ML20212K303
| ML20212K303 | |
| Person / Time | |
|---|---|
| Issue date: | 08/14/1986 |
| From: | Sege G Office of Nuclear Reactor Regulation |
| To: | Niyogi P NRC OFFICE OF NUCLEAR REGULATORY RESEARCH (RES) |
| Shared Package | |
| ML20212C579 | List: |
| References | |
| NUDOCS 8608200088 | |
| Download: ML20212K303 (20) | |
Text
~
AUG 261986 l [
UNITED STATES F,
NUCLEAR REGULATORY COMMISSION o
'*h~
WASHINGTON, D. C. 20555
- s.,..../
AUG 14 95 1
NOTE T0:
P.K. Niyogi, RES FROM:
George Sege, NRR
SUBJECT:
COMMENTS ON " CONTAINMENT PERFORMANCE DESIGN OBJECTIVE WORKSHOP" (DRAFT DATED JLLY 18,1986)
At your request, I have reviewed this Brookhaven report. The enclosed markup shows my comments.
You will note that, with one exception, my comments are stated in specific language as suggested revisions.
In accordance with the NRC objective of protecting the independence of the Brookhaven report by refraining from commenting on the way the substantive highlights of the Workshop discussions are selected and captured, I have limited my review to scope and clarity.
The reasons for some of the more significant comments marked are as follows:
Pages y and 3-1: The title of Sections 3 and 4 are too similar to indicate the distinction between their respective contents.
Insertion of " Issues of" before the present title of Section 3 would clarify the distinction.
Page 1-1, lines 3-4: Since sponsorship and preparations for the Workshop were a joint RES-NRR undertaking, the sponsorship should be ascribed either to the NRC (without designation of any particular office) or jointly to RES and NRR, rather than to RES alone.
Page 2-2, first 41 lines: The language here suggests a view that a CPD0 should be promulgated as a sop to the public.
/[
I doubt that the authors intended to create that impression.
Brookhaven should review the language of this passagf< to assure that it clearly reflects the meaning intended to be conveyed.
Page 4-1, second, third, and fourth sentences of middle paragraph:
Deletion of these sentences would reduce somewhat the imbalance J.
of Section 4, in which the first 2 of 4 pages summarize descrip-tive information provided to the participants and a part of the third page explains that information, and less than half the text addresses the discussion content. The deleted sentences are urnecessary and partly repetitious.
The fourth sentence, moreover, tends to confuse, rather than clarify, the scope differences between Sections 2 and 4.
l l
Xh
%s7 %obov
o AUG 14 986 1
Page 4-2, end of Option 5 item: The definition of " gross failure" is a significant part of this option.
Its inclusion is needed to avoid a significant incompleteness of the description, and also to avoid the apparent inconsistency of including only three criteria under a "Four Criteria" heading.
Page 4-3, second paragraph, last line:
Insertion of "at least" where shown clarifies the intent of the mechanistic option.
Page 4-3, last 7 lines: The editorial changes marked would undo the incorrect characterization of the last 55% of the two-page
" Discussion of CPD0 Options" as a " summary" of the first 45%.
Page 4-4, line 11: This discussion applies only to pressure suppression containments, not to all BWRs (e.g., not Dresden I or Big Rock Point).
We were pleased to have had the opportunity to review this draft. Please call if you need further assistance.
George Sege
Enclosure:
Marked-up Draft cc:
T. Speis B. Sheron Z. Rosztoczy L. Soffer 9
ema.-a
~
e i c,
a*
y acdny by G.Seye, eliges CONTAINMENT PERFORMANCE DESIGN OBJECTIVE WORLSHOP Draft dated July 18, 1986
?
T-t Department of Nuclear Energy Brookhaven National Laboratcry Upton, New York 11973 r
h h
Prepared for U.S. Nuclear Regulatory Comm:ssion Washington, DC 20555 Contreet No. DE-ACO2-76CHOOl6 FIN A-3298 grot.gs/OH5-
n,
- , /
111 GISISOCI The " Containment Perf ormance Design Object 1.e Wori: shop" was de-signed to obtain a broad range of 6:nowledgeable.tews concerning the issues in the development and implementation of a containment perfor-ria= i gn objective (CFDO).
It was a discus.s:=n wori: shop, involv-menea ing invited I;nowledgeable persons representing a Orcad range of view-points. and drawn f rom utilities, vendors, arch: e:t engineers. unt-
.versities, national l abor at or i es, and public inte-est groups.
The participants were requested to review bact'groun: Information concern-ing the safety goals and their status, a descri: t on of CPDO options selected for evaluation, an outline of an imolers-tation approach and recogni:ed issues of CPDD structure and impleme-ation.
The general objective of the wori: shop was to generate inf orcation that could be used in the NRC's study and decision process co :erning the f ormula-tion of a containment perf ormance design object:.e.
t t
I l
l l
i
. w--
F v
TAE<LE OF CONTENTS Page iii ABSTRACT..................;.---**************************
vii EXECUTIVE
SUMMARY
1-1 1.
INTRODUCTION................................................
2.
QUANTITATIVE CONTAINMENT PERFORMANCE DESIGN ~BJECTIVE VS 2-1 NO ACTION...................................................
l 5 3ME '.
cc~
34 CONTAINMENT PERFORMANCE DESIGN OBJECTIVE F0F.-JLATION AND
/
3-1 IMPLEMENTATION..............................................
4-1 i
4.
CONT AINMENT PERFORMANCE DESIGN OBJECTIVE OPT:3NS............
i Appendix A:
Inf ormation f or the Participants in the Containment Perf ormance Design Sjective A-1 Workshop.......................................
?
k t
.e e
C-r.
I via EXECUTIVE
SUMMARY
In this section, the main areas of consensus that emerged from the Wori: shop are summari:ed.
": participants considered the containment.to be a necessary element of a def ense-in-depth saf ety philosc;hy.
2.
There appeared to be general support f or minimum containment performance.
In addition, a maj ority of the participants agreed in principle to a CPDO provided the concerns related to f ormulation and implementation could be reasonably resolved.
3.
In the short term, the saf ety goals should include a statement that a CPDO will be developed f or une by the NRC staf f in the future.
4.
The participants appeared to favor a quantitative and high mechanistic content to any statement on containment perf ormance.
However, it was realized that a CPDO could not be all mechanistic in nature and would have to include some probabilistic elements.
5.
There is a clear need for the CPDO and the aspects pertaining to it to be well defined and unambiguous.
t Based on the 6.
No new option was proposed by the participa.ts.
various CPDO options, desirable characteristics were discussed and it appears that there was a general ag-eement on what a quantifiable CPDO should be.
7.
Consideratton of external events should be included in the f ormulation of a CPDO.
8.
Application of a CPDO should include consideration of both negative and positive aspects of human performance.
'W --......
1-1 1.
INTRODUCTION The ',' Containment Perf ormance Desi gn Obj ects.e Workshop" held at Harpers Ferry. West Virginia on May 12-13. 1086. was sponsored by the U.S.
Nuclear Regulatory Commission 71 O'fi;; cf m vm; im 7
^
nc-rr- ;t.
Brookhaven National Laboratory (BNL) was responsible f or pro-viding the arrangements f or the Workshop.
The pa-ticipants were drawn
{
from utilities, vendors, architect engineers. um s ver si ti es. nati onal laboratories, and public interest groups.
To assist in the ef ficient
{
conduct of the Workshop, the participants recei ved in advance an NRC input package, which is reproduced in Appendi:: A.
Dr. H.
J.
Kouts of BNL was chairman of the Workshop and members of the NRC and BNL staf f served as resource persons in the discussions (ref er to Section III of Appendix A).
The discussion guidelines and age".f a f or the Workshop i
are given in Sections VII and V appendi:. A respe:tively.
,)'
[
e The Workshop was held to bring together sixteen selected partici-pants, with diverse backgrounds and with special knouledge of various i
areas pertaining to nuclear power plants in gere'al and containments in particular.
It was a discussion workshop ir. ahich the partici-pants' views were obtained on CPDO options and e9 issues, the ones presented to the Workshop (ref er to Section VI*! of Append 1:: A) and any new ones that emerged from the discussion.
ArE-as of Consensus as well as areas of disagreement were identified tc; ether with the key issues and arguments involved in significant d: sagr eement s.
The gen-eral objective of the workshop was to generate inf ormation that could be used in the NRC's study and decision process concerning the f ormu-lation of a containment perf ormance design objective.
The information sought concerned the f ollowing broad quest i ons :
~
. ~ _....
s.
1-0 i
1.
What are then merits and drawbacks of adding to a CPDO the safety goals? 1)o the merits outweigh the drai$ backs?
2.
What are the principal issues in CPDD f ormulation and implemen-tetion?
How should they be resolved?
(The issues considered were those presented to the participants anc also any additional issues identified by the participants. )
b Is there a CPDD f ormulation approach that s*.culd be considered 3.
beside the presented options?
Is any of the options clearly L
superior?
Is any option clearly impractica. or inferior, so that it should be. dropped f rom f urther cons:deration?
As background to the Workshop, the partict:: ants were requested to review an NRC input package (refer to Appendix A) which included:
?
(a) background inf ormation concerning the saf ety goals and their sta-tus; (b) description of some CPDD options selected f or evaluation, i
i with a discussion of pros and cons; (c) outline of an implementation 6
approach; (d) identification of recognized issues of CFDO structure 1
and implementation; and (e) a detailed agenda, together with guide-lines with respect to discussion objectives an scope.
In addition, the proposed " Policy Statement on Saf ety Goals f er the Operation of i
Nuclear Power Plants," NRC. March 1983. Recomme.ded Revisions to the Policy Statement, EDO, February 14, 1986 and A~:.5 letters of March 19, 1986 :.nd April 15. 1986 were also provided as ee4erence materi al.
Participation in the workshcp discussions was limited to the in-vited participants, but the wor 6: shop was open to the public f or atten-dance as observers, and limited time periods ws e set aside for any comments from the observers.
The wori shop was noticed in the Federal Register.
A verbatim record was kept and copies are available f or a
b.w:
> - ( ',s
- e.
e 1 ~.
moderate fee upon request to BNL.
This report escument he hich-Tkt 1l. o r s A bfv t lights of the Workshop.
CC ;!
t'-
-t c; broa: questions are dis-Y
/
cussed in the f ollowing three sections.
e h
j 1
1
)
v 4
i e
3 i
e l
. - - * -. -. - ~
,,-,_.,-*e --*
-swaw- - -wrerw_y-<-mv----=w-4
--rw're--
--N***w9
-d-**v W-**-+a-*WP-FPe*--wr'-*v'er-v'E e*
w-*pt
---*9--
-wT7**
f
b._ m
.~_
2-1 2.
DUANTITATIVE CONT AINMENT PERFORMANCE DESIGN CBJECTIVE VS NO ACTION The main questions that the partic1 pants we e requested to ad-dress under this issue were:
"What are the merits and drawbacks of addin? a CPDO to the saf ety
- goals?
Do the merits outweigh the drawbacks?"
The participants were divided rather evenly between whether or not a quantitative CPDD should be added to the saf ety goals with per-haps a slight majority in f avor of adding a CPDO.
In addition, a sig-
[
nificant number of the participants f elt that a statement should be included in the saf ety goals which identified the need f or a CPDO but which also indicated that one could not at prese.-t be f ormulated and that work was in progress to define a CPDO.
In the f ollowing para-t.
graphs the merits and drawbacks of adding a CPD0 to the saf ety goals I
are briefly summarized.
f The participants identified several advantages of adding a CPDO
[
to tht* safety goals.
A major concern was that it might be possible a
for a particular nuclear installation to meet the top level safety goals related to of f site health ef f ects without a containment building by having a very low core damage f requency.
Most of the participants felt that a containment building was an essentia. part of e defense-in-depth safety philosophy and that the additt o. of a quantitative CPDD to the safety goal would provide some measu~e of the required effectiveness of the structure.
The need for scre measure of contain-ment ef f ectiveness was also considered importar.t by other participants
~~
who thought that the health ef f ects goals were tco f ar removed from the containment function and that an intermediate measure (namely a CPDO) was therefore needed.
Other participants f elt that there was a
- v
. ~ -.
j.
f 2-2
'need to buildup public confidence (in the afterr.ath of the Chernobyl i
accident) and that it was important to have a saf ety goal which pro-g vided ior a 1ow conditional probabi1ity of a seve-e f a ssion product L
O release given a core melt accident and that this would in turn help to reassure the public.
It was also noted by some of the participants that a CPDO would be of limited use f or operat:ng plants because the containment systems are in place and it would te dif ficult. based on cost-benef i t arguments, to require major desigt modifications to those
{
1
~ [
existing systems.
However, these participants did consider that a CPDO would be a useful goal for the design of f uture plants.
Another participant expressed an alternate view in which he noted that the existing containments were not currently measu-sng up to their perfor-mance requirements (NUREG-4220 was ref erenced to support this conclu-e sion) and that improvements were necessary to natigate small releases and prevent structural failure.
This participant was theref ore very I
supportive of a CPDD and felt that a higher CFOO should be speci fied b
f or f uture plants.
In summary, the most significant benefit of a CPDD appears to be to specifically define ef f ective.ess of the containment structure (and systems) rather than to rely or the proposed health ef f ects goals, which were considered to be to: far removed from the containment function.
The participants identified several prot". ems and objections to the eddition of a CFDO to the safety goal s.
Cne cf the main objec-tions was that the core melt frequency and health ?f f ect goals already imply a level of containment performance and the-ef ore the addition of a CPDO would duplicate the e::isting goals.
A go:d deal of concern was also raised as to how a CFDO could be developed and implemented.
For
=
C-0 e:: amp l e, a n a BWR how would the corrpl e:: interre;ationships between containment failure, suppression pool, and react:r building (all of which have an important influence on the quantaty of fission products ultimately released to the environment) be f actceed into the specifi-cation of'a CPDO.
In addition, participants wh: objected to a CPDD were concerned about developing too many goals a id the potential for overlapping of the various goals.
There was also the concern that a-
~
CPDO could constrain f uture plant designs.
The merits of a qualita-tive rather than a quantitative CPDD were also discussed.
The maj ori-ty of those participants who supported a CPDO felt that it should be quantitative in nature.
The principal reason foe preferring a quanti-tative CPDD was because of the perceived dif ficulties associated with implementation of a qualitative CPDO.
In summary, the major objections to a CPD~ appear to be duplica-tion with e::isting goals and the potential f or inhibiting f uture designs.
The mest significant problem appears to be f ormulataan of a
CPDO that accounts f or the comple::itles and diff erences in containment designs.
i 1
l 4
C-3-1 ISSHcr5 C. F C.
CONTAINMENT PERFORMANCE DESIGN OBJECTIVE FDF.MJLATION AND h
IMPLEMENTATION The main questions that the participants we-e requested to ad-dress under this issue were:
"What are the principal issues in CPDO f ormalation and implemen-tation? How should they be resolved 7" Several additional questions (ref er to Section VII.B of App end t :-
A f or details) related to earthqual:es, human actions. primary vs sec-ondary containment, tradeoffs between containment and plant perfor-mance, importance of a CPDO compared with other saf ety goals and e::isting vs f uture plant considerations were also presented tc the participants as discussion guidelines.
In the f ollowing paragraphs issues related to the formulation and implementation are discussed.
A significant number of participants who opposed a CPDO did so because of concerns related to its f ormulation and implementation.
Several very important issues were raised f or which the path to reso-0 lution was not well defined.
A major issue related to whether a CPDO could be made fle::ible enough to cover a range of accident sequences e
and plant designs.
For e::ampl e, how does one develcp a CPDD f or acci-dents that involve loss of containment heat r ee.c eal (CHR) in which containment failure might precede core melt?
Should the CPDO be the same for e::isting and future plants?
If a particular plant has a very remote site or relatively low core melt frequency should it be subject to the same CPDD as a site with a very high suerounding popul ation or core melt frequency?
In general, the participants agreed that a ccm-prehensive CPDO could be developed (refer to Section 4) which ad-dressed these issues.
However, the CPDO would have to recogns:e
-s.
-b-yl
- 3-2 containment function rather than be simply limited to loss of contain-ment integrity.
Containment function includes. f or e::ampl e, the ef-fectiveness of suppression pool scrubbing and secondary containments as mechanisma f or trapping aerosol fission procacts in a BWR.
Thus.
containment integrity might be lost relatively early in an accident sequence in a BWR with a relatively high probattlity but, provided the
[
fission products pass through the suppression pcol and the reactor building the CPDD could presumably be met."
Ancther e>: ample noted was I
a station blackout sequence in a PWR in which t*.e probability of con-tainment f ailure is very high.
However, if f a:1ure occurs af ter a long time the of f site consequences would be ve y low.
This implies that the CPDD should recognice all of these different accident se-quences and plant designs and be both mechanistic and probabilistic in nature.
In addition, the participants f elt that a more stringent CPDO This was required f or f uture plants than f or the e::1 sting plants.
simply recogni:es that it is easier to make design changes while the g
plant is on the drawing board than af ter it is built and operating.
While the participants f elt that a compre ensive CPDD could be l
developed they were concerned that it could be overly prescriptive and hence inhibit design.
Participants were also concerned as to how one demonstrates that the CPDO has been met.
In addition, our knowledge of containment performance and fission product release is still evolv-ing and subject to change.
Consequently, a h;ghly prescriptive CPDO could also be subject to change.
The particicants f elt that the CPDD should be very stable and thuc it will have te be caref ully f ormulated to prevent it f rom being subject to the changes in our understanding of containment perf ormance.
._.=
c.
to 3 3-3 The participants also felt that accidents :.s ta ated by seismic events should also be considered in the CPDO.
should, however. be noted that the more probable seismic events sir
- y cause the f ailure of systems and components that result in accide-- sequences similar to those caused by internally initiated events.
I other words, the more probable seismic events result in containment f ailure modes that are si mi l ar to those caused by accidents initiated t v internal events.
I There are, of course, very unlil:ely seismic eve-s that could directly cause structural failure of the containment or. ors of containment integrity and thus result in unique containment f ailure modes.
How-ever, thre participants felt that these low protability very severe seismic events should be considered in relationship to the ef f ect that the event itself would have on the surrounding gepulation.
Therefore, a limit must be set as to how low in probabilit. one goes when formu-lating the CPDO.
Finally, the participants fsilt that the ne;ative and positive aspects of human performance must be considere: in the implementation of a CPDO.
For e:< ampl e, it may be determined t at wetwell venting is a necessary element of a CPDO for a BWR with a. Marl I containment.
Wetwell venting is an intricate maneuver which =:11 require the opera-tors to tal:e many actions that they are current *.y not required to per-form.
Under these difficult circumstances the egetive actions that might be perf ormed by the operators must be co tidered and factored into the implementaticn of the CPDO.
t 4-1 4.
CONTAINMENT PERFORMANCE DESIGN OE4JECTIVE 0:-*ONS The main questions that the participants ws-e requested to ad-dress under this issue were:
"1s there a CFDO f ormulation approach that should be considered besides the presented (refer to Appendi:: A) op : ens?
Is any of the options (in its present f orm or in modi f ied f o-e s clearly superior?
or inferior, so that it should be Is any option clearly impractical dropped f rom f urther consideration? What are tr.e reasons f or the judgment?"
Several addi ti onal questions (refer to Se: tion VII.C of Appendix a for details) were also discussed by the participants.
As an intro-
_ a,
--..-.-. ~ -.....
}_ duction to the discussio y on this issue a presentation was given by in! ection VIII of C,
S
?
Dr.
K.
Shiu (ENL) on t various options presented\\/
/
Appendix A.
Four CP O options were presented tc he Workshop partici-s t
which helpid to f ocus the subsequent dis ussion by the partici-
- pants,
/
N
/
In add. tion \\to the f our CFDO options,/the possibility of no-y pants.
/
K (refer to Section P
q, alitative options were also discussed
--.-...,_......)
action or four CPD0 cations are briefly 2).{Inthefolloingparagraphsthe described.
A more detailed description of the cations is given in Section VIII of Append 1r A.
Option 1:
Modified ACRS Approach Conditional prcbability of a large s: ale release, given a large scale core melt:
for e::isting plants:
Not mor e tr.s-O.1 for f uture plents:
As much belea.1 as cost b e n'e f i t
~
trade-of f s justif y.
Target:
0.01
4 4-2 Option 4:
Fle::ibl e Approach Frequency of large scale release f rom cor.tainment:
not
-6 more than 10 j gy,
~
Condi tional probability of early containment failure, given a l arge-scal e core mel t:
nct more than 0.1.
Option 5:
Four Criteria Containment should survive without gross f ailure all but the very unlikely of severe accidents.
The aggregate
-6 frequency of gross f ailure should not e::ceed 10
/RY.
- Containment leal. age and venting during and f ollowing severe accidents should be limited to specified levels.
The probability of occurrence of events which exceed these levels, but are of lesser severity than gross f ailure, should be low:
they should not exceed 10~ /RY.
v y
)[
Emphasis on def ensefinhdepth should be retained.
The
}{Jh&051 conditional probability of gross containment f ailure or if - g, t excessive leakage, given core melt should not exceed 0.1.
Option 6:
Mechanistic
- No substantial leel: age for "24" hours No gross structural or major penetratien f ailure f or " 7 2
hours
- Ability to ititervene within "72" hours Maximum practacal assurance that systems on which containment ef f ectiveness depends are always operational.
Note:
Ouotation marls around figures are intended to emphasi:e that the figures a-e illustrative only and that any values to be selected for specified
- j' w,
p, 4-0 consideration will require further studv.
ELSEWillGU_TO_CEDQ_QEllGUE mp+ ion 1 basically attempts to limit the conca tional probability of a severe release given a core melt accident.
Some of the partict-pants f avored using a "large scale release" because this could be re-lated to containment function rather than loss of containment integri-ty.
However, there were concerns regarding the definition of a "large scale release" and how it would be predicted.
Option 4 is probabilistic in nature and attempts to limit the conditional probability of early containment f ailure and the frequency of a severe release.
Most participants f elt that this option needed to be modified to include more mechanistic elements.
Thus, several participants considered that a mi::ture of option 4 and option 6 was a reasonable compromise f or a CPDO.
Option 6 is at an early stage of development and basically attempts to limit the amount of leakage and T
of /eo S ' late in the accident very sequence.
prevent containment f ailure until L
Option 5 contains some of the elements of the other three options so that it is an attempt to combine probabilistic and mechanistic ele-ments but the f ormulation did not receive wide sapport f rom the par-ticipants.
However, the participants did f avor cevelopment of a CFDO 5
that includes elements of options 4 and 6ghen option 5 1s moving in the right direction.
Obviously more wort will te needed to achieve en !
appropriate balance between the.various. elements of a.C. F D.O..
+u n v' s+ pant wppeared tc ' d(a CF DO that in-r r e - ~q,
\\
s-icluded both probabilistic and mechanistic eleme ts'J
- the par-
)(
"r r. c,.
g w
tacipants were concerned with how compliance with a CFDO would be dem-ma= + r a t ed in the licht of uncertainties in containment loads.
7.
t
,r., FA ;;,...
e
~
4-4
[~D k'-
containment performance. and f i s si on product release.
Some partici-h oc pants felt very strongly that it.e various elements of a CPDD should be defined as "high confidence" le cels as opposed to mean values.
For 1:M g,
e:: amp l e, it would have to be de. :nstrated with*"high. confidence" that a particular containment has a conditional probability of early fall-ure less than a prescribed value.
Participants were divided a to what "high confidence" represe.ted and how it would be determined but clearly some way has to be devt sed to adequately consider uncertain-ties when formulating the CPDO.
Other participants were concerned about the emphasis on containne. : f ailure because this does not recog-P't 55 u r e s vpN U/on f o n %*n m utf3)
V ni:e the uni que el ements of the BW
- in which early containment fail-ure does not automatically mea 7 a severe fission product release.
These participants therefore recommended that " containment failure" be replaced with " loss of containment function."
This would allow BWRs to lose containment integrity ea-ly (perhaps by venting) but, provided the suppression pool and the reartor building were not bypassed, the containment function would be maintained.
An alternative way of e::-
pressing this is to measure pe-f ormance in terms of source term re-duction.
In other words. a s:.ar:e term to containment would be spec t-fied and provided a specified reduction an the source term could be achieved (with or without contaliment failure) then the CFDO could be me t,.
Again, thes,e all simply re: resent possible ways of specifying f7 c ent a i ntren t performance but clea-1y the perticipants felt that some combination of the above eleme7ts would have to form part of the CPDO.
e i!E r, -,.,
.-w
5 I-
+
ATTACHMENT B 1
l l
i
(
j i
i n pm F
,LyubL BROOKHAVEN NATIONAL LABORATORY IIIII ASSOCIATED UNIVERSITIES, INC.
WwJ %wu L Upton. Long Island, New York 11973 (516) 282s 2630 Department of Nuclear Energy FTS 666' October 24, 1986 Mr. Daniel Hirsch Director, Program on Nuclear Policy Adlai E. Stevenson College University of California - Santa Cruz Santa Cruz, CA 95064
Dear Mr. Hirsch:
Thank you for your letter of July 29, 1986, which provided comments on a draft report of the CP00 Workshop.
I have attempted to incorporate your comments and the comments received from other Workshop participants into a revised version of the CPD0 Wo.kshop report.
This revised report has been reviewed by NRC staff and is ready for publication.
I believe that the changes made to the report have addressed the impcrtant matters that you raised in your letter.
Please review the revised report and inform me whether or not you still wish to have your letter and Attachments published with the CPD0 Workshop report.
It is my opinion and the opinion of the Workshop Chairman that given the present form of the report it would be inappropriate and misleading to attach your letter and Attachments.
Please advise us of your decision by November 3,1986.
Otherwise I will assume that it is alright to publish the report without yuur letter and Attachments.
- Regards, Wm. Trevor Pratt, Division Head Safety & Risk Evaluation Division WTP:csc Enc.
cc:
H. J. Kouts (w/o encl.)
R. A. Bari (w/o encl.)
P. K. Niyogi (NRC, w/o encl.)
i ATTACHMEtif C
(E!
31986 NUNIVERSITY OF CALIFORNIA, SANTA CRUZ SERKELEY
- DAVIS. IRVINE. LOS ANCELES
- It!VER51DE
- SAN DtiCO
- SAN FRANCISCO SANTA BARBARA
- SANTA CRUZ YY ADLAI E. STEYENSON COLLEGE SAhTA CRUZ, CALIFORNIA 95064 (408) 429-2833 October 30, 1986 Wm. Trevor Pratt Division Head Safety and Risk Evaluation Division Department of Nuclear Energy 32 Lewis, Bldg. 130 Brookhaven National Laboratory Upton, Long Island, New York 11973
Dear Dr. Pratt:
I am in receipt of your letter of October 24, 1986, requesting that I withdraw my dissent from the final report of the NRC's Containment Performance Design Objective (CPDO) panel.
I cannot in good conscience consent.
None of the items to which I objected in the draft report has been corrected.
Indeed, the few changes in the overall text, in general, further weaken an already weak report. For example, summary item 1 has been altered to suggest containment is necessary only for the " current generation of light-water reactors," a matter with which the majority of the panel was certainly not in agreement. Item 2 has been altered from a majority of the panel supporting a CPD0 to only a few so agreeing, with an extraordinary statement appended that "many also recognized that the containment is only one element" in preventing fission product release in case of accident.
If there were one thing that the majority of the panel agreed upon, it was how essential it is to have a proper functioning containnent.
I ccatinue to dissent from items 3, 6, and 8 in the summar), as well as the representation in the abstract regarding the panel's composition:
( 3) I still vigorously oppose the recommendation that no CPD0 be included in the NRC safety goal, merely a statement that such objectives would be desirable but can t at present be formulated.
I find it incomprehensible, particularly in light of Chernobyl, that the NRC has no containment performance objectives in its safety goals.
(6) It remains incorrect to say no new CPD0 option was proposed.
I proposed such an option, detailed in my minority report, which, unlike the other CPDos, might be acceptable to the public.
(8) I continue to object to consideration in a containment objective of hoped-for positive human intervention in the midst of an accident.
Containments are supposed to be safety features to protect the public
s i
Y~
fL
>(
i i
'when operators make mistakes; it would be folly to rely on proper operator s
s
' action for aJfeature that is designed to protect the public from the operator's mistake's.
s
, And the report still inaccurately describes the representation on the panel, i
(
s Furthermore, I continue to object to the report's omissions, particularly its failure to discuss what was, in my view, the most significant aspects of the entire CPD0 exercise:
(a) that none of the containment objectives, as proposed, would be acceptable to the public if explained in plain English, and (b) that, notwithstanding the unacceptably lax standards unders consideration, U.S. reactors apparently could not meet them, often by adery wide margin (11 of 12 evaluaticns of reactors against the,various proposed goals resulted in failure to meet those goals.] The primary disagreement of the panel was over what to do in light of (a) and (b), with industry representatives arguing that the appropriate response was h, <
to not publish any containment objectives at all, to which I vigorously
(,
dissent.
It remains incomprehensible to me that, in the wake of Chernobyl, there
(
remain no containment performance design objectives for U.S. reactors, particularly in light of the apparent high probability of failure of U.S.
containments in severe accidents (e.g., the NRC's current estimate of 90%
protability of failure for the Mark I).
I believe it would be inappropriate and ndsleading to publish the report without the dissenting comments, so please incade this letter and my earlier letter and attachments, copies of which are enclosed.
A Sin r y,
L Daniel Hirsch Director Program on Nuclear Iblicy s
Enclosures:
letter of July 29, 1986
' (comments on report) (minority report) i s
s UNIVERSITY OF CALIFORNIA, SANTA CRUZ s Raruv. Davis. >RvrNe Los ANcrus RIVERSIDE
- SAN DIECO
- SAN FRANCISCO
(
SAMA BAR3 ARA
- SAMA CRVZ u
ADLAI E. STEVENSON COLLEGE SANTA CRUZ, CALIFORNIA 95064 (408) 429-2833 July 29, 1986 i
Wm. Trevor Pratt i
Division Head Safety and Risk Evaluation Division Department of Nuclear Energy i
Brookhaven National Laboratory Upton, tag Island, New York 11973
Dear Dr. Pratt:
In response to your letter of July 17 requesting comments regarding the draft report of the NRC's Containment Performance Design Objective (CPDO)
Workshop, please find attached my separate comments. Attachment I responds to specific items in the daft report; Attachment 2 is a minority report addressing matters not addressed in detail in the draft report. Please attach both of them, with this cover letter, to the final report when published.
Please note that I do not believe that the draft report is accurate in ascribing consensus among members of the CPDO panel tc all the items listed in the Executive Summary.
I, for one, do not concur with items 3, 6, and 8.
I do not believe that the safety goals should merely include a statement that a containment performance design objective will be developed somet.ime in the future. In light of chernobyl, it is urgent that containment performance standards be established now, not further deferred.
I did not agree with the proposed CPDO options, and did propose my own new options. I felt--and Chernobyl has only strengthened tgis view--that the proposed goals of 10-4 probability of core melt and 10- prcbability of containment failure (the basic structure of the various proposed quantitative safety objectives)-are far too lax and would be totally unacceptable to the public if presented in a form that does not obscure their true meani accidentsof10-gg. 'Ihe proposed " goal" of a probability of core melt per reactor-year translates into approximately a 40%
chance of a core melt in the current population of nuclear plants over their expected lives, even if NRC were prohibited from granting any new reactor licenses (instead of attempting to make granting of such licenses easier), and existing reactors met the goal (which they don't), and external initiators such as earthquakes and sabotage, generally excluded from consideration in the PBAs on which compliance with such safety goals is estimated, could be igncred (which they can't). The proposed core melt goal translates into an even more unacceptable probability of 60-70% chance of core melt if new nuclear plants are not prohibited and a modest number of additional plants are built (approximately doubling the current population.)
s
.o S e 10-1 conditional probability goal for containment failure m a core melt accident did occur (virtually inevitable if the 10-gans that if core melt goal were accepted), a containment failure expectation of 1 in 10 would be considered " acceptable," a kind of Russian roulette gamble I doubt the public would find reassuring if explained in plain English.
I do not believe the American public would accept a safety goal of a 40-70% probability of a core melt accident, nor would they find comforting the fact that current Probabilistic plants are coming in around 3 x 10-pisk Assessments (PRAs) for U.S. nuclear
, i.e.,
falling short of that grossly inadequate goal by a factor of three. Nor would they be comforted that, even excluding earthquakes and sabotage, current severe core damage accidents are occuring at about seven times the suggested safety goal rate (i.e., about once every 1500 reactor-years, or about once every seven human years with the current worldwide population of reactors). The PRAs, thus, appear to underpredict actual risk of core melt by a factor of two and yet nevertheless show reactors failing to meet the suggested core melt goal by a factor of three.
I further doubt the public would accept a containment failure goal of one in ten, nor would they find reassuring that current estimates in61cate failure probabilities as high as 90% for containments such as the Mark I, which thus would fail to meet the proposed containment performance objective by a factor of nine.
Perhaps the most troubling and spectacular aspect of the CPDO. Workshop was the conclusion in the analysis performed for the Workshop by NRC and Brookhaven staff that virtually none of the reactor types analyzea could meet virtually any of the proposed safety goals, and that appears true whether one uses old Reactor Safety Study methodology for analysis or the new " source term" calculations. In twelve evaluations of reactor containment types against various proposed CPD0 options, there were eleven failures to meet the proposed goals, a disheartening indication of the inadequacy of current U.S. containments to prevent major releases of radioactivity in severe reactor accidents.
%e fundamental discussion of the CPD0 Workshop, in my view, thus is not even mentioned in the draft report: that the proposed Containment Performance Design Objectives would indicate as acceptable core melt and containment failure probabilities that would be clearly unacceptable to the public if published in plain English, and that, nevertheless, current U.S.
plants can t mec*. even those inadequate safety goals.
For that reason, a number of the Workshop participants, particularly the industry representatives, argued for publishing no containment performance objectives whatsoever.
I dissented, and continue to do o Sin er ly, Daniel Hirsch Director Program on Nuclear Policy enclosures: Attachment 1.
Ccanents on Draft Report.
Minority report
.i Comments g Daniel Hirsch to Draft Report g the _NRC Containment Performance Design Ob]ective Workshop pa,,ge carsnent vii I do not agree with items 3, 6, and 8.
It is thus incorrect to identify them as " areas of consensus."
- 3. I believe that the safety goals should include containment performance design objectives now. It is unconscionable that after 5 years of considering quantitative safety goals, the NRC still has none aside from the vaguest, qualitative social risk statements (e.g. " nuclear should be no more dangerous than other major societal risks").
It is unconscionable, particularly in light of Chernobyl (which was occurring as our Containment Workshop met), that at this late date conteinments still aren't l
required to be designed to withstand the challenge of a full core melt (the very types of accidents where they are needed), and that no design requirements exist to assure the containment does what it is needed to do--remain intact in case of severe accident. It appears to me an abdication of responsibility to propose at this 2
late date, as this draft report does, that the proposed safety goals include no containment performance objective but merely a statement that such an objective "will be developed for use by the NRC staff in the future."
How can the public be assured that reactors are sufficiently safe if the regulatory agency has no standard for measuring their safety, particularly in regards the largest contributor to uncertainty about severe accidents-performance of the containment? How can reactor designers know how to design their containments if there is no standard against which to judge them?
How can utilities have assurance additicnal requirements will not be added later to strengthen containments if there is no clear requirement now to assure adequate containment? Most importantly, how can public health and safety be protected when the regulatory agency and regulated industry have essentially no standard for one 1
of the nest inportant components of reactor safety?
l 6.
It is not true that no new option was proposed by the participants, nor that there was agreement on what a quantifiable CPDO should be. Both I and Sholly indicated we were not happy with the numbers being proposed for the quantitative goals, feeling they were unacceptably high. I indicated I did not believe it possible the public would accept a goal of a one in ten chance of catastrophic containment failure with major release of radioactivity in case of severe accident, nor for that matter would they accept a 40-70% chance of severe accident as was being proposed with the core melt probability. I proposed as an acceptable goal a containment failure probability of one in a 4
hundred and high confidence that the true probability wasn't higher (defined as a 95% confidence level).. I proposed that the 1
1 I
\\-
probability of a core melt accident occurring somewhere in the U.S. during a person's lifetime should likewise be one in a hundred, again with high confidence. (The Workshop participants recognized that, even with statistical high confioence, the probability that something might be overlooked requires redundancy-both a low probability of core melt and a low probability of containment failure in an accident. Otherwise, there' is no redundancy, no defense-in-depth.)
I also proposed that the quantitative goals be written in a form that does not obscure their true meaning fog members of the general public.
Core melt frequency of 10- per reactor-year, as I pointed out at the Workshop, translates to about a 50-50 chance of a core melt over the next forty years or so, excluding sabotage. Public acceptance of a safety goal should not be predicated upon obscurir.g what it really means. If a 50-50 chance of full core melt is felt to be an acceptable goal by the NRC, the agency should have the courage to say so, not obscure the goal in statistical terms of scientific notation unfamiliar to the public and using essentially meaningless concepts like " reactor-year."
8.
I dissent from the assertion that a containment performance objective should include consideration of the positive aspects of human performance. 'Ihe whole theory of defense-in-depth, the entire basis for the need for containment, is that you can't count on people to do the right thing, particularly in an accident, and that you need some independent backup system that will, with high confidence, contain the results of human error so as to prevent the public from being harmed by such mistakes. To reduce the requirements for containment performance on the assumption of appropriate human performance belies the very reason for having containments in the first placo.
- p. iii I am aware of no Workshop participants who were, as indicated in l
the Abstract (and repeated in the Introduction at page 1-1),
" representing...public interest groups." I thinx it is a failure that there were none, and that the Workshop was so heavily weighted with industry representatives. Of the sixteen participants, there were two participants associated with the I
utility organization EPRI (Zebroski and Leverett), two from f
specific nuclear utilities (O'Donnell from GPU and Boyer from Philadelphia Electric), two from reactor vendors (Parrette from Corrbustion Engineering and Temme from GE), and one from a major reactor architect engineering firm (Gardner from Stone & Webster).
Of the remaining nine, three were from national nuclear laboratories (Hodge from Oak Ridge, Kouts from Brookhaven, and Von Riesemann from Sandia), two were members of the NRC's Advisory Committee on Reactor Safeguards (Siess and Kerr), two were from universities (Theofanous and myself, both from the University of California), one from an independent technical consulting firm (Sholly from MHB Associates), and cne was from American Nuclear Insurers (Mariani).
2
o
,. =
,e
.o.
UNITED STATES NUCLEAR REGULA'IORY CDefiISSION CONTAIMtENT PERFORMANCE DESIQi OIL 7ECTIVE NORKSHOP 9
Assessing th t Need for Containment Performance Design Objectives for U.S. Nuclear Reactors Minority Report By Daniel Hirsch Director Stevenson Program on Nuclear Policy University of California, Santa Cruz*
July 29,1986 The views presented herein are solely those of the author and not necessarily those of the Stevenson Program on Nuclear Policy nor the University of California, Santa Cruz.
ABSIRACT
'Ihe " Containment Performance Design Objective Workshop" was designed to assist in development of a containment performance design objective (CPDO) for possible inclusion in the Safety Goal under consideration by the Nuclear Regulatory Commission. Given that potential failure of reactor containments could be catastrophic in severe accidents, and that containments in the U.S.
currently are not required to be designed to withstand core melt accidents, the need for speedy formulation and implementation of performance and design objectives for containments is of substantial importance. The importance of having a containment strong enough to withstand severe accidents was underscored by the recent Chernobyl accident, which was occurring at the time the CPD0 Workshop was meeting in Harpers Ferry.
ii
EXECUTIVE
SUMMARY
t e main areas of consensus were as follows:
1.
Containments are essential features of the defense-in-depth philosophy of reactor safety.
2.
Proper design and performance of containments are essential'if public health and safety are to be adequately protected.
3.
The qualitative societal goals considered by the Commission for its safety goal (essentially that nuclear power should not add significantly to existing societal risks) are too far removed from reactor design, performance, and accident sequence considerations to be of use in determining "how sefe is safe enough" and whether the regulations and nuclear plants indeed match that standard.
Quantitative measures of deceptablat core melt probability and containment failure probability are needed if such a safety goal is to be of use to reactor designers in designing reactors, reactor operators in operating the reactors, regulators in regulating them, and the public in having confidence in their safety.
4.
S e quantitative goals proposed for consideration of the CPDO Workshop raised two essential problems:
a.
The goals for acceptable probabilities of core melt accidents and acceptable probabilities of containmient failure would likely be unacceptable to the public, presenting probabilities of core melt and of containment failure that were far too high to be publicly acceptable.
b.
Nevertheless, the goals could not be met by U.S. reactors.
Thus, an exceedingly unpleasant situation exists--proposed safety goals cannot be met by American nuclear power plants, despite the fact that those proposed goals would permit a frequency of core melt accidents and containment failures that would be unacceptable to the American public if presented to them in a clear fashion.
Se main area of disagreement was as follows:
5.
Because current plants couldn't meet safety goals that would be acceptable to the public, the majority of the CPDO Workshop recommended that no quantitative safety goals be published at this time.
A minority Essentea, arguing that the Chernobyl accident made even more important prompt attention to improving containment performance and design and reducing probability of core melt accidents, and that ignoring the problem would not make it go away.
iii
+
INTPODUCTION One of the primary efforts of the U.S. Nuclear Regulatory Commission over the last five years or so has been an attempt to answer quantitatively the question:
"How safe is safe enough?" This enterprise, prompted in large measure by the then-Chairman Nunzio Palladino, was marked by a certain kind of optimism.
It was apparently felt that if the NRC could establish a quantitative safety goal that was acceptable to the public, and a quantitative mechanism of measuring compliance with that goal by nuclear designs, plant performance, and the regulatory structure itself, then public acceptance of nuclear power could be enhanced and a rational basis provided to nuclear regulation.
There was considerable controversy about this approach from the-beginning. 'Ihe technique for estimating probability of occurrence of events which had not yet occurred ("Probabilistic Risk Assessment" or PRA) contains large uncertainties and relies to an unfortunate extent on user inputs, which have tended repeatedly to be considerably too optimistic. Yet PRAs were seized upon as a way of attempting to demonstrate that U.S. reactors were " safe enet'gh"; some, in fact, have argued that these PRAs demonstrated that nuclear power was "too safe," and that safety and emergency regulations should be substantially relaxed.
The basic quantitative safety goal that was proposed for consideration was a core melt frequency of 10-4 core melts per reactor-year.
It was soon realized that a core melt goal was insufficient in determining safety because it said nothing about the adequacy of the containment to prevent major releases if the core did melt.
Therefore, the Advisory Committee on
.~
Reactor Safety (ACRS), among others, recommended the inclusion in the safety goal of a containment performance design objective (CPDO), as well as a quantitative goal for core melts.
(See ACRS letters of March 19 and April 15,1986].
@e need for a performance design objective for containments in severe accidents is especially important because containments in the U.S. are currently not required to be designed to deal with core melt accidents which, before tree Mile Island, had been officially deemed non-credible by the NRC and its predecessor agency, the AEC. The containments were only r/. quired to withstand the steam release from a major pipe break, but not the challenge involved with steam or hydrogen explosions or the containment loads associated with interaction of the molten core with concrete, direct heating challenges, or a range of other tremendous stresses which are likely to rupture containments during severe accidents.
Furthermore, recent data indicate that more than 50% of U.S.
reactors at any one time are in violation of their Technical Specifications limits regarding containment isolation; i.e., they leak at rates greater than permitted because of openings left open, some of which have been very large (e.g., equipment hatches). And there is great concern about containment bypass accidents, where the cause of the accident leaves a pathway for radioactivity to get out of the cantainment without ever causing it to crack or burst'.
NRC studies, old and new, have indicatea a high probability of containment failure in severe accidents because of these factors; the current debate is primarily over the timing of the failure, and whether late failure (e.g. 24 hours2.777778e-4 days <br />0.00667 hours <br />3.968254e-5 weeks <br />9.132e-6 months <br />) might provide some time for some fission product mitigation to occur before release into the environment.
We prospect for containment failure, whether late or early, is of such importance that setting goals for containment performance and design is a very important enterprise.
2
The NRC, through its Office of Nuclear Regulatory Research, convened a special advisory group to address the question of a containment performance design objective-the CPD0 Workshop-and contracted with Brookhaven National Laboratory (BNL) to propose and analyze a variety of possible CPDO standards for consideration by the Workshop. The CPDO Workshop met in Harpers Ferry, West Virginia, on May 12-13. Ironically, the Chernobyl accident had just occurred and the entire question of containment performance took on additional urgency. The meetings were substantially colored by the accident which was still ongoing at the time.
THE CENIRAL QUESTIONS The central question that should be addressed related to an adequate safety goal in general and a containment performance design objective in specific is:
how safe is " safe enough"? Put differently, what, for the majority of the American public, would be an acceptable risk from nuclear power?
Th'is question was barely touched on during the CPD0 Workshop.
The central discussion seemed to be about what safety goal and CPD0 the NRC and the nuclear industry could " live with," and whether it wouldn't be better not to have any such goals at all. As several participants said about the apparent arbitrariness of the proposed core melt and containment objectives, they seemed to have been arrived at by approximating the order of magnitude at which plants currently were, on the assumption that a " goal" that merely described the current situation would cause no burden to licensees.
Some participants, however, felt that this process approached a serious matter backwards and begged the question of what was an appropriate level of 3
nuclear safety.
DISCUSSION OF PROPOSED QUANTITATIVE SAFE 1Y GOALS The Workshop participants discussed at some length the various proposals that had been presented to them for quantitative safety goals that would inclu' e a containment performance objective.
The basic form of the d
proposals was as follows: a core melt probability of 10-4 per reactor-year, a conditional containment failure probability given a core melt of 10-1, plus an overall probability of an accident with a major release of 10-6 per reactor-year. (The last item was to make clear that containment was part of the defense-in-depth philosophy, and that by reducing on paper one's core melt probability an order of magnitude below the safety goal, one couldn't go ahead and build a reactor without a containment).
PROPOSED GOALS UtMCCEPTABLE It. was pointed o' t at the CPDO Workshop that the proposed safety goals u
were written in an obscure fashion that would be unlikely to be understandable to the general public.
It is particularly impottant if the safety goals are to enhance public confidence and support that they be clear and accessible.
Terms like " reactor-year" tend to obscure true risks, both because of their awkwardness and because the real risk that the public is concerned about is not the probability per reactor per year, which will tend to be small, but the probability that there will be a core melt at some reactor at some point, a probability which will be much larger.
-4 Thus, the proposed core melt frequency of 10 core melts per reactor 4
~
year translates into about a 40% chance of at least one core melt at one of the current or som-to-be m-line nuclear power plants in this country over their lifetime (generally excluding the additional probability of external initiators such as earthquakes or sabotage).
A one-in-ten-thousand chance of core melt per reactor-year may sound vanishingly small and acceptable if people don't understand the term; but a 40% chance of a core melt at a U.S.
reactor would likely be unacceptable to mest of the public, even if there were high confidence one were meeting that goal.
The 40% risk of core melt assumes reactors meet the goal, that one knows with high confidence that they do, that external initiators such as sabotage and earthquakes which are generally excluded from such considerations do not occur, and that the NRC is prohibited from granting licenses for rew nuclear power plants in excess of the total currently licensed to operate or under active construction.
If there is no prohibition on new reactors, and a modest number of new plants (roughly doubling the present number) are considered in the risk equation, the safety goal would mean a 60-70% probability of a core melt during their lifetime. This is again a figure unlikely to be acceptable to the public, even though it represents a small fraction of the number of nuclear plants often proposed by government and industry representatives in the past.
5
'Ihe probability of at least one core melt is conputed as follows:
P
- 1-*
cm where Pcm = Probability of core melt in a U.S. reactor over a certain period hem = Probability of core melt per reactor-year t = number of reactors operating times the nunber of years they operate (See S._ Mitra et al, "A Study of the Implications of Applying Quantitative Risk Criteria in the Licensing of Nuclear Power Plants in the United States," U.S. Nuclear Regulatory Commission, NUREG/CR-2040, May 1981, p. 6W If the purpose is to ' inform the public clearly what the NRC believes is an acceptable level of public safety in the nuclear area, the safety goals need to be expressed in a clear, accessible form that doesn't require average citizens to go through calculations using the formula above. If the NRC feels a 40-70% chance of a core melt is acceptable, then it should say so. The public is, however, unlikely to find such a goal acceptable if clearly told what it means. And obscuring the true meaning of such safety goals because of fear of public disapproval is poor public policy.
The same must be said for the proposed conditional failure rate of 10-1 for containments. If the public is unlikely to be comforted by a " goal" of a 40-70% chance of a core melt, it is also unlikely to feel adequately protected by an announcement by the regulatory agency that it is " aiming for" a likelihood of the containment failing of one in ten. Particularly given the vast uncertainties in containment loads and performance, such a
" goal" would sound unpleasantly like a virtual certainty of a core melt, followed by a Russian roulette chance of radioactive release because the containment burste.
Therefore, if the proposed quantitative safety goals are expressed in i
6
o clear terms, rather than obscured in exponentials and reactor-years, they are not likely to be considered socially acceptable. The proposed goals thus fail to adequately answer the question: "How safe is safe enough?"
'Ibey propose a level of safety, or more precisely, danger, that is far from being considered safe enough by a majority of the American public.
The proposed goals thus fail either test:
If the purpose of safety goals is to truly determine an o
adequate level of safety, these goals would fail, If the purpose is to gain public confidence for nuclear power, o
they also would fail.
ACCEPIABLE SAETTY GOALS What then would be an acceptable level of safety?
(1) the probability of a core melt accident occurring in this country over people's lifetimes of one in a hundred, and (2) a containment failure probability of one in a hundred, both probabilities to be determined with high confidence
- not to be higher than the proposed goals.
statistically speaking, one is referring to a 95% confidence level.
7
The public has often heard it asserted that there is a one in a million, or one in ten million, or one in one hundred million chance of a severe accident. 'Ihat would be acceptable if true and if they believed it, but they don't believe it, and for good reason, in part because of the obscuring of tr'ue risk by use of terms such as reactor-year, as discusse'd above. If the public could have high confidence in the safety claims, a one in a hundred chance of a core melt in the U.S.
might, however, be acceptable, as would be a one in a hundred chance of containment failure.
Such a safety goal would be an order of magnitude stricter for containment performance, and a core melt probability about two orders of magnitude less likely than those under consideration.
U.S. PLANTS CAN'T EVEN MEET 'IEE PROPOSED SAFETY GOALS We have seen in the discussion above that the proposed safety goals are unacceptably lax--a virtual certainty of a core melt accident and a Russian roulette chance thereafter of the containment holding. We have seen that to adequately answer the central public policy question of "how safe is safe enough" in a prudent and acceptable way would require a containment goal an order of magnitude better than proposed and a core melt goal one or two orders of magnitude better. We will now examine, however, how well U.S.
plants perform against even the considerably less adequate goals under consideration.
The probabilistic risk assessments performed to date have come in at about three times the proposed safety goal of 10-4 That is where the famous NRC estimate of a 45% probability of at least one core melt in the next twenty years came from.
The figure for the current population of 8
reactors for their lifetimes (rather than just 20 years) is about 80%
probability of a core melt; about 95% if additional plants were built in about'the same number as are currently in operation or under construction.
(It is for that reason that the NRC Severe Accident Policy Statement has caused so much concern with its declaration that current plants are safe enough; if future plants, should any be built, are no safer than current plants, then there is an unacceptably high probability of severe accidents occurring.)
Furthermore, the PRAs, even though showing risks about three times higher than the proposed safety goal, still appear to substantially underestimate actual risks by at least an additional factor of two. As William Kerr noted at the CPDO Workshop, there is a rule about PRAs:
they tend to come in with a core melt probability of about 10-6, which goes up, after initial review, to about 10-5, and goes up again after further review to about 10-4 One can only wonder what the value would be with another l
layer of review.
The empirical evidence to date suggests that the true values are an additional order of magnitude above that 10-4 level. There have been two severe core damage accidents in power reactors to date. The first, at TMI, occurred after only a few hundred reactor-years.
Even if one includes the years since, one must now add Chernoby1: two severe accidents in about 3000 reactor-years, or one per 1500 reactor-years, nearly 10-3 Actual reactor experience is more pessimistic than the rather rosy, yet still unacceptable, PRA projections.
We now seem to be having severe reactor accidents at a rate an order of magnitude more f requently than the proposed core melt safety goal, approximately every 7 years or so. It is unlikely that the expectation of even a single additional severe accident would be acceptable to the public.
9
Over a person's lifetime, even assuming the reactors do meec the proposed 10-1 goal for containment performtace, that means a one in ten chance of containment failure each time there is a severe accident, and a substantial probability of numerous such events in a person's lifetime. Se probability of at least one severe core damage accident with containment breach is thus very high, unacceptably high.
(For example, the current accident rate suggests an 80% chance of five or more core melts over the life of the reactors given the current population of ones on-line and nearing completion in the U.S. and an equal number of additional ones if there is no ban on future plants.)
What about the containment failure portion of the safety goal?
The CPDO Working Group, a group of NRC and Brookhaven staff who prepared analyses of the propsed CPDOs for consideration by the Workshop, evaluated various proposed containment performance objectives against current estimates of containment performance.
In a stunning conclusion, they found that virtually none of the reactor tyces could meet virtually any of the prooosed standards. This was true whether one used the methodology of the Reactor. Safety Study (WASH-1400) or the new source term methodology!
[See Table El.3 and summary, from "Information for the Participants in the Containment Ferformance Design Objective Workshop" by CPDO Working Group, Pradyot K. Niyogi, RES George Sege, NRR; and Kelvin Shiu, BNL; April 24, t
I 1986; Appendix A, " preliminary Evaluations of Various Containment Performance Design Objective Options," page 9 and Table El.3.]
In eleven of the twelve evaluations performed, the containments examined failed to meet the proposed containment performance goals, often by very large margins (e.g., a 93% failure probability compared against a goal of 10%.) In the single evaluation where a containment met a performance l
10
Pr 11;inary ResYlts tf CPD0 Option, Eval,uatinst P
ntainment Peach Surry Bottom Limertet GE55AR Zion Options
,ge X FAILS TO MEET FAILS WM Option 1 GOAL.15
.33 47 W R$"
.05. MEETS coal
.01*
\\ FAILS 1.4(-5)5 2.9(-5)N Option 2 10 5 2.O(-5)N 2.7(-5)5 7.1(-6) 10* S-N FAII.S FA m Fission-Product 1.1(-5)5 Attenuation 4.2(-5)N pagg,g
.le FAILS
.24
.01
.9 Option 4 10-8 9(-6)
.7(-5) 7(-6)
Flexible mI.s mI.s Approach
.23 4.3(-6)
Opt 1on 5 WASH.
hew 1400 Sourc Terms NUREG-0956
- The values in these ILones are the quantitative objectives for that option.
- N = nonsolids 5 = solids tihe lower left hand values are based on the new source tem (draft NUREG-0956) Infomation with containment reevaluation; the upper right hand values are based on the reactor safety study WASH-1400.
Source: ' Preliminary Evaluations of various Containment Performance Design Gay!ctive options," Appendix A of
- Containment Perferr.ance Design ob)ective: options, Implementation, ana !ssues.* in *Information for the Participants in the Containment Perforr.ince Design objective worksnop,' prepared ty the CPDo Working croup:
Pradyot K. Nayo31, NRC office of taaclear Negulatory Researcht coorge Sege, NRC office of Nuclear Reactor Regulatacr$3 Relvin Shieu, Brookhaven National IAboratoryg U.S. Nuclear Regulatory Commission, Washington, D.C., April 24, 1986.
1 tie table summarizes results of evaluations performed by the NRC Containment Working Group comparing performance estimates for three ccritainment types against the primary proposed containment performance design objectives.
Results are available at this time for only Options 1,2, and 4. The quantitative approaches of the various CPDO options are summarized,in the first column and described in more detail in the Working Group s full report. Three containment types were evaluated: a PWR subatmospheric containment (Surry), a BWR Mark I (Peach Bottom), and a BWR Mark II (I.imerick). ftr the Surry containment, two sets of results are presented:
the lower left hand values were obtained using the methodology of the Reactor Safety Study (WASH-1400) whereas the upper right hand results are bsed on new source term information provided in draft NUREG-0956 with reevaluation of containment event trecs. 'Itte values adjacent to each option represent the quantitative ob3ectives of the CPDO option in question for compartson with the containment-specific results. (fbr example, the chart indicates that the Peach Bottom type of ccritainment is estimated to have a 93t probability of failure, as cefined in CPDO Option 1, whereas that option sets a goal of only lot failure probability.) Note that although the various options define contatnment performance in different ways, and although three different ccritainment types are used for evaluation purposes, only one of the twelve evaluations suggests ability to meet any of the proposed containment performance design co]ectives and tt.at the new source term assq:ttons do not appear to provide substantially different results.
11
e
?
goal, that same contai.nment failed the other two containment objectives options considered. Additionally, for the single containment cbjective that was met, it passed only by use of one calculational method while failing by another.
Furthermore, for those who hoped the new source term assumptions would demonstrate markedly lower risk and greater safety in current reactors than previously assumed, the analysis provides little comfort. For the particular containment type (the Surry PWR subatmospheric) for which comparisons against various CPDO options were made between the new source term NUREG-0956 and the older WASH-1400 methodologies, the evaluation concludes that:
...it appears that the new source term methodology draft NUREG-0956 provides only slightly different results. The WASH-1400 results do not satisfy any of the options while the new source term methodology only satisfy [ sic] Option 1.
Niyogi, et. g., loc. cit.
Recent reports have underscored these disconcerting conclusions regarding the inability of U.S.
containments to meet even minimal containment performance setandards. Ebr example, the Wall Street Journal of July 22,1986, reports.that an NRC reevaluation of the GE Mark I containment system, of which there are about two dozen in the U.S.,
reaffirms longstanding NRC estimates of a 90% likelihood of containment failure in a severe accident.
["NRC to Report GE Containment System on Certain Nuclear Reactors is Flawed," by Bill Paul.] That is nine times the prooosed safety goal for containment performance and an extraordinarily unacceptable failure probabilitv.
Additionally, briefings provided the CPDO Workshop during the Chernobyl accident indicated that that reactor had two containment volumes, one at 27 pounds per square inch design pressure, the other at 57.
Quite a nutter of U.S. reactors have design pressure lower than those. While it is not yet 12 l
s known whether those containment areas failed or if the accident bypassed them, either prospect raises serious questions for U.S. reactors and their containments. Furthermore, arguments that the area above the Chernobyl 6
reactor could not be a successful pressure boundary in case of accident because of the many penetrations, even if true, would merely raise an additional concern regarding U.S. containments, which also have numerous penetrations, although not as many.
'Itte empirical evidence to date is that severe core damage accidents are occurring not at 10-4 intervals, but closer to an order of magnitude more frequently,, about every 1500 reactor-years, or every 7 human years. The evidence regarding containment failure is basically one for two, as opposed to the goal of one in ten: TMI, whose containment was never tested because the core damage was arrested before it could melt through the reactor vessel and challenge the containment, and Chernobyl, where its post-TMI containment volumes (and pressure suppression pools, like ours) were either bypassed or otherwise failed. To those who say one should not consider Chernobyl relevant because our containments are different or more extensive, even if true, one must respond that that is precisely why assuring adequate containment performance is so essential here, for a containment that holds is all that keeps U.S. reactors in core melt situations from becoming the unmitigated disaster that was Chernobyl.
And, as indicated in the j
evaluation prepared for the Workshop and discussed above, U.S. reactors 1
could basically meet none of the containment objectives proposed, inadequate as they are. As uncomforting as a 10% failure goal would be to the public, l
current estimates of a 90% probability of containment failure for the Mark I l
containment are even less reassuring.
13 l
d CUNCLUSIONS l
1.
Quantitative safety goals should be written in language clearly understandable to the public.
j 2.
When so written, the proposM goals are unacceptable, proposing to set
" goals" that would permit intolerably high probabilities of core melt and of containment failure.
3.
Safety goals acceptable from a public policy standpoint would require core melt frequencies about one hundred times lower than proposed and containment failure probabilities ten times lower.
4.
Inadequate though the pr.oposed safety goals are, current U.S. plants apparently cannot meet them. In fact, U.S. plants in general fail to meet the proposed standards by a substantial margin.
While' agreeing that the proposed quantitative safety goals, if written in plain English, would propose probabilities of severe accidents and of containment failure so high as to be unacceptable to most of the American public, and while agreeing that U.S. plants in general couldn't meet the proposed goals, lax as they are, many of the participants in the CPDO Workshop had a rather strange solution to the problem. 'Ihey proposed simply that the problem be solved by not issuing any quantitative safety goals, releasing instead only vague " qualitative" societal goals plus a statement that a containment performance objective will be addressed at some unstated point in the future.
This would be irresponsible.
The NRC has for five years tried to answer two questions:
"how safe is safe enough?" and "are our nuclear plants that safe?" It is understandable that the enterprise was undertaken
~
wi:h the assumption that a socially-acceptable level of safety and the level 14
J e
.y of current plants would match. But when, after five years and tremendous sums of money, the results come back disquieting--the best safety goal the agency can come up with is orders of mag.-itude less safe than the public would accept if published and nonetheless our plants are far from meeting even that min'imal goal--it would be unforgiveable to suppress those findings.
The scientific method involves asking a question and finding the answer, not throwing out answers one does not like.
We agency asked, "Are our plants safe enough?" The answer, particularly with regards the all-important containment, appears to be "No."
Were are two reasonable public policy responses to the results of this many-years-long investigation:
bring the safety level up to a publicly acceptable safety goal, or withdraw the licenses because the Commission cannot make the affirmative safety finding required of it by the Atomic Energy Act. A third approach, choosing to defer a containment performance objective to some unspecified time in the future or failing to publish a quantitative core melt goal, particularly in terms understandable to the public, would be very poor public policy.
l 15 I
---