ML20235E447
| ML20235E447 | |
| Person / Time | |
|---|---|
| Issue date: | 05/21/1987 |
| From: | Beckjord E NRC OFFICE OF NUCLEAR REGULATORY RESEARCH (RES) |
| To: | Kastenberg W CALIFORNIA, UNIV. OF, LOS ANGELES, CA |
| Shared Package | |
| ML20235E440 | List: |
| References | |
| FOIA-87-371, RTR-NUREG-1150 NUDOCS 8707110065 | |
| Download: ML20235E447 (5) | |
Text
l
- Shu, M
o,,
UNITED STATES 8
NUCLEAR REGULATORY COMMISSION o
p WASHINGTON, D. C. 20555 5
4 MAY 21 1987 '
1 Dr. W. E. Kastenberg i
Professor arid Chairman Mechanical, Aerospace and j
Nuclear Engineering Department School of Engineering and' Applied Science
)
University of. California, Los Angeles.
Los Angeles, CA 90024 l
l
Dear Bill:
l 1150 Review You may have seen Bob Budnitz' letter on 1150.
I am sending it in the event that you have not seen it.
I note his comment on natural events, best estimates, and uncertainty.
j I enjoyed the symposium for Dave Okrent last week, and thank you for
.l the invitation.
j Yours truly, e
W
.Eric S. Beckjord, Director Office of Nuclear Regulatory Research '
Enclosure:
3 April 1987 ltr from Budnitz to Kouts
- i v.
fh71 g 5 870707 JORDAN 87-371 PDR I
e p
FutureReaurces Amciates,Inc.
A=%
2000 Center Street Suite 418 Berkeley, CA 94704 415-526-5111 s"~.
?
g
(_ y ?
- 3 April 1987
()f s
e c,
/
v.
/-
/.
1 i
l Dr. Herbert J. Kouts
{
Department of Nuclear Energy Brookhaven National Laboratory Upton, New York 11973
Dear Herb:
J This short note is a response to your request, transmitted to me yesterday from Dr. M. Khatib Rahbar, to provide any comments I might have about the NUREG-1150 methodology. I will be unable to come to Washington for the meeting you will be chairing on April 20-22, so I am providing brief comments by letter instead.
As you know, I was a member of the
- authors' team" for NUREG-1150.
Although I attended many of the meetings beginning in late 1985 in which the overall NUREG il50 reoort outline was laid out I was not a participant in discussions about the methodology used for either the systems analysis, the
~
phenomenological analysis, or the consequences analysis. The NUREG-1150 t'lanning meetings that I attended mostly covered the questions of what the report would cover, what would be its chapter by chapter format, how it would present its results to non expert decision-makers, and how the layers of documentation (main report, appendices, back-up reports) would be structured.
l My own contribution was to write the NUREG 1150 chapter on risk-reduction potential (based largely on Sandia work), much of the chapter on extrapola-tion of NUREG ll50 results from the five reference plants to the larger population of plants, and part of the chapter on regulatory applications.
I believe that the overall methodology used in NUREG-1150 is basically sound.
I The systems analysis done for each of the reference plants was only an approximate, rather than a full scope, reanalysis ---- specifically, the earlier PRA studies done for each of the reference plants were used as points-of departure, and modifications to each of them were accomplished
)
without revisiting the whole analysis in gory detail. This approach, while susceptible to some uncertainty,is in my view an appropriate approach for these five reference plants, especially considering that there are large and irreducible uncertainties in the NUREG ll50 results in any event, arising from modelling problems and from data sets that are only approximately applicable.
I
l I also believe that, given the lack of knowledg
{
phenomena in the core melt, vessel-penetration ie about many of th(
1 ment-failure parts of the NUREG Il50 analysis th, n containment, and c t
a reasonablc one:
differentiated a large number of issues and st dithey const j
analysis and by using expert judgment.
ent event trees, i
u ed them by sensitivity the "Delphi" method of polling " experts
- I will not comment much he guidance from experts in such methods but it su which I know has been so controver-j of wisdom on the technical issues addressed would have been with rely captured a wide range that, for a few of the phenomenclogical i
. In my view one weakness was not necessarily represent either a broad enough sssues, the "exper,ts" co expertise at enough depth. However,I believe thatpectrum of opinion o that this approach captured are probabl y not terribly either understated orthe ra this aspect of NUREG 1150 is from peoverstated --- ani griping I've heard about j
Iny.111 and are griping about its methodology iopic who don't like NURE nstead.
j Two other aspects of the NUREG-l!50 anal j(
initiators were considered. To be specific iTh
,{
1) n a ly important accident hurricanes, tornadoes, floods, and other so
, nternal fires, earthquakes, u!
group been found in almost every fulbscope PRA j
s act that these have as a to overall risk indices. Leaving them y
out makes no more sense to me thanto be transients. Furthermore, and despite mishortening the pt As and only including been fully feasible ---- the excuse for theimethodo!ogy understandings to the contrary, the 1
h.
enough in hand to have terms of the cost of including them r omission was usually given in technical insights obtained from NUREG
. I can only state my own belief that the studied for two of the reference plants (Surry a d liSO may be seriousiv flawM by this n t ators are now being the NUREG il50 sixth plant (LaSalle)n will have a full externalinitiatorseach Bottom) j analysis when it is completed under RMIEP useful and important, does not make up for the f i
current NUREG il50 report may be seriously fl w dThis fut n,
act that in my view the initiators.
a by its omission of these
';~
e
- 2) The second aspect that always botheaed just published, is how to present the ~~rangesapproac a--""
s actio~ n in the ver: ion as estimate values" for the key risk indicators (coreof uncertainty" and the "o release fractions, offsite consequences etn) melt frequency, radionuclides f
t 2
j i
l 1
I I
My physics training, in which I performed compl'cated experirhental measure-i ments involving teams of a dozen or mare, has convinced me that the best l
appicach in a large team effort like this is as follows: the taost knowledy-able memoers of any team of analysts should make it their b.2siness to underttand everythine that enters into the " bottom line' results, includ(ng the l
analysis methods, the data used, the modelling approaches, and how accertain-.
j ties are linked and correlated. Then these individuals (ust' ally only one or j
i l
iwo out of a large team, pmumably the ' team leaders") should figure cut 1
l
.vhat is their best estirm for the numeric 61 vtJues of these bottom 41ne
~
results, a'id why; and what is their 'best estimate for the vneertai'ity dt#jlv_-
Ijan, and why. This knowledge thould be displayed in at:mericM Term (tables or figures) to convey to the render ahele individauls' best itgL_9f.Jrow!edu about the numbers being measured or analyzed.
In my view, the responsible parties ip any complu endeaver owe their sponsors end the technical comrdunity nothing sess than this. Thereforet I-(
was never in agreement with tiae approach of dispitying only "ranes' for the key bottom-line results of NUREG-1350 - -- I belit.ved during our auttiers' team meetings, and believe today, that there is always a curve that the informed analyst can draw that represents his " state of knowledge" about the parameter (s) being measured or analyzed. Sometimes this curve may be a flat distribution, meaning that the analyst believes that there is an equal chance that the true result lles anywhere within the flat region; but this is care, and if it is actually the analyst's best judgment it conveys impoTtant.information.
In my view,it is never correct to present a
- range" without some indic.ation by the inalyst as to what the distribuiton of knowledge looks like lagks like within the range.
These two problems are my main bones to pick with the NUREG 1150 report.
I am plea *,ed that 1/ou have asked for, so that I could write down, just what they are. Notwithstanding these issues,1 believe that the methodology used was reasonable, even if I don't always agree with the details of svery part of the analysis. (In something this complicated, no single individual besides the main authors could agree wit,5 all of the details!).
I hope that this discussion provides useful insights to you in your effvrt to 4
understand the strengths and weaknesses of the NUREG 1150 methodolesy. I could write more about several specific issues but I have chosen not to dilute my niain message with lesser comments.
Despite my criticisms described above, I think it is important to go on record with my opinion that NUREG 1150 represents a msior rivanpa in documenting i
our understanding of reactor safety Irriportant irisights hhv9 been 6eveloped in this study which have not only taught us new things ubout reactor safety but have pointed out where additional knowledge is neced to reduce remain ing uncertainties further. These gr.ps in our underst6ndics exist in every arca studicd: the systeras analysis, the phenomenological analysis both in-vess,1 and ex vessel, and the consequences analysis. My opinion is that
mE.im.E.,',,
t-h"M
7.
I NUREG-il50 and its supporting rcports have been of enurmous help in pointing out what we don't yet know. (Furthermore, the insights buried in the cepporting reports haven't begun to be digested yet). When a maDr I
effoil like this is complefed, it is almdst always a big step forward, and
)
certainly thh was true for NUREG.ll50 f'
Matt important4, and despite its remaining uncertainties, the study' hts confirmed the key insf3 ts gained from the ripacering WASH-1400 inalysis 'of h
3 13 years ago: that core melt frequencies are quite low, that consequences cf most core celts are fit short of t;atastrcphic, and tnat overall societal risks are well below the %a(Sty goals" thdt scelety demands for other coniparable j
risky industrial endeavor 6.
That completes ray thoughts. If thee are any questione,1: lease don't hesitate to contact me.
i With warmest reguds, Sif,eerely yours.
- 1
)
j E%ert J. Budnitz I
'l cc:
L. Murphy, NRC i
i 4
m
-c