ML20151C924
| ML20151C924 | |
| Person / Time | |
|---|---|
| Issue date: | 01/18/1988 |
| From: | Catton I Advisory Committee on Reactor Safeguards |
| To: | Boehnert P Advisory Committee on Reactor Safeguards |
| References | |
| ACRS-CT-1912, NUDOCS 8804130272 | |
| Download: ML20151C924 (4) | |
Text
)
CT-/%
/
l'
o,,
UNITED STATES NUCLEAR REGULATORY COMMISSION y.,
n ADVISORY COMMITTEE ON REACTOR SAFE JUARDS
{
/g January 18, 1988 o,,
WASHINGTON. D. C. 20555
%, V
...+
MEMORANDUM FOR:
Paul Boehnert FROM:
Ivan Catton PEER REVIEW OF CSAU METHODOLOGY, JANUARY 12-13,
SUBJECT:
1988, BETHESDA, MARYLAND A Group was constituted to provide peer review of the CSAU method-olcgy. The Group was selected from universities and from industry.
For the most part, the Peer Review Group Members are familiar with the nuclear industry and the problems one faces when one tries to model a plant. The Peer Review Group Members are the following:
N. Todreas, MIT Chair F. Mayinger, TUM S. Panerjee, UCSB B. Mikic, MIT A.E. Dukler VH M. Thurgood J. Fell, UK G.B. Wallis, Dartmouth M. Ishii AHL L. Ybarrando The peer review was to be confined to the methodology itself. This was, however, found to be difficult because the method could only be 1
demonstrated using a code like TRAC and many of the Peer Review Group i
Members were as interested in the inner workings of the code as they were of the method. The charge to the Group was the following-l 1.
Is the methodology systematic, logical and practical?
2.
Has the NRC demonstrated a methodology which shows that best estimate methods can be used to meet the requirements of the modified rule, The Peer Review Group had many of the same concerns that we have expressed in the past; nodalization, time step and the use of engi-nearing judgement. Todreas viewed the method as a "disciplined approach that applies engineering judgement to the problem". This view was followcd 'oy Graham '/allis saying "engineering judgement is the last refuge when you don't Nye an answer".
SOME MEETING DETAILS Lou Shotkin continually emphasized that the CSAU is only a guide.
Somehow I get the feeli a th6t the CSAU is too loose. The revised g 42 g 2 88011s CT-1922 PDR
Peer Review of CSAU Methodology 1/12-13/88 Meeting January 18, 1988 ECCS Rule deliberately does not say how high the probability that the legislated limits will not be exceeded must be, only that it must be "high".
The Reg Guide says what "high" must be but not how much confidence I nust have in the answer, only that it must be "high".
One's level of ignorance may well determine one's confidence. This may place an unbearable burden on NRR.
It seems possible that "he who argues best" will get the most relief. A group like the present TPG w!ll not always be involved. A more prescriptive Reg Guide may be needed. Wallis clearly expressed concerns about the "judgement of NRC".
Here a group of prescribed IETs and SETS may be the answer.
Following Sol Levy's presentation of an over view of the CSAU, the Peer Review Group questions showed a strong interest in nodalization.
Banerjee corrented that nodalization that is not converged can be used to take you anywhere. By selecting a particular nodalization one can get good comparison with data. Jesse Fell noted that the TRAC code is made up of mocels that are just somewhat better than lumped parameter models.
I don't believe he meant this as a criticism of the code.
Dukler expressed his concerns about whether or not we have enough wisdom (experience) to make judgments. His concerns are similar to those of Wallis.
Roy Shaw described the step in the CSAU where one establishes what phenomena are important.
It was clear that past experience played an important role in the process. This lead Ybarrando to question whether or not one could repeat the process for the SBLOCA and ferret out all the important phenomena, ile feels this is a result of the large amount of data and code experience brought to bear on the LBLOCA. Althrugh he seemed to accept the procedure, Todreas felt that the method does not clearly define how the "importance" is estab-lished.
Gary Wilson (INEL) gave a minimum acceptable data base for demonstra-tion of LBLOCA CSAU:
Harviken SemiScale LOFT SCTF CCTF UPTF How this set was arrived at was not clear. This should be an inter-esting topic to pursue.
It is my view that this is what will rake the CSAU viable. Dukler was concerned that the above data set was too limited. He wanted to know why other data that addresses areas of uncertainty were not used. Banerjee was concerned that the code was too heavily tuned to LOFT.
J
Peer Review of CSAU Methodology 1/12-13/88 Meeting January 18, 1988 Rom Duffey addressed the scaling question. By plotting blowdown peak data from a number of experinents (LOFT, PBF, SemiScale, LOBI, and THTF) as a function of linear heating rate, he was able to show that 95% of the measurements fell within a plus or minus 361 deg F band.
There was no discernable effect of scale. He then corrected the curves for different decay heats, plotted the peak as a function of adiabatic ternperatare and found that 95% fell within a plus or minus 230 deg F band.
If the PBF data were removed, the band becomes even smaller. There were 1 mediate jests as to how we no longer needed a code to do the calculations. The simple scaling put it all on a curve whose spread is only nominally greater than that found using TRAC and a lot of dollars.
Some tricky things were done with TRAC to generate a response surface for uncertainty evaluation.
Some twenty extra thermally non-inter-active fuel rods were used along in seven runs where each of the important thermal hydraulic parameters were varied, one at a time.
This means that 140 separate sets of calculations were obtained with the seven runs.
It was pointed out by the comittee that there were not 140 independent sets of calculations. Many, including myself, feel that the core average conditions are not what the hot red sees.
Once it was pointed out, it became obvious that more would have to be done. Here, one does not know before hand whether or not there will be an affect but rust test for it. The results of the process were presented by Gerry Lelouche. Although he argued that they were i
complete, sound reasons were given for making sure the seven thermal hydraulic parameters were enough and that they had been lookea at in a sufficient enough number of combinations. The Group had some trouble during the presentation with definitions. They requested that things like "code dependency", "important phenomena", and "scaling" be clearly defined and spelled out in the methodology description.
l
SUMMARY
The Peer Review Group was not sure NRC will be able to do what the TPG has done, because of a lack of faith in their judgerent. Wallis is not sure NRC has the competence to audit an industry submital.
j There is a need for a more definithe statement (more prescriptive) of the CSAU methodology. The CSAU methodology does not require documen-tation supporting "phenomena selection" nor maximum and minimum values of needed parameters that will be required. The need for code audit was not clearly spelled out nor were that it should entail.
More work on nodalization is needed. One comes up with a nodalization scheme and handles it consistently, yet no intelligence is given as to i
whether nr not it is best.
In other words, this is the number you get with that nodalization. Banerjee is concerned about convergence,
.n
Feer Review of CSAU Methodology 1/12-13/88 Meeting
. January 18, 1988 parameter code than a finite difference be meaningful.
may not parameters in a bounciing 9:ay.The present effort is to ng30(?)
is just replacing one pessimism with another that is not pessimistic.
Westinghouse acknowledged this and noted that th u e as now considering a response surface approat.h.
ing to see that the Westinghouse bounding values were tot v ent from the TPG response surface 95% values of 1379 De ery differ-average peaking factor.
The Peer Review Group was concerned that no runs were ma some cross product contributors to uncerta e with Banerjee wanted all seven parameters pushed simultane u
cays some testing of combinations is in order to be sure that Wallis cross product tenns are as small as stated.
calculations of cross product terms were in o Fell also felt that some because the local flow conditi s wrong ons were unchanged from the core aver-age.
As noted by Jesse Fell, the peer review group is happy wit blowdewn example.
As a matter of fact, one could calcu back of an envelope.
e refill and reflood phases fare.
The SBLOCA will be even more Ciffi-cult.
Many of the Members were still unsettled about the scali given by Duffey and Lelouche.
ng arguments not understand what they were being told or they were Lelouche for his presentation style.
blowdown very convincing and know from personal experiencI reflood can be handled effectively.
e that Ybarrando raised the question of the CSAU and BWRs wh is less robust.
margin goes down, the QA requirements will have ere the data base The sloppy QA procedures of the past will have to be dealt up.
with.