Nuclear Engineering and Design 217 (2002) 179– 206
www.elsevier.com/locate/nucengdes
Technical note
Review of quantitative accuracy assessments with fast
Fourier transform based method (FFTBM)
Andrej Prošek a,*, Francesco D’Auria b, Borut Mavko a
a
‘Jožef Stefan’ Institute, Jamo6a 39, 1000 Ljubljana, Slo6enia
b
Uni6ersity of Pisa, Via Diotisal6i 2, 56 100 Pisa, Italy
Received 12 June 2001; received in revised form 10 January 2002; accepted 7 April 2002
Abstract
In the past 10 years various methodologies were proposed to evaluate the uncertainty of BE code predictions. One
common step to all methodologies is the use of experimental and plant data for the nodalization development and
qualification. When thermal– hydraulic computer codes are used for simulation the questions raised are: ‘How long
improvements should be added to the model, how much simplification can be introduced and how to conduct an
objective comparison?’ The proposed fast Fourier transform based method (FFTBM) assists in answering these
questions. The method is easy to understand, convenient to use, user independent and it clearly indicates when
simulation needs to be improved. The FFTBM shows the measurement– prediction discrepancies— accuracy quantification— in the frequency domain. The acceptability factor for code calculation was determined based on several
hundreds of code calculations. The FFTBM method has been applied to various international standard problems,
standard problem exercises and other experiment simulations that are presented in the paper. The result shows that
the quantitative comparison between thermal– hydraulic code results and experimental measurements with qualitative
evaluation may assist the decision whether or not the simulation needs to be improved. © 2002 Elsevier Science B.V.
All rights reserved.
1. Introduction
The assessment process of large thermal – hydraulic BE computer codes aims principally at
verifying their quality by comparing code predictions against experimental data gained mainly by
tests performed on scaled plant experimental facilities. The merit of these predictions, adopted for
* Corresponding author. Tel.: + 386-1-5885450; fax: + 3861-5612335
E-mail address: andrej.prosek@ijs.si (A. Prošek).
nuclear power plant safety analyses, depends on
many factors involving code features and user
experience.
Best estimate (BE) thermal – hydraulic system
codes predict reactor transient scenarios as realistically as possible by approximating the physical
behavior with some accuracy. Information on inaccuracy of predictions comes from the code assessment and validation process. Extensive
experimental programs have been conducted in
order to support the development and validation
activities of BE thermal – hydraulic codes. A
0029-5493/02/$ - see front matter © 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 0 2 9 - 5 4 9 3 ( 0 2 ) 0 0 1 5 2 - 8
180
A. Prošek et al. / Nuclear Engineering and Design 217 (2002) 179–206
methodology suitable to quantify the code accuracy (i.e. the code capability to correctly predict
the scenario observed during the test performed
on a scaled facility) has been developed (D’Auria
et al., 1994). It is an integral method using the fast
Fourier transform (FFT).
The fast Fourier transform based method
(FFTBM) is easy to understand, convenient to
use, user independent and it clearly indicates
when simulation needs to be improved. The
FFTBM shows the measurement–prediction discrepancies in the frequency domain.
The purpose of the paper is to present the
applications of the FFTBM to the calculation
analyses of International standard problems (ISP)
or standard problem exercise (SPE) organized by
CSNI or IAEA (i.e. ISP 21 (OECD-NEA, 1989),
ISP 22 (Ambrosini et al., 1991), ISP 27 (D’Auria
et al., 1994, Prošek and Mavko, 1997), ISP 33
(Purhonen et al., 1994), ISP 35 (D’Auria et al.,
1995), ISP 39 (D’Auria and Galassi, 1997), ISP 42
(Aksan et al., 2001) as well as Institut ‘Joz' ef
Stefan’ (IJS) calculations of IAEA-SPE-2
(Frogheri et al., 1995) and IAEA-SPE-4 (Prošek
et al., 1995), DCMN calculations of IAEA-SPE1– 4 (D’Auria et al., 1996), participants to IAEASPE-4 calculations (Mavko et al., 1997) and other
applications (Prošek and Mavko, 1995; Kljenak
and Prošek, 1996; Prošek et al., 1996; D’Auria
and Ingegneri, 1997; D’Auria et al., 1997b). Second aim was to show the maturity of the method
and its usefulness to the thermal– hydraulic code
analysts.
forth between these two representations by means
of the Fourier transform equations,
&
&
F0 ( f ) =
F(t)e2yift dt
−
F(t)=
F0 ( f )e − 2yift df
(1)
−
In the most common situations, function F(t) is
sampled at evenly spaced intervals in time. Suppose that we have N consecutive sampled values
Fk
F(tk ),
tk
k~,
k = 0, 1, 2, …, N−1
(2)
so that sampling interval is ~. The approximation
of the integral in Eq. (1) by discrete sum gives:
F0 ( fn )=
&
N−1
−
F(t)e2yifn t dt: % Fk e2yifn tk~
k=0
N−1
=~ % Fk e2yikn/N
(3)
k=0
where, fn n/N~. The final summation in Eq. (3)
is called the discrete Fourier transform of the N
points Fk. Let us denote it by F0 n :
N−1
F0 n = % Fk e2yikn/N
(4)
k=0
The relation Eq. (3) between the discrete
Fourier transform of a set of numbers and their
continuous Fourier transform when they are
viewed as samples of continuous function sampled
at an interval ~ can be written as
F0 ( fn ): ~F0 n
(5)
2. FFTBM overview
The formula for discrete inverse Fourier transform, which recovers the set of Fk ’s exactly from
F0 n ’s is:
2.1. Fourier transform of discretely sampled data
Fk =
A physical process can be described either in
the time domain, by values of F(t) as a function
of time t, or else in the frequency domain, where
the process is specified by giving its amplitude
F0 ( f ) (generally complex number indicating phase
also) as a function of frequency f, with − B
fB . For many purposes it is useful to think of
F(t) and F0 ( f ) as being two different representations of the same function. One goes back and
The discrete Fourier transform can be computed with an algorithm called the fast Fourier
transform, or FFT, which is algorithm that
rapidly computes the discrete Fourier transform.
To apply it, functions must be identified by a
number of values that is a power with base equal
to 2 and sampling theorem must be fulfilled. The
fulfillment of the sampling theorem is required to
avoid distortion of sampled signals due to aliasing
1 N−1
% F0 n e − 2yikn/N
N n=0
(6)
A. Prošek et al. / Nuclear Engineering and Design 217 (2002) 179–206
occurrence. The sampling theorem says: ‘‘a signal
that varies continuously with time is completely
determined by its values at an infinite sequence of
equally spaced times if the frequency of these
sampling times is greater than twice the highest
frequency component of the signal’’ (Lapendes,
1978). Thus, if the number of points defining the
function in the time domain is N =2m + 1, then
according to the sampling theorem the sampling
frequency is:
N 2m + 1
1
=fs = 2fmax = =
Td
Td
~
mental and error signal obtained by FFT at frequencies fn, where (n=0, 1, …, 2m) and m is the
exponent (m=8, 9, 10, 11). These spectra of
amplitudes together with frequencies are used for
calculation of average amplitude (AA) and
weighted frequency (WF) that characterize code
accuracy. For each variable the AA (variable
accuracy) is defined as the sum of error function
amplitudes normalized to the sum of experimental
signal amplitudes:
2m
% D0 F( fn )
(7)
n=0
AA =
where, Td is the transient time duration of the
sampled signal and fmax is the highest (maximum)
frequency component of the signal. The sampling
theorem does not hold beyond fmax. From the
relation in Eq. (7) is seen that the number of
points selection is strictly connected to sampling
frequency. The FFT algorithm determines the
number of points, equally spaced, which is a
power with base 2 (N range from 29 to 212).
Generally, an interpolation is necessary to satisfy
this requirement. The FFTBM application implies
analysis window (Td) selection, the number of
points (N) determination, the cut off frequency
( fcut) value determination, the fixed frequency
(minimum maximum frequency of analysis) value
selection and the weights selection. A cut off
frequency has been introduced to cut off spurious
contributions, generally negligible. The sampling
frequency (number of points) in the FFTBM interpolation algorithm is defined based on given
fixed frequency.
2.2. A6erage amplitude and weighted frequency
The FFTBM shows the measurement– prediction discrepancies in the frequency domain. For
calculation of these discrepancies the experimental
signal (Fexp(t)) and error function are needed. The
error function in the time domain DF(t) is
defined:
DF(t) =Fcal(t) −Fexp(t)
181
(8)
where, Fcal(t) is calculated signal. The code accuracy quantification for a individual calculated
variable is based on amplitudes of discrete experi-
(9)
2m
% F0 exp( fn )
n=0
and a WF is defined as the sum of frequencies
multiplied (weighted) by error function amplitudes, normalized to the sum of error function
amplitudes
2m
% D0 F( fn ) fn
WF =
n=0
m
2
.
(10)
% D0 F( fn )
n=0
The basic idea of FFTBM is to quantify the
discrepancy with single values, therefore, amplitude and frequency are averaged (see Eqs. (9) and
(10)). The purpose of the defined values is not to
replace the traditional qualitative analysis but to
additionally quantify the calculation.
The most significant information given by AA,
is the relative magnitude of the discrepancy coming from the comparison between the calculation
and the corresponding experimental variable time
history. When the calculated and the experimental
data are equal then the error function is zero (AA
is also equal to zero), characterizing perfect agreement. The WF factor characterizes the kind of
error, because its value emphasizes where the
error has more relevance either at low or high
frequencies. Depending on the transient, high frequency errors may be more acceptable than low
(in thermal– hydraulic transient, better accuracy is
generally represented by low AA values at high
WF values). Nevertheless, in the applications of
the method the need was felt for a better definitive
role of WF in the quantification of accuracy
(Ambrosini et al., 1990).
A. Prošek et al. / Nuclear Engineering and Design 217 (2002) 179–206
182
The meaning of AA and WF can best be illustrated by an example. The experimental signal
and calculated signal are given by:
Fexp(t)= {t
Fcal(t) =
if
0B t515
!
1.4t
4.5+ 0.5t
if
if
t 55
.
5 Bt5 15
(11)
Inserting Eq. (11) into Eq. (8) gives for the
error function:
DF(t) =
!
0.4t
4.5−0.5t
if
if
t 55
.
5Bt 515
(12)
The signals were sampled each second thus
obtaining 16 discrete points. These discrete signals
were input to the FFTBM algorithm. The maximum frequency selected was 0.5 Hz. According to
the Eq. (7) 32 points would be sufficient. However, in the FFTBM algorithm the time signal
discrete values are interpolated to get 29 – 212
points, depending on the input highest frequency
and the number of discrete points. Therefore, the
maximum frequency is equal to 17.066 Hz (N=
512 and Td =15 s). For the chosen cut frequency
0.5 Hz the amplitude spectra consists of eight
amplitudes (F0 ( fn ) ={7.500, 4.784, 2.392, 1.595,
1.196, 0.957, 0.798, 0.684} and D0 F( fn ) ={0.002,
1.858, 0.416, 0.319, 0.283, 0.172, 0.159, 0.150}).
The frequencies are fn = {0.000, 0.067, 0.133,
0.200, 0.267, 0.333, 0.400, 0.467}.
The AA calculated following Eq. (9) is:
AA=
0.002 + 1.858 + 0.416 + 0.319 +0.283 + 0.172 + 0.159 + 0.15
7.5 + 4.784+ 2.392 + 1.595 + 1.196 + 0.957+ 0.798 + 0.684
=
3.359
= 0.169
19.906
(13)
Similarly, following Eq. (10) and multiplying
D0 F( fn ) by fn gives the sum 0.51 and the value of
WF is then 0.152 as shown by equation:
0.510
WF =
=0.152.
(14)
3.359
A small value for WF means that the dis-
crepancy between the measured and the calculated
trends is more important at low frequencies.
When WF is large, the discrepancy comes from
various kinds of noise and consequently is less
important.
When looking at the amplitude spectra of the
difference signal we can observe that it is exponentially falling function of the frequency (see
Fig. 2). Not only spectra of ‘difference signals’
have been elaborated, but also spectra of various
measurements including pressure, differential
pressure, flow-rate, and temperature. All these
spectra show peaks in a frequency range well
below 1 Hz. It can be said that typically chosen
maximum frequency is 0.5 Hz (this gives sampling
interval 1 s). Due to typically falling amplitude
spectra most large amplitudes belongs to frequencies up to 0.1 Hz. This holds for transients lasting
a few hundreds and thousands of seconds (e.g.
small-break loss-of-coolant accidents (SB LOCAs)) with maximum frequency component below 10 Hz. When transients are shorter (e.g 10 s)
according to Eq. (7) and due to the fact that
maximum number of the discrete points is limited
to 4096 the spectra consist of higher frequency
components which are reciprocal to time interval
duration.
Typical values of WF observed in the SB
LOCA applications are from 0.01 to 0.2 when
transients last hundreds or thousands of seconds.
This can be explained by the fact that the number
of points in the interpolation algorithm was fixed
between 512 and 4096 points at the development
phase. For transients lasting 1000 s according to
Eq. (7), the maximum possible frequency is 2.048
Hz and for transient lasting 100 s is 20.48 Hz.
When transients are shorter even higher frequencies are involved in the spectra, which as a consequence increase the typical value of WF. This
additionally shows the expressed need for a better
definitive role of WF. Therefore, primarily AA is
used for accuracy quantification.
2.3. Accuracy of code calculation
The overall picture of the accuracy for the
given code calculation is obtained by defining
A. Prošek et al. / Nuclear Engineering and Design 217 (2002) 179–206
average performance indices, total weighted AA
(total accuracy):
Nvar
AAtot = % (AA)i (wf)i
(15)
i=1
and total WF
Nvar
WFtot = % (WF)i (wf)i
(16)
i=1
with
Nvar
% (wf)i = 1
(17)
i=1
where, Nvar is the number of the variables analyzed, and (AA)i, (WF)i and (wf)i are AA, WF and
weighting factors for i-th analyzed variable, respectively. Each (wf)i accounts for experimental
accuracy, safety relevance of particular variables
and its relevance with respect to pressure (Bovalini et al., 1992). Weighting factors are shown in
Table 1.
Each weighting factor (wf)i takes into account
(D’Auria et al., 1997a):
1. Experimental accuracy (wexp)i : experimental
trends of thermal–hydraulic variables are
characterized by uncertainty due to intrinsic
characteristics of instruments, the measurement method and different evaluation procedures used to compare experimental measures
and the code predictions.
2. Safety relevance (wsaf)i : higher importance is
attributed to the accuracy of those calculated
variables (such as pressure, peak clad temperaTable 1
Weighting factor components for the analyzed quantities (Bovalini et al., 1992)
Pressure drops
Mass inventories
Flowrates
Primary pressure
Secondary pressure
Fluid temperatures
Clad temperatures
Collapsed levels
Core power
wexp
wsaf
wnorm
0.7
0.8
0.5
1.0
1.0
0.8
0.9
0.8
0.8
0.7
0.9
0.8
1.0
0.6
0.8
1.0
0.9
0.8
0.5
0.9
0.5
1.0
1.1
2.4
1.2
0.6
0.5
183
ture, etc.) which are relevant for safety and
design.
3. Primary pressure normalization (wnorm)i : This
contribution is given by a factor, which normalizes the AA value calculated for the selected variables with respect to the AA value
calculated for the primary pressure. This factor has been introduced in order to consider
the physics relations existing between different
quantities (i.e. fluid temperature and pressure
in case of saturated blow-down must be characterized by the same order of error). The
measurement of the primary pressure can be
considered highly reliable. Weighting factors
have been introduced considering that the
quantities that are the object of the accuracy
evaluation are not independent from each
other. The interdependency among the quantities is complex and is fixed (on the code side)
by partial differential equations and by the
actual system status (on the experimental side).
It is impossible to characterize the function of
one quantity versus the others. This function
also depends upon the selected transient scenario. However, a limited number of transients
were selected consisting in the blow-down phenomenon (performed in the PIPER facility
available at University of Pisa) and in LOCA
experiments (performed in the LOBI facility
available at JRC-Ispra).
On this basis, the relation between the error in
predicting pressure and the error in predicting the
other quantities was identified. Clearly this process had at least two elements of arbitrariness: (a)
the choice of transients, (b) the combination of
the error to get the final weighting factors available in the Table 1.
Nevertheless, the consideration of such weighting factors ‘a posteriori’ (i.e. following the application of the method) produces overall FFTBM
results apparently more consistent than those
achieved without their use.
The weighting factor for the i-th variable, is
therefor defined as:
(w ) (w ) (w
)
(wf)i = N exp i saf i norm i
var
% (wexp)i (wsaf)i (wnorm)i
i=1
(18)
184
A. Prošek et al. / Nuclear Engineering and Design 217 (2002) 179–206
where, wexp is the contribution related to the
experimental accuracy, wsaf is the contribution,
which expresses the safety relevance, and wnorm
the
contribution
of
primary
pressure
normalization.
This introduces a degree of engineering judgment in the method development not in its application that has been fixed by a proper and unique
definition of the weighting factors. The weights
must remain unchanged during each comparison
between code results and experimental data concerning a same class of transient.
2.4. Methodology for quantifying code accuracy
Given qualified user and qualified nodalization
scheme, code assessment process involves three
steps: the first is selection of an experiment from
CSNI validation matrices (Aksan et al., 1987) (or
a plant transient), then qualitative assessment,
and finally the quantitative assessment.
The University of Pisa methodology for quantifying code accuracy is related to those suggested
by CSNI (Holmström, 1992) and US INEL
(Schultz, 1992). The steps are: subdivision of the
scenario into ‘phenomenological windows’; for
each phenomenological window methodology requires specification of key phenomena that are
distinctive for this class of transients (for example
break flow), identification of the relevant thermal – hydraulic aspects (RTA) which are peculiar
of each transient (these are events or phenomena
consequent to the physical process, for example
sub-cooled blow-down) and selection of variables
characterizing the RTA (for example average
break flow); qualitative assessment of obtained
results by visually comparing the experimental
and calculated variables trends.
Qualitative assessment is done by evaluating
and ranking the discrepancies between the measured and calculated variable trends. Assessment
results can be subjectively described as:
Excellent: code predicts the variable qualitatively and quantitatively. Calculation falls
within experimental data uncertainty band. Calculation is qualitatively and quantitatively
correct.
Reasonable: code predicts the variable qualitatively but not quantitatively. Calculation is not
within experimental data uncertainty band but
shows correct behavior and trends.
Minimal: code does not predict the variable,
but the reason for this is understood and predictable. The calculation does not lie within
experimental data uncertainty band and at
times does not even show correct trends and
behavior.
Unqualified: code does not predict the variable,
and the reason is not understood. Calculation
does not lie within experimental data uncertainty band and at times does not even show
correct trends and behavior.
The qualitative assessment gives the first indications about the calculated predictions. The qualitative assessment phase is a necessary prerequisite
for a subsequent quantitative phase. It is meaningless to perform this last phase trough the FFT
based method if any RTA is not predicted.
The quantitative assessment can be managed by
applying method based on the FFT. The most
suitable factor for the definition of an acceptability criterion is the total AAtot. With reference to
the accuracy of a given calculation, we can define
the following acceptability criterion:
AAtot BK
(19)
where, K is acceptability factor valid for the whole
transient.
The (objective) basis for quantifying the acceptability limits is constituted by the state of the art
capabilities of current system BE codes in thermal – hydraulics. Several tens of complex transients including small break LOCA, large break
LOCA, long lasting transients, have been analyzed by the method based on the FFT using all
the codes available to the international community (RELAP, TRAC, ATHLET, CATHARE,
SMABRE); several hundreds of time quantities
have been dealt with in such a way. In a number
of cases detailed and independent code accuracy
evaluations were available (this is the case of
International standard problems (ISP) 26 and 27)
leading to the conclusion that, an excellent calculation could be characterized by K= 0.3. In the
same way, a poor calculation (i.e. a calculation
A. Prošek et al. / Nuclear Engineering and Design 217 (2002) 179–206
predicting all the relevant experiment phenomena
with poor accuracy) could be characterized by
K =0.5. It can be noted that:
AAtot = 0.3 characterize very good code
predictions,
0.3BAAtot 50.5 characterize good code
predictions,
0.5B AAtot 50.7 characterize poor code
predictions,
AAtot \ 0.7 characterize very poor code
predictions.
The difficulty in getting the value AAtot =0.3
(e.g. very good knowledge of boundary conditions
and a very detailed nodalization are necessary)
and the demonstration that K =0.4 implies an
error level acceptable to experienced code users,
brought to the decision of assuming this last value
as ‘acceptability limit’ for a calculation. The same
criterion can be used to evaluate the code capability in the single variable prediction; clearly, in this
case the AA factor is the one evaluated for the
addressed variable. In particular, acceptability
factor K =0.1 has been fixed for the primary
pressure, because of its importance.
It should be noted that the FFT based method
does not allow the identification of the error
origin (i.e. user effect, wrong initial condition,
nodalization model deficiency, etc.) or to take into
account directly time shift of certain phenomena
in the quantitative analysis. Time shift is better
characterized through the necessary qualitative
assessment that must be associated with quantitative assessment.
3. Examples of FFTBM
Two FFTBM application examples are presented to show how the method works. The first
example shows, how to calculate AA and WF for
a single variable prediction (secondary pressure of
BETHSY 6.2TC). In the second example, it will
be shown how to assess the total code accuracy
from the average amplitudes of the selected variables to be considered in BETHSY 6.2TC
calculation.
185
Fig. 1. Experimental, calculated and error time trend for
secondary pressure.
3.1. Example of calculating AA and WF
In this example the FFTBM was applied to the
BETHSY 6.2TC calculation of secondary pressure
(Hrvatin et al., 2000). The BETHSY 6.2 TC test
simulates a scaled 15.24 cm (6 inch) cold leg break
transient with no high and low head safety injection and with only accumulator in the intact loop
available. During transient after scram and the
condenser isolation the pressure increases. The
pressure increase is limited by atmospheric relief
valves at 7.2 MPa. In the RELAP5 model the
relief valve was modeled with time dependent
junction which discharges mass without any delay.
Later the pressure in the calculation decreases due
to cool-down of the primary system while feedwater was isolated. In the experiment the pressure
decrease is much faster and it seems that there was
additional cooling mechanism not modeled with
RELAP5 causing differences in the results.
When calculating the AA and WF for secondary pressure shown in Fig. 1 the result is
amplitude spectra shown in Fig. 2 for the lowest
frequencies. Now the question is how to select
maximum and cut frequency. Cut frequency is the
frequency, beyond which the amplitudes are not
taken into account for calculating AA and WF.
We can see that AA and WF are characterized by
large amplitudes at low frequencies and that
higher frequency components have smaller impact
on the AA. The impact of the cut frequency
(generally negligible) is shown in Fig. 3. For the
selected variable the highest AA is obtained tak-
186
A. Prošek et al. / Nuclear Engineering and Design 217 (2002) 179–206
Fig. 2. Amplitude spectra for secondary pressure (p47) of
BETHSY 6.2TC.
ing into account all frequencies. When only half
frequencies are taken into account, the difference
in results is about 7%. For some variables the
highest AA are reached at low cut frequencies and
the AA decreases with increasing cut frequency
(D’Auria et al., 2000). Therefore, care must be
taken into account when selecting cut frequency.
On the other hand, the WF monotonically increases with cut frequency. An example is shown
in Fig. 3. Therefore, higher cut frequency gives
higher WF values and the opposite. Please note
that in the analysis the same cut frequency is used
for all calculations compared.
3.2. Example of calculating total accuracy
Generally, the applications of the FFTBM
methodology are performed for 20–25 time trends.
Fig. 4. Visual comparison for BETHSY 6.2TC experiment.
Fig. 3. Impact of cut frequency on AA and WF.
In the present case 23 time trends were selected for
accuracy quantification of BETHSY 6.2TC experiment (Hrvatin et al., 2000). Some trends are
shown in Fig. 4 together with calculated average
amplitudes to show relation between discrepancy
and quantitative measure. All selected variables
and calculated AA and WF are shown in Table 2.
The total AA and WF were calculated using Eqs.
(15) – (17) and weighting factors from Table 1. The
total AA for four time intervals range from 0.09 to
0.28 all meeting acceptability criterion.
A. Prošek et al. / Nuclear Engineering and Design 217 (2002) 179–206
187
Table 2
Example of calculating total accuracy for BETHSY 6.2TC
Number
Variable description
Time interval of analysis
0–140 s
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
Pressurizer pressure
SG pressure
Accumulator pressure
Core inlet fluid
Core outlet fluid
Upper head fluid
Integral break flow
SG DC bottom flow
Break flow rate
ECCS integral flow
Heater rod temperature (bottom)
Heater rod temperature (middle)
Heater rod temperature (bottom)
Primary side total mass
Core level
SG DC level
DP inlet-outlet SG
Core power
DP loop seal broken loop
(downflow)
DP loop seal broken loop
(upflow)
DP in pressurizer
DP SG inlet
DP across DC-UH
Total
0–350 s
0–980 s
0–2687 s
AA
WF
AA
WF
AA
WF
AA
WF
0.08
0.07
0.01
0.01
0.01
0.02
0.03
0.01
0.35
0.00
0.04
0.14
0.02
0.04
0.16
0.63
0.70
0.06
0.27
0.06
0.04
0.07
0.04
0.05
0.09
0.08
0.10
0.04
0.00
0.05
0.10
0.05
0.08
0.07
0.07
0.05
0.05
0.04
0.06
0.08
0.03
0.01
0.01
0.02
0.06
0.01
0.39
2.02
0.04
0.18
0.30
0.11
0.50
0.53
0.81
0.04
0.25
0.05
0.06
0.13
0.04
0.06
0.08
0.07
0.08
0.06
0.19
0.06
0.09
0.09
0.05
0.06
0.07
0.04
0.03
0.03
0.06
0.14
0.06
0.04
0.02
0.28
0.04
0.03
0.45
0.12
0.04
0.14
0.20
0.15
0.58
0.56
0.90
0.04
0.25
0.05
0.06
0.03
0.02
0.04
0.06
0.05
0.09
0.06
0.06
0.06
0.05
0.02
0.04
0.05
0.08
0.05
0.03
0.04
0.06
0.27
0.09
0.11
0.40
0.61
0.04
0.09
0.52
0.14
0.08
0.50
0.24
0.13
0.46
0.54
0.89
0.08
0.27
0.05
0.05
0.04
0.05
0.05
0.05
0.04
0.08
0.05
0.04
0.05
0.05
0.03
0.03
0.05
0.06
0.04
0.06
0.03
0.34
0.11
0.35
0.04
0.35
0.04
0.37
0.04
0.16
1.03
0.41
0.09
0.05
0.04
0.05
0.07
0.17
0.83
0.73
0.20
0.05
0.08
0.06
0.07
0.18
1.20
0.74
0.17
0.05
0.06
0.05
0.05
0.21
0.97
0.88
0.28
0.06
0.06
0.05
0.05
In Fig. 5 is shown influence of selecting cut
frequency on total accuracy when fixed frequency
0.5 Hz was selected. When cut frequency is half or
double of fixed frequency the AA changes by less
than 3%. In this case the total AA increases (the
total accuracy decreases) with cut frequency
which means that taking into account all frequencies gives the worst result. However, it should be
noted that single variables accuracy may be increasing and decreasing with increasing cut frequency, therefore, the variation of total accuracy
is smaller than single variable accuracy variations.
spread from SB LOCA to other accidents (including severe accidents) and that some level of maturity was reached after one decade of applications.
4. Review of FFTBM applications
The review of FFTBM applications shows the
historical progress of the method, how the use
Fig. 5. Impact of cut frequency on total accuracy for BETHSY
6.2TC experiment.
188
A. Prošek et al. / Nuclear Engineering and Design 217 (2002) 179–206
ables selected FFTBM shows two different levels
of accuracy, which was confirmed by other methods also. Direct comparison with WF results for
different calculations has meaning only for similar
average amplitudes. In that case, the AA with
larger WF is more acceptable result. Therefore, in
our case better value of WF for TRAC/PF1 than
for R4/M6 for residual mass do not change conclusions based on AA. Ambrosini et al. (1990) felt
the need for a better role of WF in the quantification of accuracy.
Fig. 6. Results of the FFTBM application to ISP-18.
4.2. Application to ISP 21
4.1. Partial application to ISP 18
Due to its promising characteristics, the
FFTBM method was then used to quantify accuracy of blind pre-test calculations of ISP 21 with
the title ‘PIPER-ONE Test PO-SB-7 on SB
LOCA in BWR- Recirculation line’. The main
objectives of ISP 21 were simulation of the
PIPER-ONE facility under the conditions of a
SB-LOCA in the down-comer region, confirmation of code capabilities describing a test in a
geometrically simplified apparatus and the test
PO-SB-7 to serve as counterpart test to BWR-related tests in other facilities.
In the analysis (Ambrosini et al., 1990), ten
variables were considered for seven calculations
and the transient duration was 350 s. At that time
no procedure was available for weighting factors,
therefore, it was decided to put the same weight to
each variable. The results for single variable average amplitudes and total accuracy are shown in
Table 4. The quantitative analysis was found consistent with conclusions reached in the Compari-
After testing on sample curves, the FFTBM
was first time applied to pre-test blind calculations
(Ambrosini et al., 1990) of ISP 18 with the title
‘LOBI-MOD 2 Small Break LOCA Experiment
A2-81’. The main objective of ISP 18 was simulation of a 1% SB LOCA in the non-nuclear LOBI
test facility at Ispra with high pressure injection
system (HPIS) partially in operation.
For the quantitative analysis two variables were
selected: intact loop hot leg density and primary
system residual mass. The results obtained are
presented in Fig. 6 and were compared with other
methods also (Ambrosini et al., 1990) whose results are presented in Table 3. It can be noted
that, in this case that quantitative information
provided is consistent for the various methods,
showing a better behavior of the first generation
code (RELAP4/MOD6) with respect to more advanced. When comparing AA for the two vari-
Table 3
Results obtained by accuracy evaluation methods for selected variables of ISP 18
Variable
Intact loop hot leg density
Primary side residual mass
Method
Dm
IA
MFE
Dm
IA
MFE
Code used
R4/M6
R5/M1-C
R5/M2
TRAC/PF1
0.29
0.65
4.07
0.14
0.2
0.37
0.66
0.89
18
0.12
0.18
1.17
0.65
0.81
16.67
0.27
0.43
4.57
0.66
0.89
28.76
0.22
0.43
4.43
Table 4
Results obtained by FFTBM for double-blind ISP 21 calculations
Variable
TF03: lower plenum
temperature
TF17: steam dome
temperature
CL70: collapsed liquid level
in the down-comer
MF43: break mass flow rate
MF48: LPCS mass flow rate
TR50: rod temperature
(level B)
TR52: rod temperature
(level D)
TR55: rod temperature
(level G)
RM65: fluid mass in the loop
Average weight
Standard weightb
a
b
ABB
Ansaldo
ENEL
JAERI
VTT1
VTT2
DCMNa
GOBLIN-EM
RELAP5/
MOD2
RELAP4/
MOD6
THYDE-B1/
MOD2
SAMBRE (1)
SAMBRE (2)
RELAP5/
MOD2
AA
WF
AA
0.084
0.063
0.089
0.109
0.065
0.382
0.196
0.071
0.480
0.149
0.053
0.145
0.266
0.081
0.314
0.266
0.081
0.315
0.038
0.094
0.077
WF
AA
0.135
0.539
0.149
0.626
0.130
0.619
0.140
0.444
0.137
0.149
0.137
0.150
0.159
0.392
WF
AA
0.155
0.326
0.139
0.108
0.148
1.102
0.150
0.194
0.040
0.253
0.042
0.253
0.156
0.077
WF
AA
WF
AA
WF
AA
0.086
0.736
0.099
0.385
0.123
0.072
0.131
0.588
0.115
0.351
0.133
0.085
0.140
0.772
0.124
0.274
0.118
0.332
0.084
0.667
0.109
0.432
0.065
0.102
0.063
0.790
0.116
0.876
0.122
0.287
0.063
0.790
0.116
0.880
0.122
0.288
0.171
0.534
0.147
0.106
0.140
0.066
WF
AA
0.161
0.152
0.061
0.133
0.094
0.625
0.108
0.167
0.110
1.575
0.110
1.473
0.069
0.117
WF
AA
0.133
0.394
0.126
0.455
0.060
0.740
0.087
0.485
0.111
1.106
0.107
1.038
0.113
0.314
WF
AA
WF
AAtot
WFtot
AAtot
WFtot
0.071
0.113
0.128
0.289
0.115
0.260
0.127
0.087
0.282
0.157
0.312
0.116
0.338
0.120
0.062
0.600
0.157
0.574
0.111
0.578
0.111
0.077
0.285
0.150
0.307
0.102
0.279
0.115
0.095
0.646
0.162
0.626
0.104
0.603
0.101
0.087
0.647
0.162
0.610
0.103
0.580
0.100
0.077
0.029
0.102
0.175
0.123
0.180
0.122
A. Prošek et al. / Nuclear Engineering and Design 217 (2002) 179–206
PA01: lower plenum pressure
Participant and code used
Post– test calculation.
Analysis performed in year 2001.
189
190
A. Prošek et al. / Nuclear Engineering and Design 217 (2002) 179–206
son report of ISP 21 (D’Auria et al., 1989) on the
basis of qualitative observation of the submitted
curves. The considerable improvements achievable
in accuracy by the tuning of variables affecting
code behavior during post-test analysis was further interesting outcome. In this analysis the RELAP4/MOD6 behaves much worse than more
advanced RELAP5/MOD2 code. The analysis
also revealed the need for development of an
objective procedure in order to identify the experimental variables to be considered in the quantification of accuracy and to define the weighting
factors for the calculation of AAtot and WFtot.
Table 4 shows the results of using standard
weights as defined in Table 1. The difference in
the results using average weights and standard
weights is less than 13%.
4.3. Application to ISP 22
Next FFTBM application was to ISP 22 entitled ‘Loss of Feedwater Transient in Italian PWR
(SPES Test SP-FW-02)’. The main objectives were
simulation of the integral behavior of the SPES
facility under the conditions of complete loss of
feedwater to secondary steam separators, core dry
out due to lack of heat sinks and opening of
pressurizer relief valve and primary side refilling
by actuation of emergency feedwater and pressurizer drainage to the primary coolant system. For
ISP 22 ‘blind’ pre-test predictions and extensive
‘open’ post-test analyses were performed.
The FFTBM analysis was performed in two
steps (Ambrosini et al., 1991). The first one was, a
qualitative approach to get a preliminary evaluation of the results obtained in each code application. By the visual observation of the curves
submitted by participants to pre-test and post-test
analyses, the method was aimed to state how
much each application can be considered representative of the phenomena actually occurring
during the experiment. In the second step, the
technique previously applied to ISP 18 and ISP 21
was used (see Sections 4.1 and 4.2). This approach
was considered as a pilot study towards the definition of an objective procedure for the quantification of code accuracy. For the accuracy
quantification 14 variables were considered. The
Fig. 7. Results of the FFTBM application to pre-test ISP 22
calculations as a function of time interval.
choice of these variables was made to identify the
most representative ones for the characterization
of transient evolution, avoiding any possible redundancy. As in application to ISP 21 equal
values were adopted for the weighting factors.
For pre-test applications, 14 code calculations
were analyzed considering three different time
intervals, as the submitted curves had different
time duration’s (from 0 to 1000 s, from 0 to 4000
s and from 0 to 6000 s). The results for the
selected time intervals are shown in Fig. 7. The
participants used the following codes: Energoproekt (BULG) used RELAP5/MOD2, CEA
(CEAF) used CATHARE 2 code, Institut Jozef
Stefan (IJSJ) used RELAP5/MOD2 code, JAERI
(JAER) TRAC-PF1 code, Studsvik Energieteknic
(STUD) used TRAC-PF1/MOD1 code, University of Pisa (UPIS) used RELAP5/MOD2 code
and University of Stratchlyde (USCL) used
TRAC code.
From Fig. 7, it can be drawn that an analysis of
AA as a function of time is necessary to evaluate
accuracy of calculations and their acceptability
during the various phases of the transient. The
reason for greater discrepancies in the second time
interval is due to too early emergency feedwater
actuation. In the last time interval these discrepancies are averaged over a larger time
interval.
A. Prošek et al. / Nuclear Engineering and Design 217 (2002) 179–206
In the post-test calculations four organizations
participated: CEA using CATHARE 2 code
(CEA1 and CEA2), University of Pisa (UPIS)
with RELAP5/MOD2 code, University of Dresden (UDRE) with RELAP4/MOD6 and BHA
BHA Atomic Research Centre from India (INDI)
with RELAP4/MOD6. The obtained results are
presented in Fig. 8. In addition to the analysis
with average weighting factors analysis with standard weighting factors was performed in 2001.
General improvement with respect to pre-test results was obtained for all organizations except
CEA (partly due to error compensation effect
present in the pre-test calculation). The evaluation
of post-test results also confirmed that RELAP4
code (and similar ones) were not able to deal with
this kind of transient even if the experimental data
were available.
Reanalysis with currently used standard weighting factors showed that the AAtot are lower (results are better) than in the analysis using average
weighting factors. Also it can be noted that for
this limited set of variables some calculations fulfil
the FFTBM criteria set for total accuracy (AAtot
below 0.4). The accuracy of INDI calculation was
assessed with only 11 out of 14 variables required.
Reanalysis of other post-test calculations with the
same eleven variables as for the INDI calculation
produces better AAtot values (on the average 0.06smaller values).
Fig. 8. Results of the FFTBM application to post-test ISP 22
calculations in the time interval (0 – 6000 s).
191
4.4. Application to ISP 27
Application of FFTBM to the ISP 27 entitled
‘BETHSY Experiment 9.1B; 2ª Cold Leg Break
without HPSI and with Delayed Ultimate Procedure’ shows the maturity of the method. The
maturity was showed in that the method was
sensible in highlighting the differences between
pre- and post-test calculations for the same user,
normally originated by an ad-hoc code tuning
operated in post-test analysis and by the code use
at the international level. In this Italian–French
application (D’Auria et al., 1994) the full FFTBM
method was applied to pre- and post-test calculations as described in the Bovalini et al. (1992).
The full FFTBM method include 20–25 variables
selected representing relevant thermal– hydraulic
aspects and standard weighting factors as shown
in Table 1.
The main objectives of the ISP 27 were simulation of integral plant behavior under SB LOCA
conditions, loop seal clearing and heat transfer
during boil-off period, behavior of an uncovered
core under reflux-condenser mode of steam generators and primary side refilling by low pressure
injection system (LPIS). The blind predictions
were requested, however also ‘open’ post-test
analyses were processed. In total more than 40
calculations (28 blind, 18 post-test and two open)
from 19 countries were submitted and nine different system codes were used. Among the 90 variables selected for the comparison with the
calculated results, 25 variables have been selected
for a quantitative evaluation of ISP 27 pre-test
results. Concerning the post-test submissions, the
majority of participants provided only a limited
set of variables, considered as significant for the
comparison.
Taking into account the course of events during
the transient and very different time duration of
submitted calculations, to compare each other,
the results of a considerable number of participants on the basis of common time windows,
three analyses have been carried out: from 0 to
3000 s, from 0 to 6000 s and from 0 to 8000 s. The
results are shown in Table 5. For each calculation,
the participant, country, code and number of
variables are specified, and the results for 1/WFtot
and AAtot for the three selected time intervals.
192
A. Prošek et al. / Nuclear Engineering and Design 217 (2002) 179–206
Table 5
Results obtained by FFTBM for pre-test ISP 27 calculations
ID
Participant/country/code
NV
Time interval of analysis
0-3000 s
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
ZFK Dresden, Germany, R5/M2
KINS, Korea, R5/M3
LTKK, Finland, CATH. 2 V1.2(31.)
Kurchatov Institute, Russia, R5/M3
CIAE Beijing, China, R5/M2
JINS, Japan, R5/M2
JAERI, Japan, CATH. 2 V1.2(31.)
Studvisk Nuclear, Sweden, R5/M3
NNC, England, NOTRUMP
Kurchatov Institute, Russia, R5/M2
JAERI, Japan, CATH. 2 V1.2(21.)
Paul Scherrer, Switzerland, R5/M2
GRS, Germany, ATHLET M1.E
Tractabel, Belgium, R5/M2
UKAEA, England, R5/M2
VTT, Finland, R5/M3
Texas AMU, USA, R5/M3
UKAEA, England, R5/M3
KAERI, Korea, CATH. 2 V1.2(21.)
IPEN/CNEN, Brazil, TRAC/PF1
DCMN/ENEA, Italy–Croatia, R5/M2
AEKI, Hungary, R5/M2
Jožef Stefan Institute, Slovenia, R5/M3
Jožef Stefan Institute, Slovenia, R5/M2
NPPRI, Slovakia, R5/M2
25
25
25
25
25
24
22
25
24
25
22
25
25
25
25
25
25
25
23
22
22
25
25
25
22
0-6000 s
0-8000 s
1/WFtot
AAtot
1/WFtot
AAtot
1/WFtot
AAtot
9.2
10.0
9.4
9.7
8.8
8.6
8.5
8.4
7.3
9.4
8.5
10.0
8.4
8.8
8.6
9.7
9.6
8.6
8.1
8.2
8.0
10.0
8.4
8.4
9.1
0.30
0.68
0.62
0.40
0.31
0.40
0.47
0.25
0.51
0.46
0.47
0.50
0.39
0.41
0.27
0.38
0.39
0.35
0.49
0.35
0.36
0.52
0.38
0.39
0.40
*
*
*
*
*
*
*
10.5
12.3
14.9
11.1
14.3
10.4
14.0
11.5
11.5
12.0
10.6
11.5
12.0
12.6
15.8
14.8
11.9
10.6
*
*
*
*
*
*
*
0.26
0.51
0.60
0.33
0.39
0.36
0.44
0.29
0.33
0.49
0.30
0.41
0.35
0.36
0.55
0.40
0.55
0.56
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
10.9
14.2
10.8
11.6
10.8
12.5
14.1
13.8
14.9
9.9
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
0.39
0.44
0.30
0.42
0.40
0.38
0.55
0.39
0.44
0.64
*Anticipated prediction of the end of the transient.
Note: CATH., R5/M2, R5/M3 correspond to CATHARE, RELAP5/MOD2, RELAP5/MOD3, respectively.
In the post-test analysis for only nine calculations all variables were available therefore the
analysis with six or seven variables was performed. Table 6 shows that, the results of quantitative analysis with 7(6) variables are different
from the analysis with 25(24) variables. The qualitative analysis showed that seven variables are
not enough to characterize completely the transient, therefore, the overall accuracy for the addressed calculations.
Later another post-test study was performed to
compare ISP 27 results obtained by different
nodalizations for RELAP5/MOD3.1 code (Prošek
and Mavko, 1997). The base and detailed model
consisted of 196/207/191/754 and 398/408/396/
1554 (volumes, junctions, heat structures and
mesh points), respectively. The same time windows and variables were used as described above
and it was shown that slightly higher accuracy
was only obtained for the first windows with the
same total accuracy equal to 0.39 in time interval
from 0 to 8000 s. Nevertheless, the primary pressure accuracy was fulfilled for the detailed model
only.
4.5. Application to ISP 33
First FFTBM application (Purhonen et al.,
1994), facilities simulating VVER was to ISP 33
entitled ‘PACTEL Natural Circulation Stepwise
Coolant Inventory Reduction Experiment’. The
main objectives of ISP 33 were to study natural
A. Prošek et al. / Nuclear Engineering and Design 217 (2002) 179–206
convection circulation in a VVER plant, series of
quasi steady natural circulation periods, single
phase natural convection and two-phase natural
convection with continuous liquid flow and convection flow under reflux-boiler mode conditions.
The ‘double-blind’ pre-test analyses and ‘open’
post-test analyses were performed.
Among 21 submitted results, 19 blind calculations have been analyzed with the FFTBM using
standard weighting factors (Purhonen et al.,
1994). For the analysis five time windows were
selected. In selecting variables special attention
193
was paid to their importance with respect to
nuclear safety. However, due to unavailability of
variables, only nine variables were selected for the
analysis. In the post-test calculations all 16 calculations were analyzed. The comparison of blind
and post-test calculations was done for ten calculations which are shown in Table 7. It is obvious
that, the post-test results have a better accuracy
then the pre-test calculations. The obtained results
of the accuracy of calculations gives an indication
of the capabilities of the codes and code users to
predict behavior of this kind of transient.
Table 6
Results obtained by FFTBM for post-test ISP 27 calculations
ID
Participant/country/code
NV
Time interval of analysis
0–3000 s
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
2
4
5
6
8
9
10
18
20
JAERI, Japan, CATH. 2 V1.2(31.)
NNC, England, NOTRUMP
JAERI, Japan, CATH. 2 V1.2(31.)
CIAE (A), China, R5/M2
CIAE (B), China, R5/M2
KINS, Korea, R5/M3
JINS (B), Japan, R5/M2
Kurchatov Institute, Russia, R5/M3
KAERI, Korea, CATH. 2 V1.2(21.)
Paul Scherrer, Switzerland, R5/M2
Texas AMU, USA, R5/M3
VTT, Finland, R5/M3
Paul Scherrer (A), Switzerland, R5/M2
Paul Scherrer (B), Switzerland, R5/M2
ZFK DRESDEN, Germany, R5/M2
Jožef Stefan Institute, Slovenia, R5/M3
Jožef Stefan Institute, Slovenia, R5/M2
INEL (open), USA, R5/M3
JINS (A), Japan, R5/M2
Kurchatov Institute, Russia, R5/M2
NNC, England, NOTRUMP
CIAE (A), China, R5/M2
CIAE (B), China, R5/M2
KINS, Korea, R5/M3
Kurchatov Institute, Russia, R5/M3
KAERI, Korea, CATH. 2 V1.2(21.)
Paul Scherrer, Switzerland, R5/M2
INEL (open), USA, R5/M3
Kurchatov Institute, Russia, R5/M2
6
7
6
7
7
7
7
6
6
7
7
7
7
7
5
7
7
6
7
6
24
25
25
25
25
22
25
25
25
0–6000 s
0–9000 s
1/WFtot
AAtot
1/WFtot
AAtot
1/WFtot
AAtot
9.2
8.0
7.9
8.1
7.9
8.3
7.9
10.9
8.3
8.0
9.1
8.1
7.9
7.8
7.8
9.2
9.7
8.4
7.6
11.4
7.4
9.5
9.1
9.5
9.4
9.5
8.8
8.4
8.8
0.49
0.27
0.31
0.29
0.28
0.39
0.38
0.37
0.46
0.51
0.37
0.42
0.51
0.50
0.33
0.28
0.31
0.41
0.42
0.26
0.42
0.36
0.41
0.45
0.38
0.33
0.27
0.30
0.36
*
*
*
*
*
*
11.4
14.2
10.7
9.3
9.4
10.2
9.2
8.8
9.3
9.8
10.7
10.9
11.7
11.2
*
*
*
*
14.2
11.5
11.3
12.8
11.6
*
*
*
*
*
*
0.32
0.37
0.54
0.57
0.45
0.46
0.57
0.57
0.38
0.30
0.32
0.43
0.32
0.33
*
*
*
*
0.43
0.34
0.29
0.33
0.37
*
*
*
*
*
*
*
*
*
*
*
9.7
8.9
8.5
9.3
10.0
10.3
10.7
11.5
10.7
*
*
*
*
*
*
*
12.0
11.5
*
*
*
*
*
*
*
*
*
*
*
0.52
0.65
0.64
0.56
0.34
0.41
0.52
0.35
0.51
*
*
*
*
*
*
*
0.34
0.39
*Anticipated prediction of the end of the transient.
Note: CATH., R5/M2, R5/M3 correspond to CATHARE, RELAP5/MOD2, RELAP5/MOD3, respectively.
194
A. Prošek et al. / Nuclear Engineering and Design 217 (2002) 179–206
Table 7
Comparison of pre- and post-test calculations for ISP 33 in time interval (0–6600 s)
ID
1
2
3
4
5
6
7
8
9
10
a
b
Participant/country/code
Geschellschaft für Anlagen- und Reaktor Sicherheit (GRS) mbH, Germany,
ATHLET Mod1 E
Hochschule für Tecnik, Wirtschaft und Socialwessen Zittau, Germany, ATHLET
Mod1 E
University of Pisa, DCMN, Italy, CATHARE 2 V1.3Ea
Turkish Atomic Energy Authority, Turkey, RELAP5/MOD3 ver. 5m5
Russian Research Center Kurchatov Institute, Russia, RELAP5/MOD3 ver. 5m5
‘Jožef Stefan’ Institute, Slovenia, RELAP5/MOD3 ver. 5m5
Nuclear Power Plant Research Institute, Slovakia, RELAP5/MOD2/RMA
Nuclear Research Institute, Czech Republic, RELAP5/MOD3 ver. 5m5b
Russian Research Center Kurchatov Institute, Russia, SCADAP/RELAP5/MOD2
Research Center Rossendorf Inc., Germany, ATHLET Mod1 E
Pre-test
Post-test
1/WFtot
AAtot 1/WFtot
AAtot
92.9
0.21
80.8
0.17
94.4
0.22
80.1
0.19
83.0
121.1
102.1
97.4
104.1
109.0
104.1
81.1
0.29
0.24
0.28
0.19
0.24
0.21
0.30
0.27
103.0
106.1
73.5
101.5
89.9
75.9
85.7
80.1
0.24
0.19
0.38
0.22
0.21
0.15
0.20
0.26
In pre-test CATHARE 2 V1.2E.
In pre-test RELAP5/MOD2.5/SRL.
4.6. Application to ISP 35
First application of FFTBM to containment
code calculations was to ISP 35 entitled ‘NUPEC
Hydrogen Mixing and Distribution Test (Test
M-7-1)’. The main objectives of ISP 35 were to
study helium (in lieu of hydrogen) distribution
phenomena in a model containment, with particular objectives associated to natural convection
effects, effects of spray water addition in inner
containment regions and validation of containment codes. The ‘blind’ post-test predictions were
requested and supplementary ‘open’ post-test calculations were possible. The results calculated by
ten pre-test and four post-test participants are
summarized here (for details refer to D’Auria et
al., 1995). The analysis was performed for one
phenomenological window: from 200 to 1790 s.
For the analysis 19 variables was selected: dome
pressure, temperature in nine different nodes and
He concentration in nine different nodes. For the
present first application new weighting factors
were arbitrary calculated (please note that this is
deviation from the original method): one for pressure and as the ratio between node volumes and
total volume for temperatures and He concentra-
tions. Namely, when the FFTBM started to
spread around the world researchers in the severe
accident area asked for the possibility to use the
method. To use the method the need was definition of weighting factors and acceptability limits
for total accuracy. This is responsibility of the
containment analysts. Nevertheless, for comparing single variables with experimental trends the
FFTBM can be used as it is.
The results of application to ISP 35 are shown
in Table 8. The definition of weighting factors was
just an example of application of the methodology, not accounting for other influences than
dome related quantities. With reference to available post-test calculations, the FFTBM analysis,
performed in the same window, confirms the better accuracy characterizing the post-test calculations, as a result of improved modeling of
boundary conditions, changes in the nodalization
and sensitivity analyses. Concerning the global
code calculation accuracy, further efforts might be
necessary to refine the values of the weighting
factor components, to be utilized in the further
containment analyses. While comparing the results obtained by CONTAIN code, user effect in
code utilization is identified.
A. Prošek et al. / Nuclear Engineering and Design 217 (2002) 179–206
4.7. Application to ISP 39
Application to ISP 39 is the first application of
FFTBM to severe accident. The ISP 39 is entitled
‘FARO-Test L-14 on Fuel coolant interaction and
quenching’. The main objectives were benchmarking the predictive capabilities of computer codes
used in the evaluation of fuel coolant interaction
and quenching phenomena on basis of the FARO
test L-14. Particular emphasis was on vessel pressurization, premixing aspects, debris formation
and cooling, quenching and steam production
rates, quantification of hydrogen formation rate.
It was ‘open’ exercise, with experimental results
provided. The results for time interval 1.42–4.5 s
are shown in Table 9.
This application (D’Auria and Galassi, 1997) is
peculiar in definition of weighting factors. First
the best calculation for pressure (PMED) was
identified. This AA value was used as reference to
normalize AA for other variables. In this way the
ratio between AA for PMED and AA for variable
gives the weighting factor for that variable (for
example the best ranked is calculation 15, therefore, wf for ENERTOT is 0.096/0.14 =0.69).
One of the conclusions of the performed study
was that, it confirmed the capabilities of the
FFTBM method in ranking generic calculation
results. Qualitative evaluation allows and allowed
better understanding of the results produced by
FFTBM and constitutes a pre-requisite for the
195
application of the FFTBM, as already mentioned.
For such analysis the reader is referred to (D’Auria and Galassi, 1997). The use of method is more
and more powerful when more and more applications are completed addressing the same phenomena. Addressing the same phenomena and the
weighting factors can be assigned on a physical
basis. In this sense, the presented application to
the study of corium– water interaction should be
considered as pilot one.
4.8. Application to ISP 42
The ISP 42 entitled ‘ISP-42- PANDA Test
‘TEPPS’’ main objectives were to study transient
and quasi steady– state operation of a passive
containment cooling system, primary coolant system and containment system behavior coupled,
primary coolant system under low pressure natural convection conditions, steam condensation in
the presence of non-condensables, mixing and/or
stratification of light or heavy gases in the containment and mixing and/or stratification in large
water pools.
The ISP 42 was performed in the PANDA
facility as a sequence of phases A–F (Auber and
Dreir, 1999). The quantitative analysis was performed for the phase A: passive containment
cooling system start up. Globally 49 calculations
were performed by the participants belonging to
ten European organizations. Eight different thermal –hydraulic computer codes were used.
Table 8
Comparison of pre- and post-test calculations for ISP 35 in time interval (200–1790 s)
ID
1
2
3
4
5
6
7
8
9
10
Participant/Country/Code
GRS, Germany, RALOC
TRACTEBEL, Belgium, MELCOR 1.8.2
SNL, USA, CONTAIN 1.12
NNC Ltd., UK, COMPACT
KEMA, Netherlands, MAPP 4.0
JRC Ispra, CEC, CONTAIN 1.12
JAERI, Japan, CONTAIN 1.12
IVO, Finland, RALOC
University of Pisa, Italy, FUMO
AEA, UK, CONTAIN 1.12
*Calculations not submitted.
Pre-test
Post-test
1/WFtot
AAtot
1/WFtot
AAtot
134.1
134.6
132.3
126.4
143.3
174.8
138.3
121.3
150.2
141.0
0.064
0.056
0.066
0.067
0.055
0.091
0.068
0.047
0.067
0.057
142.4
*
*
*
128.7
*
175.8
*
127.7
*
0.026
*
*
*
0.033
*
0.040
*
0.034
*
196
ID
3
15
2
1
11
13
9
10
8
6
12
5
wf
1
0.69
2.2
0.21
2.2
2.1
1.7
Pressure
(PMED)
Energy
(ENERTOT)
Steam temperature
(TMEDSTEAM)
Mixture water
level (LTC)
Average
temperature-elevation
400 mm (TAVG
400×1E-3)
Average
temperature-elevation
800 mm (TAVG
800×1E-3)
Average
temperature-elevation
1200 mm (TAVG
1200×1E-3)
0.105
0.096
0.080
0.074
0.065
0.075
0.100
0.061
0.116
0.273
0.065
0.134
0.086
0.140
0.089
0.181
0.286
0.367
0.151
0.418
0.202
0.656
0.524
0.171
0.068
0.043
0.063
0.055
0.109
0.091
0.063
0.047
0.293
0.130
0.281
0.591
0.156
0.463
0.171
0.114
0.102
0.077
0.333
0.147
0.210
0.293
0.211
0.269
0.041
0.044
0.074
0.080
0.052
0.053
0.080
0.084
0.060
0.080
0.096
0.056
0.050
0.047
0.065
0.070
0.040
0.035
0.076
0.062
0.039
0.066
0.073
0.038
0.050
0.056
0.053
0.065
0.044
0.037
0.041
0.088
0.053
0.099
0.091
0.048
AAtot
Rank
0.062
0.067
0.070
0.077
0.079
0.079
0.081
0.094
0.123
0.154
0.159
0.187
1
2
3
4
5
6
7
8
9
10
11
12
A. Prošek et al. / Nuclear Engineering and Design 217 (2002) 179–206
Table 9
Results obtained by FFTBM for ISP 39 in time interval (1.42–4.5 s)
A. Prošek et al. / Nuclear Engineering and Design 217 (2002) 179–206
197
Table 10
Results obtained by FFTBM for ISP 42.
ID
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Participant/country/code
CEA Grenoble, France, ENEA Bologna,
Italy, CATHARE
Nuclear Research Institute Rez, Czech
Republic, RALOC
Paul Scherrer Institut (A), Switzerland,
RELAP5/MOD3.2
VEIKI Budapest, Hungary, CONTAIN
Nuclear Research Institute Rez, Czech
Republic, CATHARE
Nuclear Research Institute Rez, Czech
Republic, CATHARE
Paul Scherrer Institut (B), Switzerland,
RELAP5/MOD3.2
VEIKI Budapest, Hungary, CONTAIN
Paul Scherrer Institut, Switzerland, GOTHIC
Paul Scherrer Institut, Switzerland, SPM
Paul Scherrer Institut, Switzerland, GOTHIC
University of Pisa, Italy,
RELAP5/MOD3.2.2
GRS Köln, Germany, COCOYS
NRG Arnheim, Netherlands, SPECTRA
Taek Ankara, Turkey, RELAP5/MOD3.2.2
Università Politecnica Catalunya, Spain,
RELAP5/MOD3.2
Time interval 0–3900 s
Time interval 0–5400 s
AAtot (10 var.)
AAtot (22 var.)
AAtot (10 var.)
AAtot (22 var.)
0.37
0.19
NA
NA
0.19
0.18
0.18
0.19
0.35
0.26
0.26
0.22
0.26
0.33
0.16
0.18
0.17
NA
0.13
NA
0.26
0.12
0.34
0.16
0.22
0.20
0.17
0.17
0.26
0.13
0.18
0.12
0.21
0.20
0.15
0.13
0.14
0.11
0.17
NA
0.17
0.11
0.16
0.17
NA
0.12
0.13
0.11
0.31
0.12
0.30
0.31
0.20
0.08
0.18
0.40
0.27
0.11
0.26
0.27
0.17
0.08
0.18
0.35
NA, calculation ended earlier.
For the FFTBM analysis (Aksan et al., 2001)
the time intervals selected were from 0 to 3900 s
(because three calculations ended around 3900 s)
and from 0 to 5400 s. The analysis was performed
with ten variables delivered by all participants and
22 variables, significant for the analysis (including
liquid masses, pressures, temperatures and mass
flows). When not all variables were delivered, it
was assumed that the precision of a participant in
calculating the not delivered variables would be
the average of the precision gained in calculating
the delivered parameters. The results for two time
intervals and with different number of variables
used for total accuracy quantification are shown
in Table 10. Again it can be seen that ten variables are not enough to completely characterize
the transient. The calculated accuracy is good for
most of calculations, both for results obtained by
primary system and containment codes. In this
study AAtot again proved to be congruent and
realistic, providing a valuable, objective and
quantitative accuracy evaluation, likely helpful in
the code assessment process.
4.9. Application to IJS calculations of
IAEA-SPE-2
In the study (Frogheri et al., 1995) 20 variables
were selected to characterize the calculation in the
three time intervals for post-test calculations using
different nodalizations (detailed, middle, simple
and miniature with 102/112/99, 71/76/84, 41/43/60
and 30/31/49 volumes/junctions/heat structures,
respectively). The results of the FFTBM analysis
presented in Table 11 confirm the conclusions
previously published in (Mavko et al., 1995)
where, it was stated that phenomena during the
IAEA-SPE-2 SB LOCA experiment were pre-
198
A. Prošek et al. / Nuclear Engineering and Design 217 (2002) 179–206
dicted satisfactorily by RELAP5/MOD2 and
MOD3 code versions, that even the simplest
model produced satisfying results compared with
the detailed one and that only down-comer level
was predicted wrongly for simple and especially
miniature model. Namely, the AA for downcomer level was calculated as 0.35, 0.34, 0.43 and
0.61 for detailed, middle, simple and miniature,
respectively. This perfectly agrees with the
FFTBM criterion set to 0.4 for single variable.
The total AA below 0.3 shown in Table 11 again
confirms the conclusion from previous study that
the experiments were predicted satisfactorily even
with the simplest model.
4.10. Application to IJS calculations of
IAEA-SPE-4
This application was performed to IAEA-SPE-4
experiment (Prošek et al., 1995). This is 7.4% SB
LOCA, with the break located at the top of the
external vessel down-comer, no HPIS available
and secondary feed and bleed initiation. The main
objectives of the standard problem exercise were
to extend the database for VVER-440/213 type
NPPS, to validate the computer codes being
adapted for VVER safety analyses and to facilitate the exchange of information on application of
advanced thermal– hydraulic computer codes.
The results of the FFTBM application are
shown in Table 12. In this application 22 variables
were selected to characterize the calculation in the
five time intervals for both pre- and post-test
calculations. The post-test calculations were char-
acterized with greater accuracy than pre-test. In
this calculation, the same input deck was used to
compare accuracy of three versions of RELAP5
code. In this case the best results obtained were
with RELAP5/MOD2 code.
4.11. Application to DCMN calculations of
IAEA-SPE-1 – 4
The study described in (D’Auria et al., 1996)
deals with the application to tests performed in
PMK facility for the primary pressure accuracy.
IAEA-SPE-1 simulates a 7.4% SB LOCA with
break located in the upper head of the downcomer, without the hydro– accumulator injection
and without high pressure injection system.
IAEA-SPE-2 simulates the same LOCA like
IAEA-SPE-1 but with available hydro–accumulators. IAEA-SPE-3 simulates a transient originated
by the opening of a break located in the upper
part of ascending riser in the steam generator.
In the study only primary pressure was considered, shown in Table 13. The main conclusions
were that, for IAEA-SPE-3 the results were clearly
unacceptable, that more complex transient lead to
worse results than simple ones (steam generator
tube rupture in IAEA-SPE-3), going from pre-test
to post-test the assumptions made in the modeling
could be adjusted resulting in a better representation of the transient and that advanced codes and
recent applications produce better results in the
simulation of the considered transients with respect to the precedent ones.
Table 11
Application of FFTBM to IJS IAEA-SPE-2 calculations for 20 selected variables
Code used
Nodalization
Time intervals
0–200 s
RELAP5/MOD2
RELAP5/MOD2
RELAP5/MOD2
RELAP5/MOD2
RELAP5/MOD3
Detailed
Medium
Simple
Miniature
Detailed
0–500 s
0–996 s
1/WFtot
AAtot
1/WFtot
AAtot
1/WFtot
AAtot
21.7
22.32
24.32
20.46
22.97
0.079
0.076
0.066
0.073
0.077
25.38
26.35
26.54
29.19
24.24
0.151
0.141
0.148
0.162
0.137
24.68
24.62
25.5
30.58
26.13
0.169
0.156
0.176
0.204
0.154
A. Prošek et al. / Nuclear Engineering and Design 217 (2002) 179–206
199
Table 12
Application of FFTBM to IJS IAEA-SPE-4 calculations for 22 selected variables
Calculation
Time intervals
0–50 s
0–350 s
0–1200 s
0–1400 s
0–1800 s
1/WFtot
AAtot
1/WFtot
AAtot
1/WFtot
AAtot
1/WFtot
AAtot
1/WFtot
AAtot
Pre-test
M2A
M2B
M3.1A
M3.1B
M3A
M3B
14.5
14.5
14.5
14.7
16.1
14.4
0.060
0.070
0.066
0.063
0.072
0.072
23.1
22.1
18.3
21.3
19.2
18.8
0.131
0.138
0.168
0.161
0.164
0.132
23.7
23.8
24.0
24.0
23.5
21.6
0.149
0.160
0.187
0.236
0.188
0.177
23.7
24.0
23.7
23.8
23.6
22.2
0.197
0.207
0.237
0.267
0.238
0.231
27.5
27.6
27.8
27.2
28.0
26.3
0.185
0.194
0.225
0.248
0.222
0.211
Post-test
M2
M3.1
M3
14.0
14.3
17.0
0.064
0.070
0.071
22.3
20.7
21.3
0.114
0.152
0.152
22.1
22.2
22.0
0.134
0.172
0.178
22.4
22.9
22.9
0.183
0.226
0.231
26.1
27.4
29.1
0.175
0.214
0.217
M2-RELAP5/MOD2/36.05, RELAP5/M3-MOD3 5m5, M3.1-MOD3.1; A, break on top of the DC (normal location); B, break in
the middle of DC.
4.12. Application to IAEA-SPE-4
(Italian– Slo6ene – Hungary– IAEA cooperation)
The analysis of code accuracy for IAEA-SPE-4
pre- and post-test calculations is presented in
(Mavko et al., 1997) while, qualitative analysis is
presented in (Prošek and Mavko, 1996). In total
18 pre-test and 22 post-test calculations were performed by 19 organizations from 15 countries. In
Table 14 the results for primary pressure accuracy
and total accuracy were compared with for 13
organizations performing both pre-test and posttest calculations. As only 12 variables were available for each participant the analysis is somewhat
less complete. However, the limit (0.1) for the
primary pressure accuracy allows such a comparison. For the best calculations the same result is
obtained by ranking, by pressure accuracy or
total accuracy. An exception are CATHARE calculations, one with high pressure accuracy giving
relatively low total accuracy (calculation three)
and another with poor pressure accuracy (calculation seven) giving quite good total accuracy. Nevertheless, in the post-test, the calculation seven
achieved second best primary pressure and the
best total accuracy. On the other hand, calculation three achieved the best pressure and was
ranked again very low. Most calculations predict
primary pressure very good with one exception
(calculation 13). According to acceptance criteria,
all calculations from Table 14 are very good
calculations which shows that the transient was
not complex. As similar WFtot were calculated in
all cases, the accuracy is judged only based on
AAtot results. In general the post-test calculations
shows better accuracy than pre-test. All but one
participant improve the pressure. However, it is
interesting that four calculations not improving
the post-test total accuracy were ranked among
the six best pre-test calculations. It seems that
changes (tuning the results) in the model did not
always contribute to better results, but we are
limited to make any further conclusions because
of limited number of variables used in the analysis. The validity of predictions is also limited by
discrepancies highlighted in the qualitative accuracy evaluations. Several calculations did not predict the core dry-out occurrence and should have
been excluded from further quantitative accuracy
evaluations subject to assessment criteria. However, the dry-out in the experiment constitutes a
typical bifurcation phenomenon controlled by minor variations of boundary conditions whose prediction is beyond the capabilities of the present
200
A. Prošek et al. / Nuclear Engineering and Design 217 (2002) 179–206
generation codes, and the overall transient scenario is not affected by the phenomenon. Both
these observations led to the decision to include
all calculations in the quantitative accuracy
evaluation.
4.13. Application to SB LOCA database
The application of FFTBM to SB LOCA database was performed to evaluate accuracy of RELAP5/MOD2 and CATHARE2 code. The studies
are described in D’Auria et al., (1997b) and
D’Auria and Ingegneri, (1997). The database consists of 12 SB LOCA tests performed in four
integral test facilities (LOBI, SPES, BETHSY and
LSTF). Additionally, six tests are counterpart
tests (BL-34, BL-44, SB-03, SB-04, 6.2 TC and
SB-CL-21). The RELAP5/MOD2 was used to
simulate all transients while CATHARE was used
for simulation of counterpart tests only.
The results of FFTBM application are shown
in Table 15. In all tests the same set of 24
variables was used. However, as some variables
were missing or not available in the database this
number vary from 13 to 24. When comparing the
results to acceptance criterion all calculations
fulfill criterion 0.4 for total accuracy. When comparing RELAP5/MOD2 and CATHARE2 code,
the accuracy is similar. Some calculations are
better with RELAP5/MOD2 and some with
CATHARE2. The mean total accuracy for counterpart test is 0.3 for both RELAP5/MOD2 and
CATHARE2. The highest accuracy achieved was
0.14 and the lowest 0.39.
4.14. Application to SB LOCA spectrum
In the study by Kljenak and Prošek, (1996)
eight different scenarios were selected to compare
the AA for 19 variables. Because these variables
were not standard, special weighting factors were
defined based on the phenomena identification
and ranking table (PIRT) process. Additionally,
an analysis using equal weighting factors was
performed. Eight different break sizes were compared with 5.08 cm (2 inch) break (which was
used instead of an experiment): 3.81, 4.13, 4.45,
4.76, 5.40, 5.72, 6.03 and 6.35 cm. This is a special
application of FFTBM. The results are shown in
Fig. 9, compared with case without weighting
(AAtot is mean of single variable AA). Different
slope of the line on the left and right from 5.08
cm is partly due to the fact that FFTBM gives
different results if we compare with smaller signal
to larger or larger to small. Nevertheless, the
trends are monotonic which shows that the more
similar are the breaks the more similar are also
transients. This fact supports FFTBM method in
that we can physically expect that the larger are
differences in the break size the more different are
the transients.
5. Results and discussion
The presented results of FFTBM applications
suggest that the FFTBM is a suitable method for
accuracy quantification of thermal–hydraulic
code calculations if the latter can be compared
Table 13
Application of FFTBM to DCMN IAEA-SPE-1–4 for primary pressure
Experiment
Time interval
Calculation
Code used
1/WF
AA
IAEA-SPE-1
0–997 s
IAEA-SPE-2
IAEA-SPE-3
IAEA-SPE-4
0–847 s
0–1819 s
0–1800 s
Pre-test
Pre-test
Pre-test
Pre-test
Pre-test
Pre-test
Post-test
Post-test
Post-test
RELAP4/MOD6
RELAP5/MOD1-EUR
RELAP5/MOD2
RELAP5/MOD2
RELAP5/MOD2.5
CATHARE
RELAP5/MOD2.5
RELAP5/MOD3.1
CATHARE
7.3
9.8
20.9
11.3
26.8
9.4
24.7
29.2
14.6
0.117
0.157
0.227
0.310
0.091
0.181
0.105
0.067
0.068
A. Prošek et al. / Nuclear Engineering and Design 217 (2002) 179–206
201
Table 14
Application of FFTBM to IAEA-SPE-4 in time interval (0–1790 s)
ID
1
2
3
4
5
6
7
8
9
10
11
12
13
Participant/country/code
University of Zagreb, Croatia, RELAP5/MOD2
Lappeenranta University of Technology, Finland,
RELAP5/MOD3.1
Institute de Protection et de Sûreté Nucléaire, France,
CATHARE2 V1.3E
GRS, Germany, ATHLET
Siemens AG, Germany, RELAP5/MOD2.5
Electric Power Research Institute, Hungary, MELCOR
University of Pisa, Italy, CATHARE2 V1.3E
Institute for Atomic Energy, Poland, RELAP5/MOD2
Kurchatov Institute, Russia, RELAP5/MOD3
‘Jožef Stefan’ Institute, Slovenia, RELAP5/MOD2
‘Jožef Stefan’ Institute, Slovenia, RELAP5/MOD3
‘Jožef Stefan’ Institute, Slovenia, RELAP5/MOD3.1
Texas A&M University, USA, RELAP5/MOD3.1
Pre-test
Post-test
Pressure
Total
Pressure
Total
AA
WF
AAtot
WFtot
AA
WF
AAtot
WFtot
0.11
0.08
0.03
0.04
0.26
0.22
0.04
0.03
0.09
0.08
0.06
0.05
0.28*
0.23
0.04*
0.03
0.08
0.04
0.31*
0.03*
0.06
0.05
0.29
0.03
0.16
0.13
0.20
0.15
0.07
0.23
0.09
0.10
0.11
0.16
0.03
0.03
0.03
0.05
0.05
0.06
0.04
0.05
0.04
0.05
0.39
0.31
0.35*
0.23
0.20
0.30
0.22
0.24
0.27
0.30
0.03
0.03
0.04*
0.03
0.03
0.04
0.03
0.04
0.03
0.04
0.11
0.09
0.08
0.08
0.08
0.11
0.08
0.09
0.08
0.15
0.05
0.03
0.04
0.05
0.05
0.05
0.04
0.04
0.04
0.06
0.30
0.26
0.24
0.19*
0.22
0.30
0.20
0.25
0.25
0.26
0.04
0.04
0.04
0.03*
0.03
0.04
0.03
0.03
0.03
0.04
*Analysis performed for 11 variables.
with experimental results. It was mainly used for
SB LOCA calculations. Table 16 shows the lowest calculated total average accuracy for each ISP
(among the submitted ISP calculations to which
FFTBM was applied) or organization calculation
(among several calculations with different code
versions). For each test, the time interval (end of
transient), the type of calculation (calculation
submitted to ISP or SPE, or non-submitted calculation performed by organization later), the
code used, the number of selected variables, the
primary pressure accuracy and the total accuracy
(with standard and average weights) are shown.
It can be seen that for primary system calculations the set criterion K=0.4 for total accuracy
was achieved while for severe accidents the
method is presently limited to ranking of calculations (there are no set criteria). The total accuracy using standard weighting factors ranges
from 0.14 to 0.39 for best calculations and may
support the acceptability limit K= 0.4 as reasonable criterion. However when observing obtained
total accuracy it can be noticed that complexity
of the transient is an important factor. The fact
that many best ISP calculations were obtained by
RELAP5 code can be partly attributed to the
fact that it is a widely used code. Table 16 also
shows that the tests simulated VVER transients
are ranked in the first half. The reason may be
less complex transients. This study has taken step
also in the direction of defining the relation between results obtained by standard weighting factors and average weighting factors. The
correlation factor r=0.92 indicates that when a
sufficient number of variables is used in the
FFTBM analysis the influence of engineering
judgement (in the specification of weighting factors) on the results is less important. When looking at the primary pressure AA the correlation
between it and total AA calculated with standard
weighting factors is r=0.3 indicating hardly any
relation. That means, care must be taken while
comparing different transients with primary pressure accuracy only.
The proposed method has still some weak
points which could be resolved, therefore, further
investigations may be performed, especially to
give meaning also to the phase spectrum.
A. Prošek et al. / Nuclear Engineering and Design 217 (2002) 179–206
202
Table 15
Results of FFTBM application to SB LOCA database
Facility
LOBI
LOBI
LOBI
LOBI
LOBI
LOBI
SPES
SPES
BETHSY
BETHSY
LSTF
LSTF
a
b
Test
BL-06
BL-12
A1-82
A2-81 (ISP-18)
BL-34b
BL-44b
SB-03b
SB-04b
6.2 TCb
9.1b (ISP 27)
SB-CL-21b
SB-CL-18 (ISP 26)
RELAP5/MOD2
CATHARE2 V1.3U
NV
AAtot
WFtot
23
23
21
20
24
24
22
22
23
13
21
18
0.29
0.24
0.14
0.28
0.33
0.33
0.32
0.26
0.28
0.39
0.31
0.15
0.02
0.04
0.04
0.04
0.05
0.05
0.05
0.06
0.06
0.08
0.05
0.08
NV
AAtot
WFtot
a
a
a
a
a
a
a
a
a
a
a
a
24
24
22
22
24
0.29
0.39
0.26
0.28
0.23
a
20
a
0.35
a
a
0.05
0.05
0.06
0.05
0.05
a
0.05
a
Not available.
Counterpart tests.
Nevertheless, the obtained results showed that
the method is good enough for quick assessments
of which variables or phases of the transients need
to be improved. One important result is also that
even if better methods for code accuracy assessment would be developed the result would still
depend on the complexity of the transient. Without
this information, setting criteria with respect to
what is good or not would be less meaningful.
However, such information again requires some
engineering judgement. At present moment it
seems that the large number of applications is one
of the main advantage of choosing existing
FFTBM for accuracy assessment of code calculations for experiments. For single variable comparison also other powerful indices of accuracy may be
used as proposed by Kunz et al., 2002. This may
be more important for code developers. However,
even in this case it should not be forgotten that, the
improving one parameter might deteriorate the
other parameters. Therefore, the information on
total accuracy is in favor to choose FFTBM.
The applications presented could be a reference
database for new users of FFTBM when to stop
improvements to their simulation. This is one of
the first attempts to present best (as judged by
FFTBM) results of several calculations in one
place.
6. Conclusions
A methodology suitable to quantify the BE
codes accuracy with respect to experimental results has been developed. This integral method is
using the FFT to represent the code predictionsexperimental data discrepancies in the frequency
domain. The weighting factors and acceptability
criteria that are part of the evaluation process
were set based on best engineering judgement.
The AA and WF were calculated for several code
calculations of experimental results to estimate
Fig. 9. Application of FFTBM to SB LOCA spectrum.
A. Prošek et al. / Nuclear Engineering and Design 217 (2002) 179–206
203
Table 16
Summary of most accurate ISP, SPE or organization calculations as judged by FFTBM
Test
Time interval
Type of calculation
Code used
NV
AApr
AAtot
(stand. wf)
AAtot
(avg. wf)
ISP 35b
ISP 42 b
ISP 39b
200–1790 s
0–5400 s
1.2–4.5 s
MAPP 4.0
SPECTRA
COMETA
2d-JRC
RELAP5/
19
22
19
NA
0.03
0.11
0.03a
0.08
0.11a
NA
0.14
NA
21
0.09
0.14
0.23
MOD2
RELAP5/
9
NA
0.15
NA
18
0.02
0.15
0.21
20
0.12
0.15
0.22
22
0.08
0.18
0.24
10
0.04
0.18
0.18
11
0.08
0.19
0.30
24
23
0.10
0.06
0.23
0.24
0.33
0.30
22
22
0.08
0.08
0.26
0.26
0.38
0.39
20
0.10
0.28
0.34
24
23
0.12
0.13
0.29
0.29
0.40
0.42
14
0.21
0.30
0.43
25
NA
0.3
NA
21
0.04
0.31
0.44
24
0.04
0.33
0.37
25
0.09
0.39
0.57
13
0.11
0.39
0.41
LOBI A1-82
0–4944 s
Best ISP (post-test)
Best ISP
Best ISP (open
exercise)
PISA calculation
ISP 33c
0–6600 s
Best ISP (post-test)
ISP 26
0–1000 s
PISA calculation
IAEA-SPE-2 c
0–996 s
IJS calculation
0–1800 s
IJS calculation
0–350 s
Best ISP (post-test)
0–1790 s
Best SPE (post-test)
BETHSY 6.2 TC
LOBI BL-12
0–2179 s
0–3768 s
PISA calculation
PISA calculation
SPES SB-03
SPES SB-04
0–2034 s
0–1637 s
PISA calculation
PISA calculation
ISP 18
0–4939 s
PISA calculation
LOBI BL-34
LOBI BL-06
0–2400 s
0–7000 s
PISA calculation
PISA calculation
ISP 22
0–6000 s
Best ISP (post-test)
ISP 27
0–8000 s
Best ISP (pre-test)
LSTF SB-CL-21
0–2113 s
PISA calculation
LOBI BL-44
0–2350 s
PISA calculation
ISP 27
0–8000 s
IJS calculation
ISP 27
0–8000 s
PISA calculation
IAEA-SPE-4
c
ISP 21
IAEA-SPE-4
c
MOD3
RELAP5/
MOD2
RELAP5/
MOD3
RELAP5/
MOD2
RELAP5/
MOD2
CATHARE2
V1.3E
CATHARE2
RELAP5/
MOD2
CATHARE2
RELAP5/
MOD2
RELAP5/
MOD2
CATHARE2
RELAP5/
MOD2
RELAP5/
MOD2
RELAP5/
MOD3
RELAP5/
MOD2
RELAP5/
MOD2
RELAP5/
MOD3.1
RELAP5/
MOD2
Best ISP, it means that this is the pre- or post-test submitted calculation with the lowest AAtot calculated.
a
Special weights selected.
b
Severe accident.
c
VVER transient.
quantitatively the accuracy of the respective code
predictions. It was shown that most accurate calculations as judged by FFTBM achieved the total
average accuracy below 0.4 which was set as
‘acceptability limit’ for calculation (see Section
2.4).
204
A. Prošek et al. / Nuclear Engineering and Design 217 (2002) 179–206
The application of the method to the international standard problems, standard problem exercises and other experiment simulations showed
the FFTBM as a convenient mathematical tool
for quantitative assessment of experimental results. The results of such quantitative assessments
together with qualitative evaluation may assist the
decision whether or not the simulation needs to be
improved.
Appendix A. Nomenclature
ISP
IVO
JAERI
JINS
JRC
K
KAERI
KINS
LOBI
LSTF
LTKK
D0 F
F0
AA
ABB
ACC
AEA
ATHLET
BETHSY
BL
CEA
CIAE
DC
DCMN
Dm
DP
ECCS
ENEA
ENEL
f
F
GRS
IA
IAEA
IJS
INEL
error function amplitude spectra
time function amplitude spectra
average amplitude
Asea Brown Boveri
accumulator
Atomic Energy Authority
Analyses of Thermo–hydraulics in
Leaks and Transients
Boucle d’etudes thermo–hydrauliques systeme
broken loop
Commissariat a l’Energie
Atomique
China Institute of Atomic Energy
down-comer
Dipartimento di Costruzioni Meccanice e Nucleari
maximum value of error function
differential pressure
emergency core cooling system
Ente per le Nuove tecnologie,
l’Energia e l’Ambiente
Ente Nazionale per l’Energia
Elettrica
frequency
time function
Geschellschaft für Anlagen- und
Reaktor Sicherheit (GRS) mbH
Index of agreement
International Atomic Energy
Agency
Institut ‘Joz' ef Stefan’
Idaho National Engineering
Laboratory
MAAP
MFE
NPPRI
NV, Nvar
PRZ
PWR
RELAP
SG
SNL
SPE
SPES
t
T
~
UH
UK
UKAEA
VEIKI
VTT
VVER
WF
wf
ZFK
International Standard Problem
Imatran Voima Oy
Japan Atomic Energy Research
Institute
Japan Institute of Nuclear Safety
Joint Research Center
acceptability factor
Korea Atomic Energy Research
Institute
Korea Institute of Nuclear Safety
Loop for Off-normal Behaviour
Investigation
Large Scale Test Facility
Lappeerannan teknillinen korkeakoulu (Lappeenranta University of
Technology)
Modular Accident Analysis
Program
Mean Fractional Error
Nuclear Power Plant Research
Institute
number of analyzed variables
pressurizer
pressurized water reactor
Reactor Excursion and Leak Analysis Program
steam generator
Sandia National Laboratories
standard problem exercise
Simulatore per esperienze di
sicurezza
time
transient time
sampling interval
upper head
United Kingdom
United Kingdom Atomic Energy
Authority
Villamosenergiaipari Kutató Intézet (Institut for Electric Power
Research)
Valtion Teknillinen
Tutkimuskeskus
Voda-Vodianye Energeticheski
Reaktoryi
weighted frequency
weighting factor
Research Center Rossendorf Inc.
A. Prošek et al. / Nuclear Engineering and Design 217 (2002) 179–206
ZF
error function
Subscripts
calc
d
exp
i
max
norm
pr
s
saf
tot
calculated
duration
experimental
i-th variable
maximum
normalized
primary pressure
sampling
safety
total
Superscripts
m
power of 2
References
Aksan, S.N., Bessette, D., Brittain, I., D’Auria, F., Gruber,
P., Holmström, H.L.O., Landry, R., Naff, S., Pochard,
R., Preusche, G., Reocreux, M., Sandervag, O., Städtke,
H., Wolfert, K., Zuber, 1987. CSNI Code Validation
Matrix of Thermo-Hydraulic Codes for LWR LOCA and
Transients. CSNI Report No. 132.
Aksan, S.N., D’Auria, F., Bonato, S., 2001. Application of
Fast Fourier Transform Based Method (FFTBM) to the
results of ISP-42 PANDA test calculations: Phase A.
Proceedings of the ICONE-9.
Ambrosini, W., Bovalini, R., D’Auria, F., 1990. Evaluation
of accuracy of thermal – hydraulic code calculations. Energia Nucleare 7, 5 – 16.
Ambrosini, W., Breghi, M.P., D’Auria, F., 1991. Evaluation
of code accuracy in the prediction of the OECD/CSNI
ISP 22, University of Pisa DCMN-NT 176 (91).
Auber, C., Dreir, J., 1999. ISP 42 Panda test Phases — History of events, Boundary and Initial Conditions, Calculation parameters, PSI Internal Report TM-42-98-40,
ALPHA 835-1, Villingen (CH).
Bovalini, R., D’Auria, F., Leonardi, M., 1992. Qualification
of the Fast Fourier Transform based methodology for
the quantification of thermalhydraulic code accuracy,
DCMN Report, University of Pisa, NT 194(92).
D’Auria, F., Galassi, G.M., 1997. Accuracy Quantification
by the FFT method in FARO L-14 (ISP 39) open calculations, University of Pisa, NT 309(97).
D’Auria, F., Ingegneri, M., 1997. Qualitative and quantitative analysis of CATHARE2 code results of counterpart
test calculations in LOBI, SPES, BETHSY, LSTF facilities, Proc. of NURETH-8, vol. 2, pp. 1195 – 1203.
D’Auria, F., Mazzini, M., Oriolo, F., Paci, S., 1989. Com-
205
parison report of the OECD/CSNI International Standard Problem 21 (Piper-one experiment PO-SB-7), CSNI
Report No. 162.
D’Auria, F., Leonardi, M., Pochard, R., 1994. Methodology
for the evaluation of thermalhydraulic codes accuracy.
Proc. Int. Conf. on New trends in Nuclear System Thermohydraulics, Pisa, pp. 467 – 477.
D’Auria, F., Oriolo, F., Leonardi, M. and Paci, S., 1995.
Code Accuracy Evaluation of ISP35 Calculations Based
on NUPEC M-7-1 Test. Proc. Second Regional Meeting:
Nuclear Energy in Central Europe, Nuclear Society of
Slovenia, Ljubljana, p. 516 –523.
D’Auria, F., Eramo, A., Frogheri, M. Galassi, G.M., 1996.
Accuracy quantification in SPE-1 to SPE-4 organised by
IAEA. Proceedings of the Fourth International Conference on Nuclear Engineering (ICONE-4), Vol. 3, American Society of Mechanical Engineers and Japan Society
of Mechanical Engineers, pp. 461 – 469.
D’Auria, F., Galassi, G.M., Belsito, S., Ingegneri, M., Gatta,
P., 1997a. UMAE application: Contribution to the
OECD/CSNI UMS Vol. 2. DCMN Report, University of
Pisa, NT 307(97) Rev. 1.
D’Auria, F., Galassi, G.M., Leonardi, M. and Galetti, M.R.,
1997b. Application of Fast Fourier Transform method to
evaluate the accuracy of SBLOCA data base, Procedings
of NURETH-8, Vol. 1, pp. 358 – 366.
D’Auria, F., Mavko, B., Prošek, A., 2000. Fast fourier
transform based method for quantitative assessment of
code predictions of experimental data, V: Proceedings,
(ANS Order, 700282), Washington, American Nuclear
Society.
Frogheri, M., Prošek, A., D’Auria, F., Leonardi, M.,
Mavko, B., Parzer, I., 1995. Application of the FFT
Method to the IAEA-SPE-2 and SPE-4. IJS Report, Institut ‘Joz' ef Stefan’, IJS-DP-7101, Rev.1.
Holmström, H., 1992. Quantification of code uncertainties,
Proceedings of the CSNI Specialist Meeting on Transient
Two-Phase Flow – Current Issues in System Thermal – hydraulics, Aix-en-Provence, France.
Hrvatin, S., Prošek, A., Kljenak, I., 2000. Quantitative assessment of the BETHSY 6.2 TC test simulation, Proceedings of the Eighth International Conference on
Nuclear Engineering (ICONE-8), Baltimore, MD, USA.
Kljenak, I., Prošek, A., 1996. Analysis of Similarity Between
Small-Break Loss-of-Coolant Accident Scenarios Based
on Identification and Ranking of Basic Phenomena, Proceedings of the third Regional Meeting ‘Nuclear Energy
in Central Europe’, Nuclear Society of Slovenia, pp.
285 – 292.
Kunz, R.F., Kasamala, R.F., Mahaffy, J.H., Murray, C.H.,
2002. On the automated assessment of nuclear reactor
systems code accuracy, Nuclear Engineering and Design
211, 245 – 272.
Lapendes, D.N., 1978. Dictionary of Scientific and Technical
Terms, second ed. McGraw-Hill Book Company, New
York, St. Louis, p. 1041.
Mavko, B., Parzer, I., Petelin, S., 1995. A modeling study of
the PMK-NVH integral test facility. Nucl. Technol. 105,
231 –252.
206
A. Prošek et al. / Nuclear Engineering and Design 217 (2002) 179–206
Mavko, B., Prošek, A., D’Auria, F., 1997. Determination of
code accuracy in predicting small-break LOCA experiment. Nucl. Technol. 120, 1 – 19.
OECD-NEA, 1989. CSNI International Standard Problem
N.21-Comparison Report vols. 1 and 2ª, OECD-CSNI
Report N.162.
Prošek, A., Mavko, B., 1995. Comparison of RELAP5/MOD2
and RELAP5/MOD3.1 Calculations Using FFT method,
Proceedings of the second Regional Meeting: ‘Nuclear
Energy in Central Europe’, Nuclear Society of Slovenia,
Ljubljana, pp. 533 – 540.
Prošek, A., Mavko, B., 1996. Application of qualitative analysis to IAEA-SPE-4 calculations, Proceedings of the third
Regional Meeting: ‘Nuclear Energy in Central Europe’,
Nuclear Society of Slovenia, Ljubljana, pp. 246 – 253.
Prošek, A., Mavko, B., 1997. Accuracy Quantification of
RELAP5/MOD3.1 Code for BETHSY 9.1B Test, Proceedings of the Annual Meeting on Nuclear Technology’97,
INFORUM Verlags-und Verwaltungsgesellschaft mbH,
pp. 169– 172.
Prošek, A., Parzer, I., Mavko, B., Frogheri, M., D’Auria, F.,
Leonardi, M., 1995. Application of the FFT Method to an
IAEA-SPE-4 Experiment Simulation, Proceedings of the
Annual Meeting on Nuclear Technology ‘95, INFORUM
Verlags-und Verwaltungsgesellschaft mbH, Germany, pp.
115 – 118.
Prošek, A., Kljenak, I., Mavko, B., 1996. Quantitative Similarity Analysis of Small-Break Loss-of-Coolant Accident Scenarios, Proceeding of the Fourth International Conference
on Nuclear Engineering (ICONE-4), Vol. 3: Safety and
Reliability, American Society of Mechanical Engineers and
Japan Society of Mechanical Engineers, pp. 79 – 86.
Purhonen, H., Kouhia, J., Holmström, H., 1994. OECD/
NEA/CSNI International Standard Problem No. 33-Comparison Report, NEA/CNSI/R(94)24.
Schultz, R.R., 1992. Methodology for quantifying calculational capability of RELAP5/MOD3 code for SBLOCAs,
LBLOCAs, and operational transients, presented at First
Code Assessment and Maintenance Program (CAMP)
Mtg., Villigen, Switzerland.