ML20197D803
| ML20197D803 | |
| Person / Time | |
|---|---|
| Site: | 05000000, Davis Besse |
| Issue date: | 08/19/1985 |
| From: | Catton I Advisory Committee on Reactor Safeguards |
| To: | Boehnert P Advisory Committee on Reactor Safeguards |
| Shared Package | |
| ML19301C663 | List: |
| References | |
| FOIA-86-110 NUDOCS 8605150056 | |
| Download: ML20197D803 (3) | |
Text
.
s'
'o UNITED STATES NUCLEAR REGULATORY COMMISSION c
3 E
ADVISORY COMMITTEE ON REACTOR SAFEGUARDS o,,
f wasmscTom. o. c. 20sss August 19, 1985 MEMORANDUM FOR:
Paul Boehnert, Senior Staff Engineer FROM:
I. Catton, ACRS Consultant
SUBJECT:
ECCS SUBCOMMITTEE MEETING, JULY 31, 1985, WASHINGTON, DC Appendix K Revision Effort There is no doubt that the present EM models are too conservative and that some of the conservatism should be eliminated. More importantly, we all know of circumstances where EM modeling has resulted in designs that were not the best achievable. Set points are based on Chapter 15 results rather than physics. Simulators train operators to respond to non-physical behavior. It is my view that we should incorporate what we know about the physical processes into the codes, then add margin to the forcing function (decay heat and initial stored energy). Limiting values such as peak clad temperature could then be treated in a sensible way. The importance of the vendor or utility model goes further'than the large break loss of coolant accident. The codes will be used for small breaks and transients as well as for establishing operating procedures and the basis for simulator fidelity.
Good realistic modeling will depend on advanced codes like TRAC and RELAP-5. At this point, we really don't know how well the codes are dble to predict the various phenomena. There have been no reviews for several years. One must not forget that we are still extrapolating from small scale integral facilities to full scale LWRs. This extrapolation will require skilled analysts and a certain amount of faith. Further, codes with known improper physical representation will require modifica-tion. I was pleased to hear that Westinghouse planned to use TRAC as a basis for their new model. A detailed review of the code and their capabilities should be a part of a future ECCS Subcommittee meeting. We should ask, for example, for a detailed description of recent calcu-lations made for comparison with CCTF data. CCTF, being a fairly large integral facility, will allow us to determine whether or not an in-ternally consistant set of results are produced by the advanced codes.
You must defenc the answer you gst with a code in an engineering or fp1lg6860421
~
h \\\\v PEDROO6-110 PDR
p' Catton's ECCS 7/31 Coments August 19, 1985 scientific way. This means you must defend and argue that the details of the code are adequate to allow extrapolation from your data base to a full size LWR. Compensating errors and scaling will be the key issue.
The recent effort by GE that resulted in the SAFER code being accepted as a licensing tool was given as an example of how one should proceed in the development of a new licensing code. The gains made by GE were primarily a result of their EM code being so poor. Almost anything would have been an improvement. The statistical analysis presented by GE to argue that they could define a 95% uncertainty boundary was not convincing. There are too many approximate representations of physics for one to establish uncertainty with much rigor. Further, the code results are sensitive to time step magnitude and nodalization and we have no basis for assuming that the approaches taken for small systems will extrapolate to full size plants.
The report sumarizing the past ten years of research will be a vary valuable part of the Appendix K revision process. At present, the Staff does not intend to make the report part of the package that is sent out for public coment. To send it later than the proposed Regulatory Guide and not part of the overall package uncouples the basis for change from the change itself. With past history in mind, I believe the sumary report must be part of the package or the package should not go out.
Pumps On or Off The B&W approach is very simple. Turn the pumps on if you loose sub-cooling margin and turn tnem off when inventory is too low. This simple approach probably has a number of advantages. It does not, however, allow them to distinguish between a SBLOCA and an overcooling transient.
The B&W responses to questions about overcooling and to those of Reed's about natural circulation induced hot spots in the loop left me with the feeling that their analytical base for decision making is weak. The B&W plants have a history of troubles with their secondary side. As a result, I think they should be asked to devote a little more time to the i
questions raised at the Subcomittee meeting.
The CE approach was novel. Tripping pumps in opposite loops to buy time for the operator to make decisions is certainly a sensible thing to do.
CE gave good arguments for knowing core inventory while trying to argue that there was too much ambiguity in core inventory alone. The Westing-house analysis shows that steam generator U-tube voiding can occur and with it loss of natural circulation under some circumstances. The pressure then rises until the PORVs lift and the combined break /PORY flow is sufficient to reject enough energy. What happens to a CE PWR that has no PORVs? The approach taken to distinguish between overcool-ing and a break seems sensible.
hup
Catton's ECCS 7/31 Comments August 19, 1985 The time available to a Westinghouse PWR operator before he must take action seems to be enough for decisions to be made. It was particularly comforting to hear that they could show with their Best Estimate model that the window leading to core uncovery with excessive peak clad temperatures does not exist. They will trip the pumps for other rea-sons.
Multiple Instrument Tube Failures The Staff review showed that there was a possibility of a problem. This was based on a break flow model valid for a different set of circum-stances than is the case for instrument tube failures at the seal table.
The instrument tubes are about sixty feet along with a relatively small gap between the pipe wall and the instrument sensing line. I believe the flow rate will be much lower than used by the Staff, possibly by a factor of ten. One would have to loose an inordinate number of tubes before a problem would arise.
Davis-Besse Incident The Davis-Besse incident was an example of what good operators can do when everything goes wrong.
It was also a demonstration of where the core melt " outliers" are going to come from -- maintenance practices.
PRAs at present do not, to my knowledge, account for the complex inter-action that takes place between operators and the equipment they must depend on.
During the month that followed the incident, the NRC Staff tried to do some calculations mimicing the incident. The mainframe computer at INEL was not usable and LANL was not able to fully meet the stated needs of NRR. The Nuclear Plant Analyzer Development Program has been underway at the two laboratories for three or four years. One wonders how a mainframe computer could be down for a month or why a request for computations could not be completely fulfilled. One wonders why a PWR version of the code developed at BNL for BWRs isn't available.
f\\ \\ \\ t>
..