ML20148H629
| ML20148H629 | |
| Person / Time | |
|---|---|
| Issue date: | 11/14/1987 |
| From: | Catton I Advisory Committee on Reactor Safeguards |
| To: | Boehnert P NRC |
| References | |
| ACRS-CT-1907, NUDOCS 8801270267 | |
| Download: ML20148H629 (5) | |
Text
e T-MW Tpt UnM pa aseg
/
o UNITED ST Al t..,
! \\*
NUCLEAR REGULATORY COMMISSION ~~
~'~'
n 3,.. '
{c ADVISORY COMMITTEE ON RE ACTOR SAFEGUARDS o, hm WASHINGT ON, D. C. 20555 g
/
November 14,19#];,7,,;,gg_ ;
t.
~
MEMORANDUM FOR:
Paul Boehnert FROM:
Ivan Catton
SUBJECT:
Thiru ICAP Specialist Meeting Bethesda, Maryland 21-23 October 1987
GENERAL COMMENT
S The meeting was very interesting.
It was initiated with Brian Sheron telling the foreign participants that the USNRC was going to severly reduce funding of code development.
He said that major thermal hydraulic research in the US is completed.
This somewhat unsettled some of the foreigners. Now they will have to pay for what they get or do without. I think Sheron's position is long overdue.
Both LANL and INEL were asked to present a plan for code improvement listing the deficiencies to be addretsed in the improvement process. The lists presented by the two laboratories were similar both to each other and to lists we were seeing ten to flfteen years ago.
One has to wonder what the labs have been doing during this period.
There has been a great deal of work that could and should have been brought to bear.
I am left with the firm conviction that code developers cannot read anything other than a computer output.
A certain fuel vendor tried to use TRAC as a licensing tool. They found it to be user unf riendly, to have too many compensating errors, and to take too long to do computations. As a result, they gave up on TRAC after spending something in excess of $1M.
The Germans have had similar problems lead ng KWU to use a modified version of RELAP5 for licensing calculations. They have put in a 2D downcomer and modified the core calculations to account for some multi-D behavior.
The Germans are going I
to contribute money through EXX0:1 to complete the needed code development work on RELAP5.
There was a great deal of concern expressed by many about "user dependent uncertainties".
In this regard, it was found that the codes were both time step and node sensitive.
Some of the calculated oscillating behavior g12 7 871114 DESIGNATED ORIGINAL
[pg$
CT-1907 PDR
=.
P. A. Boehnert 2-November 14, 1987 disappears when the time step is sufficiently reduced.
I believe this;is a result of the extreme damping used in the numerical algorithms limiting the growth of the calculated instabilities.
It seems to me that further versions of TRAC-PF1 are not warranted.
With the Germans help, RELAPS will be made tolerable and its uncertainties quantified.
The present effort to determine TRAC calculational uncertainties will yield a computer inefficient tool adequate for use in the near term. The ef.fo.rt put into RELAPS will yield a more computer efficient tool for interim use allowing time for the next generation of LWR simulat, ion tools to be developed.
1 An ICAP code improvement steering group is going to be formed.
This long overdue step will lead to a group that will guide ICAP in the developmen-tal assessemnt of TRAC-PF1/ Mod 2.
The underlying assumption is that the ICAP participants will contribut-the necessary resources (money and labor). The result will be a final release by June of 1989.
DETAILED REMARK _S, S
i RELAPS-- A list of code improvements were justified by listing goals and deficiencies:
Goals improved physical modeling faster execution easy portability improved documentation l
Deficiencies interfacial processes flow regimes I
wall heat transfer process models numerics The goals are motherhood and were the basis for the advanced code program at the cutset.
It is not clear how much more can be done in meeting the objectives or gotis with the present code.
Some predict a maximum speed increase of about 30% which does not justify a great deal of effort.
Portability is an important aspect of any government developed code.
This should have been the case at the outset.
It is not clear that NRC requires any kind of standardization that will lead to portabil*;ty.
Portability is, as a result, paid for as an added task upon completion of the code. This is a silly way to do business.
RES should be esked why they do not specify a standard to avoid such problems.
The list of deficiencies is the same list that was evident ten years ago.
Many of the deficiencies are code inadequacies rather than la*.4 of understanding the physics, the physics are understood and in many cases correlations exist.
One can get a measure of the time delay between when
P. A. Boehnert November 14, 1987 the technical community reports on a process and when the code developer incorporates it into the code.
It appears to be about three to four '
years.
The problem is either that the code structure determines the physics or the developers do not read the literature.
A Mod 3 version of RELAPS was described that would meet the goals'and eliminate the list of deficiencies.
In my view such a code would just be more of the same. One does a cost benefit on most research'prbjects. It seems that requiring such an analysis for code development projects has been overlooked and ought to be considered.' Many of the code improvements that would lead to a Mod 3 were based on something being done by LANL for TRAC. This leads me to conclude that the codes are rapidly, if not already, becoming so similar as to be fully duplicative.
In tires of a decreasing budget, duplication is hard to justify.
TPAC--The list of needed improvements for Mod 2 was lengthy and touched on ainost every aspect of the code.
If the elements of the list are necessary, Mod 2 will be a new code.
It is hard for me to believe that we know so little.
Winkler (from Germany), however, says that TRAC /PF1 Modl is not doing the job. His ICAP effort will put all calculations on hold until TRAC is fixed. It seems that a Mod 2 has been developed by LANL i
for DOE for more than two years and is now in use at Savanah River and the Foreign participants want it. One 'can hardly blame them. Maybe its time DOE picked up a little more of the cost and carried on with TRAC-PF1/ Mod 2.
One has to wonder why eight man-years are still needed if the code is in use. Why not just release it to the Foreigners for use at their own risk?
The need for discipline in the code development and improvement process became very apparent.
There needs to be a specification for a standard FORTRAN developed by NRC and imposed on all who are involved.
The responsibility for changes and their implementation needs to be in one place. During discussions, one of the participants from England wanted to I
know why LANL and INEL separately develop cures for some of the code problems.
In the face of decreasing resources, one has to ask the same question.
Belgium--The Belgians had a great deal of trouble with code steady state initialization.
They were using RELAP5. Here most of their problems were in obtaining a consistent set of plant data acceptable for calculations. A good steady state calculation, however, does not guarantee a good transient calculation. To reproduce the dynamic characteristics, one must work with a number of adjustable parameters that are unimportant to the steady state.
This is another testimonial to the need for good plant transient data. One has to wonder what happened to the EPRI program in this area and why more use isn't made of a plant "flight recorder". With presently available PC data acquisition capability, an EPRI-type flight recorder is a fairly cheap device (probably under $25,000).
P. A Boehnert November 14, 1987 Finland--A number of machine dependency problems were encountered with RELAP5/ Mod 2 Version 36.05. The code was found to be essentially unusable.
An appeal was made for a pre-processor to help with initialization.
FRG--The FRG ICAP participation includes Siemens and GRS.
It is interesting that the US indostn is r.ot participating whereas the Germans are. The only US industry evident at the meeting was a company owned by Siemens.
RELAPS/ Hod 2 is now the licensing tool of KWU. A'RELAP-like code is used by GRS for licensing.
The FRG objectives have changed to reflect the licensing use of RELAP5.
They do not w' ant to use TRAC anymore and will do so only to carry out commitments made as part of the 2D/3D and
~
ICAP Programs. The NRC ought to request that the Germans carry out the CSAU methodology on RELAP5 as a part of their contribution to ICAP.
KWU has modified RELAPS to handle upper plenum injection and downcomer multi-dimensional characteristics.
The USNRC should request the code with the modifications in it for their own use.
The mods appear to be very sensible approaches to rather difficult problem 4.
With these modifica-tions, a 300 node calculation of a 42 second tr ansient in a 1300 MWe.GPWR took 2-3 hours. This is to be contrasted with 50-60 hours for an equiv-alent TRAC calculation. The Gennans calcualated oscillatory behavior everywhere in contrast to data being smooth. This is probably due to a combination bad numerics and poor condensation modeling. ISPRA studies showed that the time step had to be reduced by a factor of ten to get rid of the oscillations.
This means a possible factor of ten increase running tine (500 hours0.00579 days <br />0.139 hours <br />8.267196e-4 weeks <br />1.9025e-4 months <br />?).
Italy--The Italians are using RELAP5/ Mod 2 with added neutron kinetics to handle ATUS type transients.
They found the us sal problems with the code. They made a special plea for help with d aling with "user dependant uncertainties". Here user dependant uncertaint es include nodalization and time step magnitude, i
Japan--The Japanese studies reveiled similar problems to those nuntioned above. They are changing the reflood model and using the EPP.I correlation j
for void fraction to improve the present model in TRAC. Their efforts with condensation modeling show some of the same code developer tendencies we see in the US.
The Japanese had better be careful or they will fall into the same trap of code dependent physics that has troubled the US i
effort for the past ten years, i
Spain--The Spanish found troubles with the steam generator modeling, pressures on the secondary side were underpredicted.
Sweden--The Swedes were diusturbed by the number and frequency of appearance of new versions of the codes. They are using RELAP5 as a licensing tool and TRAC for plant transient calculatons. They would like j
to see the codes frozen until they can establish uncertainties for l
licensing.
.....s P. A. Boehnert November 14, 1987 Switzerland--The Swiss are using RELAP5 and TRAC-BWR. They noted that; they have a very cooperative relationship with INEL (does this mean lots of free consulting on code problems?).
They noted the usual litany of code problems.
United Kingdom--The English are using TRAC-PF1/tiodl for their calculations within the ICAP effort. There are others, however, who favor.RELAP5.
They found finite difference problems that were severe enough' to stop the assessment process.
These problems only became evident when'they looked at a multi-dimensional downcomer.
One cann'ot help but aonder what else is impacted by these problems.
It should be noted that this error does not become apparent when the code is used within guidelines developed through use and code adjustment (dials).
The English would like very much to have the U.S. develop and release TRAC Mod 2.
It turns out that they have four years to assess it as they will not make licensing calculations until 1991 or so. Given that, as noted by Zuber, a code needs 4-5 years to mature, their view is understandable.
If they feel strongly about the flod2 version, perhaps they should help DOE pay for it.
It is not clear that the code will be available in a timely enough manner to be useful to NRC.
ISPRA--ISPRA wanted a new version of RELAP5 and became impatient so they developed their own.
The work at ISPRA uncovered the very unsettling result that the code is extremely sensitive to time step reduction. When they reduced it below the automatic set point, a lot of the code calculated oscillaticas disappeared.
I comented on this above where I noted that calculations of 50 hours5.787037e-4 days <br />0.0139 hours <br />8.267196e-5 weeks <br />1.9025e-5 months <br /> might become 500 hours0.00579 days <br />0.139 hours <br />8.267196e-4 weeks <br />1.9025e-4 months <br />.
They also found tha; large changes in calculated results could occur as a result of small chad,ges in nodalization.
l
_