 

Wang,X. 603 

Wang, Y.P. 389 

Waseem, S.M. 471 

Williams, G. 349 

Wolfe, R.W. 247 

Wong, Y.L. 17,123, 265 

Wu,L.Y. 389 

Wu,Z.Y. 165 

Xiao,S.Y. 331 

Xie,B. 231 

Xin, M. 239 

Xu,Y.L. 17,537,611 

Yang,J.N. 9,503 

Yang, Q. 571 

Yoshida,!. 131,579 

Yoshida, K. 481 

YU.H.Y. 187 

Yun, C.B. 555 

Zeng,X. 361 

Zhang,! 511 

Zhang, L.X. 519 

Zhang, M.Z. 265,519 

Zhang, R.R.C. 611 

Zhao,X.Q. 123 

Zheng, S.H. 123 

Zhu,P. 195 

Zou,Y.S. 339,369

Ths OOO F X ^s dje foe return 01 êne^/ai on t pê date 

showi un'ess previously recalled ^ nes may be 

PCI s red To iate re.um. 

DATE DUE

Earthquake Engineering Research - HKU Libraries - The University ...

You also want an ePaper? Increase the reach of your titles

Delete template?

Save as template?