ML20137N803
| ML20137N803 | |
| Person / Time | |
|---|---|
| Issue date: | 01/29/1986 |
| From: | Advisory Committee on Reactor Safeguards |
| To: | |
| References | |
| ACRS-T-1484, NUDOCS 8602040253 | |
| Download: ML20137N803 (175) | |
Text
. 0Y GI N A 8888
-/ 9 n
UN11ED STATES NUCLEAR REGULATORY COMMISSION IN THE MATTER OF:
DOCKET NO:
ADVISORY COMMITTEE ON REACTOR SAhEGUARDS SUBCOMMITTEE ON CLASS 9 (SEVERE) ACCIDENTS A
\\
(.)
LOCATION:
WASHINGTON, D.
C.
PAGES:
1 - 124 DATE:
WEDNESDAY, JANUARY 29, 1986 I
4%.. '
c
=. ~
\\
mem huaha y st wf u as Loba
}aJ.otRemovefrom ACRS0ffice ACE-FEDERAL REPORTERS, INC.
b O
Official Reporters 444 North CapitolStreet
\\
Washington, D.C. 20001 D
(202)347-3700 e60204025:4 az.o n o T
43 I' Tift
CR25584.0 1
DAV/sjg 1
UNITED STATES OF AMERICA
- m
)
2 NUCLEAR REGULATORY COMMISSION J
3 ADVISORY COMMITTEE ON REACTOR SAFEGUARDS SUBCOMMITTEE ON CLASS 9 (SEVERE) ACCIDENTS 5
Nuclear Regulatory Commission Room 1046 6
1717 H Street, N.W.
Washington, D.
C.
7 Wednesday, January 29, 1986 9
The subcommittee meeting convened at 1:00 p.m.,
Dr.
I 10 William Kerr presiding.
l i
11 ACRS MEMBERS PRESENT:
I2 DR. WILLIAM KERR I
O, DR. CHESTER P.
SEISS 13 DR. CARSON MARK MR. DAVID A. WARD 14 CONSULTANTS PRESENT:
MR. MYER BENDER.
MR. PETER R.
DAVIS I
16 DR. IVAN CATTON DR. SAM C.
SAUNDERS 17 l DEAN HOUSTON, ACRS Staff Member 18 l
19 1
20 1
21 l
22 i
?
23
(')
\\,/
24 Am-Federd Reporters, Inc.
25 L
PUBLIC NOTICE BY THE s
UNITED STATES NUCLEAR REGULATORY COMMISSIONERS' ADVISORY COMMITTEE ON REACTOR SAFEGUARDS WEDNESDAY, JANUARY 29, 1986 The contents of this stenographic transcript of the proceedings of the United States Nuclear Regulatory Commission's Advisory Committee on Reactor Safeguards J
(ACRS), as reported herein, is an uncorrected record of the discussions recorded at the meeting held on the above date.
No member of the ACRS Staff and no participant at
()
this meeting accepts any responsibility for. errors or inaccuracies of statement or data contained in this transcript.
4 a
e
\\
(~
s 5840 01 01 2
1 DAVbur 1
PROCEEDINGS
(
2 DR. KERR:
(presiding)
The meeting will come to 3
order.
4 This is a meeting of the Advisory Committee on 5
Reactor Safeguards, Subcommittee on Class 9 Accidents.
6 I am the subcommittee chairman.
My name is 7
William Kerr.
8 Other ACRS members here today are Carson Mark, 9
Dave Ward -- he says he will be along soon -- and Chet 10 Siess.
11 We have as consultants Mr. Bender, Mr. Catton, 12 Mr. Davis, and Mr. Saunders.
13 The purpose of the meeting is to begin discussion U-s 14 of NUREG-1150, entitled " Nuclear Power Plant Risks in 15 Regulatory Applications."
16 Dean Houston is the ACRS staff member present for 17 the meeting today.
18 The rules for participation in the meeting were 19 announced as part of the notice of the meeting published in 20 the Federal Register on December 24th of 1985.
21 A transcript of the meeting is being kept and 22 will be made available as stated in the notice.
23 I would ask that each speaker identify himself or 24 herself and use a microphone.
25 We have received no written comments from members ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800 336-6646
5840 01 02 3
1 DAVbur 1
of the public, nor have we been asked to make time for oral
()
2 comments from members of the public.
3 We will proceed with the meeting, but I might say 4
that in contrast to what one might gather from the published 5
agenda, I would hope to end the meeting by 4:30.
6 Are there any comments before we call upon 7
Mr. Ernst or questions?
8 (No response.)
9 DR. KERR:
Mr. Ernst, the floor is yours.
10 MR. ERNST:
Thank you.
11 (Slide.)
12 We have a three-part agenda today.
13 First, I will give just a few opening remarks on 7b 14 the purpose, the contents, the present schedule and status, 15 things of that nature with regard to the preparation of 16 NUREG-1150.
17 We met with the ACRS -- I believe it was sometime 18 in middle to late last summer and discussed in general what I
19 we.ere about -- the need to rebaseline the risk estimates 20 on six reference plants, the interf ace with IDCOR and things 21 of that nature.
22 So I won't repeat that kind of information, but I 23 will sort of give a status report.
24 You were passed out a draft outline of 25 NUREG-ll50.
We will discuss that not in a lot of detail, ACE-FEDERAL REPORTERS, INC.
[
202-347-3700 Nationwide Coverage 800-336-6646
5840 01 03 4
1 DAVbur 1
but certainly we are here to answer questions.
()
2 Joe Murphy then has a presentation on methodology 3
and uncertainties in this risk rebaselining work and then an 4
important aspect of this, he will also be discussing the 5
current approach to assessing reasonable ranges of 6
uncertainties with regard to the risk information.
7 Mark Cunningham then will discuss some of the 8
major issues that we are addressing in this uncertainty 9
analysis and give a few words on how we are going about this 10 process.
I won't call it a process for resolving the 1
11 issues.
What I will call it is a process for coming to a 12 regulatory position on a reasonable range for these major 13 issues.
14 So that is basically what we are about today.
15 (Slide.)
16 First, just a few quick words on the purpose of 17 1150, somewhat repetitive from last summer, but that is okay 18 to heve one slide repetitive, I guess.
19 The basic purpose is to provide a current 20l understanding of the likelihood of core melt and risk from 21 severe accidents at six nuclear power plants.
We won't call 22 them representative of the entire industry, but we will 23 characterize them -- as for these six plants of six 24 different containment types, we will characterize the risk.
25 We have in these six plants six different s.
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-33H646
5840 01 04 5
1 DAVbur 1
containment designs.
The plants are being updated as to TMI (l
2 fixes.
We are using current source term information, not 3
only the science, the chemistry, and so forth of source 4
terms but also updated information on containment 5
performance.
6 So that is one purpose -- is to document what we 7
know to the best of our knowledge with regard to these six 8l plants.
9 The second purpose is to sit back and try to 10 assess the use of such information for various types of 11 regulatory application.
12 In going through these plant analyses, you get a 13 number of insights, "what if" kinds of things, what if the 14 plant might have been built dif ferently, because you are 15 really looking for the important outliers and you are 16 drawing on experience from other plants, from our ASAP 17 modeling, generic modeling base, and things of that nature.
I 18 l DR. MARK:
Excuse me.
You made a reference to 19 information concerning containment performance.
20 What kind of information exists on that subject?
21 MR. ERNST:
This would basically be the results 22 of containment loads and containment performance working 23 groups and that kind of information, whatever exists in the 24 way of information today.
We are not generating new
(-
25 information for 1150.
V) l ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 80 4 336-6646
5840 01 05 6
1 DAVbur 1
DR. MARK:
I can understand that you can talk 2
about pressure levels at which the containment might or 3
might not remain intact, but containment performance is a 4
great deal more than that, and I wasn't aware you had an 5
awful lot of information on it.
6 MR. ERNST:
Whatever is available will be used.
7 There are -- well, I hesitate even to talk about it at this 8
stage of the game.
9 There is a meeting that is going to be held at 10 CB&I, for example, on their insights on the work they have 11 done for IDCOR on containment performance at high 12 temperature levels, Mark I's at high temperature levels, and 13 l so forth, where they have come up with some judgments on
()
i 14 where the containment might fail.
15 We are going to look at information like that.
16 Whatever is available at the present time we will look at, 17 and the information will wind up with some error bands on 18 the judgments we use in trying to calculate risks based on 19 this information.
20 DR. MARK:
Thank you.
21 MR. ERNST:
In assessing the usefulness, one 22 thing we are looking for is to mesh very closely with NRR's 23 ongoing work in trying to develop a methodology and process 24 by which the risk important features of other plants -- what
}
are these features and what might be done to rectify 25 ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-336-6646
I 5840 01 06 7
1 DAVbur 1
features if they happen to be risk outliers?
(m}
2 To be helpful in this process, we are asking the 3
analysts to draw upon the information in their detailed 4
analysis of the six reference plants and document in 1150 5
those important insights you get from this that might be 6
useful to extrapolation to other plants.
7 Clearly, NRR has the responsibility for 8
developing the methodology, drawing upon all inf o rma tion,
9 and we think 1150 is one aspect of that available 10 information.
11 Other areas of usefulness, PRA has for some time 12 now served the agency in providing information to help 13 plant-specific and generic regulatory decisions.
We will 14 take a look at the existing information we have compared to 15 the old information and see whether or not we have dif ferent 16 kinds of insights that might be drawn at this time compared 17 to what was in, for example, NUREG-1050, which was published 18 a year and a half ago.
19 MR. WARD:
Mal, I really always have to -- I 20 always blanch when you say something like PRA has been 21 useful for providing plant-specific information for 22 decisions.
23 It certainly provides design-specific 24 information, but I still have a lot of trouble with whether 25 the PRA is capable of telling us anything at all about the O
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-336-66M
6 -
5840 01 07 8
1 DAVbur 1
plant as opposed to the design.
()
2 MR. ERNST:
By the plant, you mean the people and 3
the procedures?
4 MR. WARD:
Yes, even the equipment and the 5
systems in the plant.
They are all in the performance, but 6
also the people and the proc dures and that sort of thing.
7 PRA just has a big blank ther 8
There doesn't seem to b h interest in 9
developing ways for PRA to deal usefully within those 10 areas.
11 MR. ERNST:
I don't think I would agree 12 completely if I understand what you are saying.
e 13 It is true PRA can come up with some judgments on O
14 the importance of one design versus another design, with 15 certain assumptions on how the equipment is going to work 16 and things like that.
17 You can't assume equipment is going to work r
18 perfectly, and even comparing one design to another, you 19 have certain assumptions with regard to equipment operation 20 and the behavior of people.
21 We are trying -- I am sure not as well as we 22 would like to succeed, but we are trying to get at the 23 sensitivity of maintenance procedures.
We spent a fair 24 amount of time, for example, on the human reliability 25 aspects of Peach Bottom with respect to the ATWS scenario.
l l
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800 336-6M6
5840 01 08 9
1 DAVbur 1
We spent a lot of time on venting of Mark I I
2 contairments.
3 MR. WARD:
But even there, the way you are 4
approaching human reliability, it is sort of a design 5
standpoint.
You aren't talking about reactor operators at 6
Peach Bottom.
You don't know anything about them.
There is 7
no information on them going in.
8 You are talking about a designer's SRO or 9
something.
10 MR. ERNST:
To some extent, that is true.
To 11 some extent, where we rely perhaps on some simulator data, 12 and so forth, that might be generated, like at La Salle, 13 there has been a fair amount of simulator work done by the 14 operators.
So that, you could say, was somewhat plant 15 specific.
16 But I will agree with you, particularly when you 17 get in the management area and things like that, there is no 18 l way to factor that kind of thing into the PRA.
19 DR. KERR:
Now that Mr. Ward has made his sales 20 pitch on human performance.-- I am not denigrating it.
I 21 agree with you, but I don't believe Mr. Ernst is going to be 22 able to include it in his report.
23 MR. WARD:
I don't know that it is necessary to 24 ignore it.
25 (Laughter.)
O ACE FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-336-6646
5840 01 09 10 1
DAVbur 1
MR. ERNST:
The best we can do -- and maybe Joe
(
2 might want to make a comment on this -- the best we can do 3
is try to characterize how that might vary in the 4
assumptions made.
5 It clearly has some uncertainty bands around it.
6 That I agree.
7 DR. MARK:
You have mentioned taking data from 8
six plants, I believe.
9 Do these to any extent overlap with the plants 10 being studied for A-45, the generic issue of decay heat 11 removal?
12 MR. ERNST:
I am not sure, 13 DR. MARK:
I am wondering why they don't exactly f-t_)
14 overlap.
15 MR. ERNST:
Zoltan is back there.
I am not sure 16 what A-45 plants there are.
17 MR. ROSTOCZY:
I don't know which are the plants 18 for that.
I cannot state.
19 MR. WARD:
Just looking at the list, Carson, I 20 don't think there are any that are common.
21 DR. MARK:
My question is:
why the hell aren' t 22 they the same?
23 DR. SAUNDERS:
It would have been cheaper to do 24 it that way.
25 DR. KERR:
Please continue.
)
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-336-6646
5840 01 10 11 1
DAVbur 1
We will make note of the question.
gs J
2 MR. ERNST:
Okay.
That is probably enough on 3
that slide.
4 DR. KERR:
From that slide, I gather that only 5
internal initiators or internal events are being treated as 6
limited?
7 MR. ERNST:
That is correct.
There may be 8
insights on external events.
There will be a subsection on 9
external events, but not a plant specific analysis.
10 DR. KERR:
This is based on a lamp post 11 criterion?
The place to look for the lost quarter is where 12 the light is and not where the quarter is?
13l MR. ERNST:
I hope it is more than that.
\\#
14 DR. KERR:
We have numerous -- well, numerous is 15 two or three -- PRAs that tell us that the principal risk is 16 from seismic events.
That may be incorrect.
So we do a 17 risk rebaselining, which presumably is going to give us a 18 fairly good idea of what the risk is from these six 19l referer.ce plants, and we say very little about what appears 20 to be a strong risk contributor.
21 MR. ERNST:
What will be described in 1150 will 22 be some of the insights gained from other PRAs and a 23 description of the process that the agency is going.through 24 to try and address external events in NRR's process for 25 implementing severe accident policy.
b3 ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800 336-6646
5840 01 11 12 1
DAVbur 1
There is a Commission paper being put together 2
now to try to treat that.
3 The basic rationale for not having it is that the i
4 proper way to approach the problem and the proper way to 5
weight risk insights that you get from external versus 4
6 internal is an area of controversy that seems like it is 7
probably more important for the agency to look for the 8
vulnerabilities, which is part ei the implementation plan 9
process, than it is to try and come up with a risk statement I
10 at this time on the six reference plants.
j 11 The issue has not been forgotten.
It is just 12 that it seemed like, considering the timing problem of 13 trying to interface with NRR's need in working with IDCOR, O
I 14 plus the state of knowledge of the credibility of the risk 15 judgments, made it suspect whether or not we should -- and 16 we rightly or wrongly made the decision not to include it, 17 but not to forget it either from the standpoint of the 18; agency's implementation of severe accident policy.
19 DR. KERR:
It just seems to me that implicit in 20 this is a decision that the external events are not very i
21 important.
22 MR. ERNST:
That is not my characterization, nor i
23 is it the characterization --
1 24 DR. KERR:
But you tell me that 1150 is going to 1
25 be used to make decisions on what to do about severe ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800 336-6646
5840 01 12 13 1
DAVbur 1
accidents and what to do about plant safety and what to do
()
2 about other plants.
3 MR. ERNST:
It will provide some information for 4
this process, and excluded from 1150 is bottom line risk 5
determination for external events.
6 There is other information available to NRR for 7
making that kind of judgment.
As I say, there is a staff 8
paper being put together to try and outline an agency 9
posture on just how to do that without forgetting that 1150 10 does not cover that aspect.
11 It will say that it is worthwhile spending the 12 money to do the risk analysis as opposed to just money for 13 vulnerabilities.
Then 1150 could be updated in that area at O
14 some later time.
15 DR. KERR:
1150 is said to be a risk 16 rebaselining, not a vulnerability rebaselining?
17 MR. ERNST:
That is right.
18 DR. KERR:
So in the context of 1150, risk is 19 deemed to be important, but in the context of something 20 else, vulnerabilities, what is important?
21 '
Help me here.
22 MR. MURPHY:
I think there's two points that lead 23 us to that decision.
24 One is that as we look at the existing PRAs that 25 have included seismic events, the problems that have caused ACE-FEDERAL REPORTERS, INC.
l 202-347-3700 Nationwide Coverage 800-336-6646
5840 01 13 14 1
DAVbur 1
seismic to be high in them, have tended largely, although
(~x()
2 not exclusively, to be plant specific, so that it has very 3
little generic applicability.
4 We have got things like ceilings of control rooms 5
falling down and buildings banging together, things of this Rather than say component fragility dominating, it 6
sort.
7 tends to be some strange portion of the structural design or 8
something like that.
9 This makes its use, as we intend 1150 to give 10 information to 10-R, to be a little less meaningful.
11 Because it is plant specific rather than generic, it is very 12 difficult to extrapolate.
131 The other thing is that the uncertainties in the (1)
I 14 seismic parameter are very large, and seismic analyses, of 15 course, are very expensive.
We are investigating right now 16 our ability to come up with methods that can do a reasonable 17 seismic analysis, and if we can find those, then these will 18 be part of the discussions Mal mentioned, where we look at 19 whether we want an assessment of vulnerability that is 1
20 sufficient or whetner we want to go on to an assessment of 21 risk.
22 Before we can make that we have to have the 23 methods.
Methods such as SSMRP are extremely expensive, and 24 considering the uncertainty and the basic hazard curve on 25 that, it makes you wonder whether it is cost effective to ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-336-6646
5840 01 14 15 1
DAVbur 1
utilize it for this purpose.
O 2
3 4
5 6
7 8
9 10 11 12 13 O
14 15 16 17 18 19 20 21 22 23 24 25 ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-336-6646
5840 02 01 16 1
DAV/bc 1
, DR. KERR:
One more comment and I'll let this
,,()
2 go.
But, if I remember the review of the GESSAR standard 3
plan, about 50 percent of the risk from the seismic hazard 4
was attributed to relay chatter.
That turned out to be a 5
significant contributor.
6 Now, that's plant-specific in that plants have 7
different relays.
But it seems to me relay chatter is a 8
generic issue if, indeed, it occurs in large seismic 9
events.
10 MR. MURPHY:
I would agree.
11 DR. KERR:
And it could turn out to be a big 12 contributor there.
It's calculated to be.
It bothers me a 13 bit to think that we are rebaselining risk, which I think is O
14 impor' tant to do, yet we may be leaving out a very 15 significant contributor.
I don't know whether it's a 16 significant contributor or not, but at least some 17 calculations seem to indicate that it is.
18 Somebody had his or her hand up back here.
No?
19 okay.
Continue.
I 20, (Slide.)
21 MR. ERNST:
This is a chart put together just to 22 give you a feel for the various pieces of this review and 23 how they get put together.
The elements in green, if you're 24 familiar, they are associated with the bins that we have.
(~T 25 The green is the SARRP at Sandia.
The red ones are the ASEP
\\_)
1 1
ACE-FEDERAL REPORTERS, INC.
202 Nationwide Coverage 800-336-6646
,, _, ____ - _-347-3700
5840 02 02 17 1
DAv/bc 1
bins at Sandia, and some work being done at INEL.
The
(~T 2
yellow represents other funding.
The process is basically
(.)
3 to develop a containment tree based on the results of 2104 4
runs and other information to come up with some source term 5
bidding, to try and pick the dominant, also using ASEP 6
information, pick the dominant accident sequences, and then 7
get source term code package code runs made on those, which 8
falls into here.
9 This also goes into uncertainty analyses.
This 10 is somewhat misstated.
It's more than just source term 11 uncertainty analyses, we also include the front end in that.
12 Some other projects.
We have a fair amount of 13 work going on at INEL on the venting of Mach I containments, A-14 the reliability and usefulness of that kind of process.
15 We have a large amount of work going on at 16 Brookhaven on the HRA aspects of Peach Bottom.
And we have,
17 in addition to that, some additional kinds of uncertainty 18 l analyses in what we call the PRUEP program at Sandia.
19 This feeds into consequence analyses, risk re-l l
20 baselining, selection of risk production options, and value 21 impact analysis.
Actually, at this point, it splits into 22 two points, one of them going into the documentation from 23 Sandia of all of the integrated SARP activities.
But that 24 information is fed into the NUREG 1150 preparation.
25 DR. KERR:
Excuse me.
What is HRA?
O-ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800 336-6646
5840 02 03 18 2
DAV/bc 1
MR. ERNST:
Human Reliability Analysis.
So
)
2 that's generally the activities that are ongoing.
3 DR. KERR:
Did Mr. Ward hear that?
4 MR. WARD:
Sure.
But, like I say, it's a generic 5
human.
6 MR. ERNST:
That's correct.
7 DR. KERR:
Did you have a question, Pete?
8 MR. DAVIS:
Yes.
I don't see any provision in 9
that flow diagram for incorporation of industry-sponsored 10 PRA's or the IDCOR analysis.
Is that something that you're 11 going to neglect as part of this exercise, or does that get 12 folded in along the way someplace?
13 MR. ERNST:
That gets folded in basically in your O-14 uncertainty analyses.
Mark will have more to say about 15 this.
We're picking perhaps a dozen or so major 16 contributors to the uncertainties.
And in coming to an 17 agency position on the range of parameters to be considered 18 for those issues, we do consider the full range of available 19 information, which, as appropriate, would include the IDCOR 20 assumptions as well as other assumptions, yes.
21 MR. BENDER:
I hope that I can be heard.
I'm 22 looking at the chart you have up here.
Three elements of it 23 seem most important.
First, the risk baseline.
- Secondly, 24 the selection of risk reduction options.
Thirdly, the value 25 impact analysis.
ACE-FEDERAL REPORTERS, INC.
202 347-3700 Nationwide Coverage 800-336 4646
5840 02 04 19 2
DAV/bc 1
Is it possible to illustrate how those three
()
2 things are done by some example?
I'll pick one if you want 3
me to.
Let's say that we --
4 MR. ERNST:
Venting is a good one to pick, for 5
example.
Venting of Mach I containments, or filter vents, 6
for example.
7 MR. BENDER:
What things might I have to think 8
about in terms of risk baselining in order to decide whether F
9 venting is a risk reduction option I should consider?
10 MR. ERNST:
In that particular case, in the Mach 11 I vending analysis that we're doing, if venting is useful, 12 then you have to consider can you vent, will you vent at the 13 proper time?
What are the downstream effects of venting, O
14 for example, at Peach Bottom?
15 One thing I believe we learned -- I'm thinking 16 from memory now -- is that if you vented high pressure, the
'l 17 downstream duct work will disappear because it's not made to 18 take that kind of temperature pressure steam flow.
And if I
19 i you had to do this locally, which, in some sequences,'you 20 might have to do locally, the operator wouldn't survive.
21 Now, there's a plant-specific operator 22 consideration.
Therefore, that leg becomes a zero from the 23 standpoint of the likelihood of success.
24 MR. BENDER:
Let me back up for a moment.
25 The Peach Bottom people suggest that early ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-33MM6
i
~
5840 02 05 20 1
DAV/bc 1
venting is one strategy.
I think some people have accepted
()
2 that as a good idea.
The circumstances under which you 3
might use it and the impact that it might have on risk I
4 depends to some degree on the kind of accident you are 5
presuming.
6 MR. ERNST:
That's correct.
7 MR. BENDER:
But, that particular venting, if 8
done early enough, doesn't introduce the kind of hazards you 9
just described.
As a matter of fact, it would be done when 10 the system, when the environment in the containment is still 11 relatively cool and the rate of venting would be relatively 12 slow.
13 Somehow, I don' t get much of a feeling for how to O
14 combine the event and the circumstance.
15 MR. ERNST:
It's not clear to me that that's 16 true.
I think, in some of the accident sequences, you could 17, have some high temperature, high pressure kinde of 18 !
situations with a fairly high steaming rate.
19 MR. BENDER:
Some of them could and some of them 20 !
couldn't.
So, in trying to come to some judgment about it, 21
.it seems to me you're going to have to address a spectrum of 22 events.
23 MR. ERNST:
Exactly what we're doing.
Now, there 24 are cases where you might bypass the pool, for example, in i
j
(-}
25 which case the venting exercise would not accomplish the end j
v l
ACE-FEDERAL RF.'ORTERS, INC.
202-347-3700 Nationwide Ci verage 800 336 6M6
[-
4
5840 02 06 21 1
DAV/bc 1
that you're trying to achieve.
And if you have a residual
()
2 risk from situations where venting really did not help you, 1
3 then one might say, well, does that look like it's going to 4
happen often enough, and is the risk high enough to warrent 5
putting in an additional manmade system, which you might
.i 6
call a filtered vent?
^
7 MR. BENDER:
If I look at the question called 8
risk baseline, what am I doing?
What is that exercise?
1 9
MR. ERNST:
Risk rebaselining is to take that 10 plant as is and try and characterize as best we know how the 11 kinds of sequences that' would occur and the kinds of core 12 melt threats and offsite consequences that would be 13 represented by those sequences.
I. I'T
~#
14' MR. BENDER:
Is there a probable v.alue associated i
15 with each sequence of events?
16 MR. ERNST:
There will De.
i 17 MR. BENDER:
Is there a judgment basis for 18 deciding when it's acceptable and when it isn't?
19 MR. ERNST:
What we're doing is providing 20 information.
Acceptibility becomes NRR's game.
What we 21 provide is the information, the uncertainties, our 22 judgments, the analysts' judgments regarding critical 23 assumptions, factors, and things of that sort, in as 24 scrutable a manner as we can provide it.
25 DR. KERR:
Mal, in connection with the use by i
ACE-FEDERAL REPORTERS, INC.
i 202-347-3700 Nationwide Coverage 800-33M646
5840 02 07 22 1
DAV/bc 1
NRR, will NRR be able to identify what is meant by core melt f~w
(_)
2 frequency as one might use it to compare it with safety 3
goals?
4 I ask this question because it at least has 5
seemed to me that what we mean by core melt is shifting some 6
from the' WASH-1400 days.
At the last ACRS meeting, I asked 7
Mr. Denton what he meant by core melt.
He said he meant 8
that the core had completely melted and had gone through the 9
vessel.
10 I'm not proposing that.
I'm simply saying it 11 seems to me there is a bit of a spectrum in what is being 12 meant by core melt by the various people we're talking 13 about.
()
14' And in order that it be useful, say, in terns of 15l the safety goals, it seems to me one needs to be able to 16 identify in these descriptions something that one can match 17 against what's being talked about in a document like Safety 18 Goals.
19 Is that being thought about?
20 MR. ERNST:
The core melt we're coming up with is 21 to the extent we can predict the PRA's, the kind of core 22 melt you're talking about, where you have a high potential 23 for going through the vessel.
24 DR. KERR:
I'm not talking about a high
("%
25 potential.
Well, high potential still has a probability L)
ACE-FEDERAL REPORTERS, INC.
202-347 3700 Nationwide Coverage 800 336-0646
5840 02 08 23 1
DAV/bc 1
inherent in it, which is unspecified.
()
2 MR. ERNST:
We're trying to consider recovery 3
actions as best we possibly can, even in the core damage 1
4 sequences.
5 DR. KERR:
No.
What I'm trying to find out --
6 I'm probably not asking the question well.
Let's suppose 7
that the safety goal finally says 10 to the minus 4 per year 8
is the frequency for core melt.
Bad core melt or something 9
else.
Will one then be able to go to the rebaselined plan 10 and say:
Here is the thing I should compare with the safety 11 goal?
12 MR. ERNST:
The core melt we have there will be 13 what you should compare.
7_
's_)
14 DR. KERR:
It won't be ambiguous like the 15 potential for core melt.
When you say " core melt" in 1150, 16 it's the same thing, or at least you could compare it with 17 the core melt that might be described?
18 MR. ERNST:
As realistically as we think we can 19 do it today.
20 DR. SIESS:
Realistically what?
Molten core in 21 the vessel?
Or, molten core on the floor?
22 MR. ERNST:
This would be core on the floor.
23 DR. KERR:
Thank you.
24 MR. ERNST:
To some extent, I mentioned "what if"
{
25 kinds of studies to try to be helpful to NRR.
The analysts ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-336-6646
5840 02 09 24 1
DAV/bc 1
are being instructed not only to look at the results of the GIj 2
ASEP analysis but also other things that they know about in 3
their experience that, if it were different in the plant 4
they're analyzing, to highlight that difference for the 5
usefulness to NRR.
6 We feel like it's very important for the analyst 7
who is deeply involved in that process to try and identify 8
the thought process, the what if process.
9 That's, to some extent, where these additional 10 plant studies come about.
11 As far as NUREG 1150 preparation is concerned, 12 this is an NRC document.
However, we do have some 13 contractual assistance in preparation for that document, O
14 the vast majority of that, probM.rly 95 percent or so being 15 in the preparation of the technical appendices.
16 The document itself will be an NRC-written 17 document.
The entire document will be NRC-reviewed and 18 approved.
19 (Slide.)
20 A few words about QA and OC.
This again has 21 boxes similar to the one you just saw and tries to give you 22 a feel about the OA/QC kind of process that's going on hand 23 in hand with NUREG 1150 preparation.
24 I forget the exact number but I think it's on the 25 order of about 25 percent of the total funding is really
("}
\\_-
I l
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 80 4 336-6646
5840 02 10 25 1
DAV/bc 1
being spent in the QA/QC area.
[~')
2 For several sources on the front end, we have us 3
what we call the ASEP senior consulting group.
And there's 4
a OA team also that's looking for consistency of approach 5
through all the six plant analyses.
So that OA and OC, two 6
different groups, comes into play.
7 In the red squares, the containment event tree 8
development was being looked at by several people, expect in 9
the containment event tree or containment performance loads 10 and performance areas.
We have OA/QC activities ongoing by 11 several people heavily involved in source terms, and this 12 takes place in the source term binning activities and the 13 l source term uncertainty analyses.
O
\\/
14
'There's also a OA effort here which should be 15 listed.
We have Brookhaven OA-ing the source term code 16 package runs.
17 DR. KERR:
Mal, when you talk about a OA/0C 18 activity, are you talking about OA/0C for this process, or 19 the effect that QA and QC have on risk?
20 MR. ERNST:
No, the QA/QC for the work being done 21 on NUREG 1150.
It's our internal process to assure a good 22 product.
23 MR. WARD:
Remember, Bill, you always used to say 24 that the NRC didn't have its own QA program; now they have.
25 DR. KERR:
This is being developed at Brookhaven, O
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 804 336 6646
5840 02'11 26 1
DAV/bc 1
I just thought I heard.
And it's only for this project.
()
2 MR. ERNST:
There are a few blocks here.
This 3
one should be a colored block.
There are a few blocks here 4
that do not have a formal OA/OC process involved at the 5
present time.
That basically isn't here.
6 However, we have substantial reviews by ourselves 7
and the staff at NRC and, of course, each lab has its own 8
internal OC/OA process for the work.
9 When NUREG 1150 comes out, it's coming out in 10 draft form.
There will be additional OA/0C so to speak at 11 that time.
12 For the prepacaticn.sf this NUREG itself, we 13 clearly have our own involvement in the technical aspects of O
14 the review.
We'll be writing the NUREG itself with NRR 15 looking over our shoulder and, in some cases, perhaps even 16{
helping out with the writing.
l 17' We do have several consultants to look at this 18 work and knock us on the head if it looks like the work 19 needs some improvement.
20l After publication, we'll go to a public comment i
l 21 process on the document and I intend to run a rather large 22 workshop to critically review the draft before we finalize 23 it.
24 25
(~g
\\-)
l ACE-FEDERAL REPORTERS, INC.
202 347 3700 Nationwide Coverage 80 4 336-6646
5840 03 01 27 1
DAVbw 1
DR. MARK:
Roughly, how many FTEs are devoted to
()
2 that formal 0A/0C ef fort?
3 MR. ERNST:
Roughly 25 percent of our funding is t.
4 supporting that QA/QC effort.
l 5
DR. MARK:
And the funding?
6 MR. ERNST:
We're right about 6 million.
7 DR. MARK:
We're talking about 1.5 million.
8 MR. GILLESPIE:
That's 6 million for you guys.
9 We're putting more into it also.
We're probably in total 10 close to about S14 million, and about 25 percent of that is 11 on quality control.
12 DR. MARK:
Hot dog!
13 MR. GILLESPIE:
We're up in the millions of
\\"
~
14 dollars in the control of quality.
15 MR. WARD:
If that's just called technical review 16 instead of OA, does that make you feel better?
17 DR. MARK:
If it were 10 percent, it would make 18 me feel even better.
19 MR. ERNST:
Some of this is a mixture, because it 20 doesn't just take place at the end.
The OC teams, for 21 example, on a particular plant -- Joe can speak to this 22 better than I, because I recall the QC teams meet several 23 times just on the front end, and in OA, the senior 24 consulting group meets several times on the plant, but early t
25 on, they're trying to help provide guidances to make it more l
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800 336 6646
5840 03 02 28 1
DAVbw 1
efficient, but it is independent input, and it does run sort
()
2 of parallel with the whole process.
3 So it is a mixture between what you've 4
conventionally called end of the line QA and actually 5
production line OC, making the product good as you go 6
along.
7 (Slide.)
8 A couple of words on schedules.
This is the 9
latest schedule.
I can't get it all on there, but you've 10 got copies in your handouts.
Schedules has been a real 11 bugaboo.
We established a schedule last June, which was 12 admittedly simply impossible to reach, because it assumed f-13, that everything went according to Hoyle, and nobody found a
(_/
14 problem.
There were do starting problems, no problems with 15 source term code package runs.
The senior consulting group 16 agreed with everything done in the plant analysis.
Those 17 things just don't happen.
18 I am very interested in a quality product.
These 19 schedules have been slippirg to make room for the quality.
20 I have made some decisions which said don't redo something 21 on the basis of a certain QC or QA finding was not that risk 22 important.
It might have been proved.
The complete 23 validity of the whole thing and be perfect, not even 24 perfect, since this is PRA, and I don't know what perfect
("]
25 means in PRA.
It might have had something that everybody U
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800 33M686
l 5840 03 03 29 1
DAVbw I
agreed was not right.
But why spend another $20,000 doing a
()
2 source term code package run, if the run itself is not going 3
to make much risk difference.
4 So there's been some of those kind of decisions, 5
but where there's clearly a comment that comes back from the 6
process that says improve the product, and that improvement 7
is risk important, it gets done, and that causes a slip.
8 DR. SIESS:
What do you mean " improve"?
9 MR. ERNST:
People agree that, yes, that comment 10 on the modeling for Peach Bottom, for example, is, indeed, 11 true, that the ASEP team, putting that modeling together 12 have made an area, because the plant is really configured 13 this way or has this procedure, or the thermal hydraulic 14 analysis says it really goes this way instead of that way.
15 That kind of improvement..
16 DR. SIESS:
That is correcting errors.
17 Suppose the comment comes back that you did it 18 right, but you made the wrong assumptions.
Why didn't you 19 take this into account, or why didn't you take that?
20 Does that mean go back and do it over?
Do you 21 agree, or what?
22 MR. ERNST:
Joe can speak.
He's on the senior 23 consulting group for this effort.
I think it's both,of 24 these.
It's not just protecting errors.
It's finding out 25 other things, other assumptions that should be considered.
ACE-FEDERAL REPORTERS, INC.
202 347 3700 Nationwide Coverage 800-336-6646
5840 03 04 30 2
DAVbw 1
DR. SIESS:
The final point is that when
()
2 everybody agrees something is okay.
3 MR. ERNST:
Yes.
Or you reach a decision.
I 4
don't think that the process of human nature says everybody 5
is going to agree on everything, but you should at least 6
flag the issue and make a decision.
7 This schedule is probably not even good right 8
now.
I am quite sure there'll be a date on there that's 9
changed.
Even if I had it up-to-date when I caught the 10 Metro down here, it's the best we know it right now.
If I 11 have anything to predict at all, it is that the schedule 12 will slip some more.
We will be finding some more 13 problems.
I'm still. interested in quality work.
If I had O
14 to bet money, I would bet that NUREG 1150 will not come out 15 in the middle of August as advertised.
16 Right now it's iffy.
We can still make it.
But 17 based on past experience, I just can't imagine everything 18 isn' t going to happen beautifully from this point.
19 DR. KERR:
So it might be the end of August.
20.
MR. ERNST:
It might even be into September or i
21' something like that.
22 DR. CATTON:
On the diagram that you had on the 23 previous page, there's a big arrow, which comes from BMI 24 2104 runs.
25 MR. ERNST:
That was input, as well as the ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Covetage 800 336-6646
5840 03 05 31 1
DAVbw 1
original estimate a year ago, ASEP information on dominant
()
2 sequences.
That and other information was used to pick the 3
dominant sequence that should be looked at in what we call 4
binning to determine the sequences that should be run, using 5
the source term code package.
6 DR. CATTON:
Where does the arrow from 0956 enter 7
the matrix?
8 MR. ERNST:
As a matter of interest, I don' t 9
believe that there's a single sequence in 0956 that is being 10 used in 1150, because of minor changes in how the sequence 11 goes, based on the rebase lining work and other 12 information.
So I don't think there's a one to one 13 correspondence between any sequence we have now and what's O
14 in 0956.
My understanding is that the revised 0956 will 15 have the source term code package information.
16 DR. CATTON:
But it won't be ready for you to do 17 anything in time for August, will it?
18 MR. ERNST:
Yes, they're using our stuff in 19 0956.
They're going to use the NUREG 1150 in fo rma tion.
And 20,
I think we're through now with all the source term code 21 package runt That information will be in 0956.
22 DR. KERR:
I'm confused now, because I thought 23 0956 was supposed to be the basis for 1150.
Now I seem to 24 be hearing that 1150 is going to be the basis for 0956.
fs 25 Is that the current status of things?
U ACE-FEDERAL REPORTERS, INC.
202-347 3700 Nationwide Coverage 800 336-6646
1 l
5840 03 06 32 i
l 1
DAVbw 1
MR. ERNST:
They probably should have let
()
2 somebody from 0956 speak to that.
3 MR. GILLESPIE:
Mal, let me.
0956 describes a 4
modeling package, a code package.
The original 0956 was the 5
original BMI 2104 suite of codes, of which 19 runs were 6
done.
Some time has gone by.
We've gotten criticism on 7
that.
Ther was too much manual interface between different 8
models in the BMI 2104 suite.
That was basically 9
streamlined and changed to what is today a source term code 10 package, which the revised 0956 is going to describe.
It's 11 going to describe a different code package than was in the 12 original 0956.
It will describe the source term code 13 package you used today.
That source term code package is O
14 there and running.
15 The fact that we haven't got it written up yet in 16 0956 does not preclude 1150 from using that code package.
17 So they're off using it now in the new 0956, where we're 18, doing some things now to translate or make a step from the i
19 original code package of a year ago to the current code 26 package to show the relationship of the changes and what j
i 21 differences they made.
And the new 0956 will, in fact, 22 describe the code package which is being used in 1150.
So 23 it's a set of documents.
One feeds into the other.
But 24 1150 took the code package and went off and made their 25 runs.
ACE. FEDERAL REPORTERS, INC.
202 347 3700 Nationwide Coverase 800 336-6646
i l
i 5840 03 07 33 1
DAVbw 1
They're not dependent on 0956 doing exercising
()
2 the code iteratively for them.
j 3
Now for our examples in 0956, you remember 4
chapters 3 and 4 had some codes and examples.
Those curves
(
i 5
and examples will now be drawn from the approximately 30 6
runs of the source term code package that are being 7
exercised iteratively to support 1150.
8 So the code package is not dependent on 1150.
9 We're going to use some of the work done for 1150 as part of j
10 our illustration in 0956.
11 DR. KERR:
So if one, reads in the current version 12 of 0956 that the Staff now recommends that this be used for 13 all the calculations, that was true when it was released, C) 14 but I can disregard
- t. hat at this point?
15 MR. GILLESPIE:
No, because we're probably going 16 to show the difference between BMI 2104 's original package 17 and the source term code package as it exists now.
It is a 18 matter of efficiencies, feedback, loops in thermal 19 hydraulics to make more consistencies.
There were some 20 internal inconsistencies in conservation.
21 Help me out.
I'm past my second sentence.
22 MR. MEYER:
Ralph Meyer, from Research.
+
23 We talked about these before, but just to remind 24 you quickly.
There were several cases where in different 25 parts of the original BMI 2104 suite of codes, we were using L
ACE-FEDERAL REPORTERS, INC.
202 347-3700 Nationwide Coverage 80 4 336-6646
5840 03 08 34 1
DAVbw 1
different models to calculate the same thing.
It was a core
(( )
2 concrete interaction model called INTER that was in the 3
MARCH code.
Then there was CORCON Block 1, used external to 4
the MARCH code.
The same set of calculations.
There was an 5
in-vessel fission product release model called FP LOSS in 6
the MARCH code.
J 7
Externally, we were using CORSON in the same set 8
of calculations at dif ferent times.
We were calculating the 9
same phenomenon.
10 DR. SAUNDERS:
So how did you make them 11 consistent?
Take the two outputs and average the two?
12 MR. MEYER:
What we've done for the BMI 2104 13 suite of codes, what we attempted to do was to tolerate the O
14 model that we thought was not the best choice in areas that 15 we believed were not very critical.
i 16 In the final code package, we have eliminated the 17 models that are not the best choice and have only a single
~
18 model for doing the calculations.
I 19' DR. CATTON:
Actually, I was trying to ask a much 20 simpler question.
I understand what BMI 2104 is and 21 what NUREG 0956 is, but there have been a lot of criticisms 22 of some of the phenomonology and the impact that certain 23 things may have on certainty.
And I'm not talking about 24 hypercube sensitivity.
That's not what I'm talking about.
25 Some of these criticisms are fairly well written ACE-FEDERAL REPORTERS, INC.
202 347-3700 Nationwide Coverage 80433M446
5840 03 09 35 1
DAVbw 1
out in the ACRS letter.
()
2 Before you draw an arrow into this matrix, don' t 3
you have to address those uncertainties?
4 MR. MEYER:
We certainly do address those 5
uncertainties.
And let me say that the criticisms have been 6
raised in a number of areas which we call issues.
We're 7
quite forward about this.
We list issues where people have 8
taken different opinions.
9 You may believe that your opinion is correct, and 10 therefore, that we should confirm to your opinion on a 11 certain subject.
12 We don't necessarily believe that just because we 13 receive a call that it is correct and our position is
()
14 wrong.
15 We are investigating all of those issues.
We'll 16 discuss the status of the resolution, our attempt to 17 resolve those issues in the final report.
We are not going 18 on oblivious to that, but neither are we coming to a halt, 19 because there are differences of opinion in several 20 important areas.
21 I'd like to make one comment on a point that you 22 raised, Dr. Kerr, and that was, you said in our draf t 23 version that we recommended the use of this code package, 24 and now we have a new code package.
I think the implication 25 was, should you disregard our previous statement.
O ACE-FEDERAL RI: PORTERS, INC.
202-347 3700 Nation aiue Coverage 800-336 6646
~
5840 03 10 36 1
DAVbw 1
DR. KERR:
It was a question.
()
2 MR. MEYER:
The fact of the matter was, in the 3
draft report when we made that statement, we had already 4
introduced the notion of the code package.
At that time, we 5
knew what changes we planned to make, in order to resolve 6
the egregious errors, the things that could be fixed, that 7
ve all agreed needed to be fixed, and the recommendation was 8
made with reference to the code package that was to be 9
assembled from the BMI 2104 suite of codes.
10 DR. KERR:
I misread the 0956 then.
11 MR. GILLESPIE:
The question is, will the final 12 reflect those words?
The final will probably have those 13 words in it, but we'll change it back to the words.
The new O
14 ccde package will be there rather than the year old BMI 2104 15 suite.
16 DR. KERR:
My misunderstanding.
17 Yes, Mr. Bender.
18 MR. BENDER:
If I could go back to Dr. Catton's 19 question for just a minute.
I think-I would certainly agree 20 with you, Ralph, that just because something is criticized 21 does not mean you have to respond in a direct way and change 22 what you're doing because of the criticism, but it does seem 23 to me that at the front end, you have to say something about 24 the sensitivity of the result of the criticism, and that 25 ought to be out there before you go through these expensive O
ACE-FEDERAL REPORTERS, INC.
202-347 3700 Nationwide Coverage 800 3M-6646
_. - = - _ _ - _ - -
5840 03 11 37 l
1 DAVbw 1
exercises.
And I would think you'd plan to do that, but you i
2 didn't say.
3 MR. MEYER:
What is being done in 1150 that was i-j 4
not done in 0956 is an explicit accounting for the related 5
uncertainties in the analysis.
Now we reported on an 6
uncertainty study which maybe it was a sensitivity study 7
called Quest, where three sequences had been looked at to 8
try and get some idea of the degree of uncertainty resulting 9
from these various controversial areas.
i 10 That was very selective and only addressed three l
11 sequences.
It was very preliminary.
We have done two j
l 12 things because of the importance of that.
l 13 We agree with you on the importance of that.
One
(
14 is for the risk rebaselining work that's being done.
There 15 is built right into the program an uncertainty analysis.
So 16 for every case that they do, they will have an estimate of j
17 uncertainties.
18 In addition to that, we have started a new 19 uncertainty study at Brookhaven, which will try and evaluate 1
20 the uncectainties a little more rigorously than we were able 21 to do new in the risk rebaselining work.
i 22 So really, we're doing everything that we think 23 is possible on this uncertainty analysis.
2 24 l
(:)
i I
i l
ACE. FEDERAL REPORTERS, INC.
I
.,. _.... ___ _ _ _ __,. _ _ _. __ _ _,._ _ 3 3MM6 202 347-3700 Nationwide Coverase 800
5840 04 01 38 1
DAV/bc 1
MR. BENDER:
I didn't intend to express the
()
2 opinion that you didn't, but the sequence for that doesn't 3
come out very well.
How much of the evaluation of 4
criticisms is going to be out there before you go through 5
this extensive numerical analysis, which you're going to do?
6 It seems to me you ought to make that evaluation 7
early, put it before whomever it is that's judging and get 8
some response from them.
9 MR. MEYER:
I've seen the comments we have 10 received from the general public.
They're quite 11 voluminous.
I haven' t read quite all of them yet.
- But, 12 Jocelyn, you can correct me if I'm wrong.
I think we've 13 flushed out most of the important technical criticisms early CE) on in our peer reviews.
14 i 15 We're getting some additional information.
We 16 certainly are making changes in our 0956 report in response 17 to these comments.
18 But, in terms of the basic technical areas, I 19 think we've pretty much had them identified from our peer 20 review process.
I 21 So I don't think it's premature to go through the 22 calculational exercise.
It's true, if you were to look at a 23 graph of knowledge versus time, we're not at some plateau up 24 at the top where we know most of what there is to know.
r 25 And so what we do will soon be outdated.
l
\\-))
i ACE. FEDERAL REPORTERS, INC.
202-347 3700 Nationwide Coverage 800-336-6646
- -. - - - - - - ~. _ - -
5840 04 02 39 1
DAV/bc 1
MR. BENDER:
Doing some analysis, I would not
()
2 quibble with.
But doing some analysis is not the same as 3
systematically going through the criticisms and saying:
4 If this assumption is incorrect, the result of 5
the analysis might be sensitive to it in terms of its end 6
evaluation to some degree because...
7 MR. MEYER:
I think we are doing exactly what you 8
said.
There was a meeting at Sandia just about three weeks 9
ago where we assembled the panel of experts that we've been 10 using for the source term work and some other people, where, 11 systematically, through the containment event trees and 12 through the source term work, we listed exactly the issues 13 of controversy you're talking about; indicated what we O
14 thought the range of variation of the key parameters would 15 be, polled the experts to see what weight they would put on 16 the various values in the range in order to try and perform 17 the uncertainty evaluation that's going on.
18l So I really think we are doing exactly what,you 19 suggest.
20 MR. ERNST.
We are trying to mesh in here pretty 21 closely that process as described in the uncertainty letter 22 that I sent to Jack Hickman.
I think there's a copy passed 23 out on that and you'll hear more words from Mark Cunningham 24 on this process, too, or from Joe.
Well, you're going to 25 talk about the issues though.
So I think you'll hear more ACE. FEDERAL REPORTERS, INC.
202-347 3700 Nationwide Coverage 800 336-6646
U 5840 04 03 40 1
DAV/bc 1
about that.
2 I think we are offering it in an integrated 3
fashion.
4 DR. KERR:
Please continue.
5 MR. ERNST:
Let me do something a little bit out 6
of sequence here.
7l (Slide.)
8 Let me say a couple of words about the future 9
schedule.
I still say August '86 down there for the 10 futures, but I have also said that if I had to bet a 11 paycheck on it right now, I wouldn't bet a paycheck on 12 August '86.
It's still possible but not likely, no matter 13 what kind of degree I put in it.
14 The technical report should be finished and 15 [
depend on the report or the plant in the May-September kind 16 l of timeframe.
We would look for public comment in the 17 September-November kind of timeframe.
We have a workshop 18 currently planned for early fall.
A final NUREG 1150 19 considering all public comment likely in the spring of '87.
20 There are currently plans to do B&W and CE.
B&W 21 should be of extreme interest because of the less forgiving 22 nature of thermal hydrolics in the plant.
23 Bailt into the process is a continuing look at 24 results of new research and the containment of the core 25 concrete area and, as warranted, an update of 1150 if it ACE-FEDERAL REPORTERS, INC.
)
202-347-3700 Nationwide Coverage 800 336-6646
5840 04 04 41 1
DAV/bc 1
looks like substantial changes are occurring that would make 2
big differences in the conclusions we draw.
4 3
Then one would consider updating 1150.
At the 4
present time, I had only planned on analyzing these two 5
plants and providing a supplement that gives the reactor 6
information about those two plants.
7 I did have some slides on the content of 1150.
8 We've gone into some of that already in sort of peripheral 9
conversations here.
What I'd like to do is go on with Joe's 10 talk on the methodology, and I'd like for him to start with 11 the source term, because I think we're going to lose some 12 people from DAE.
13 So it's sort of in the middle, but we' ve been 14 talking about source terms so maybe we could sort of wrap 15 that subject up.
Then Joe could come back to the beginning 16 and talk about-the accident sequences, and then Mark can l
17 talk -- Joe will also be talking about uncertainty.
And 18 then Mark will have some words to say about the specific 19 issues we've identified and this range of the issue and 20 degrees of belief that Ralph's been talking about.
21 Then, time permitting, if I could come back for 22 more on the contents of 1150.
But I think you can read the 23 detailed outline, so maybe that part will just be questions 24 and answers, on 1150.
25 Does that suit?
ACE-FEDERAL REPORTERS, INC.
202 347 3700 Nationwide Coverage 800 336-6M6
5840 04 05 42 1
DAV/bc 1
DR. KERR:
Mr. Davis.
()
2 MR. DAVIS:
Can I slip in a quick question before 3
you go on?
4 I was wondering what the consequence would be of 5
missing this schedule.
Is there some pivotal regulatory 6
decision that depends on 1150 output, or some other policy 7
statement, like the severe accident policy statement or the 8
safety goals?
9 MR. ERNST:
No, it's not safety goals.
10 Basically, severe accident policy, the original schedule was 11 set to try to make as good an interface as possible between 12 the ongoing work between NRR and IDCOR to develop a 13 methodology of extrapolation to other plants, r~N hi 14 We have been able to provide some information, 15 l i
clearly not as timely as I would like, and clearly not as 16 much as Zoltan would like, but we've been trying to do the 17 best we can.
That was the driving thing.
18 l The second driving thing, and maybe Zoltan has a i
19 l word to say about that, is there are some times listed in i
20 the severe accident policy for coming to the closure and 21 getting on with the process.
So,' clearly, that drives, 22 too.
23 There's a strong inclination to try to take the 24 extent of knowledge that we have and see whether or not some 25 regulatory changes need to be made.
The position right now 7s(A ACE FEDERAL REPORTERS, INC.
202-347 3700 Nationwide Coverage 800 336-6 4 6
5840 04 06 43 1
DAV/bc 1
is we need to see what draft 1150 says before we can really
()
2 start doing anything useful in the regulatory process.
So 3
that's clearly a driving force.
4 I don't know whether Zoltan has anything to add 5
to that or not.
6 MR. ROSTOCZY:
Basically, it's the same thing,
7 that it would depend on it in two senses.
One of that is 8
the result, the information that we can use without having a 9
printed report.
Again, when it becomes available, we intend 10 to use it, so that's one dependence, 11 The second dependence is the public comment 12 aspect of it; 1150 is supposed to put out for public comment 13 the methodology of these studies that are being done and the
(')
k 14 actual calculations for the six reference plants are 15 supposedly to be put before the public to see, study, 16 comment on, whatever.
17 For that purpose, we need the completed report 18 whenever it is ava,ilable for public comment and when those 19 comments come back is when we have one of our inputs for our 20 work.
21 MR. DAVIS:
Thank you.
22 MR. MURPHY:
Mr. Chairman, with your 23 forebearance, I'll sit instead of stand.
I either have one 24 ligament in my knee that's a lot longer than it was a couple 25 of hours ago or I've got one that's in two pieces; I'm not O
ACE-FEDERAL REPORTEPS, INC.
202 347-3700 Nationwide Coverage 800-336-66 4 1
5840 04 07 44 1
DAV/bc 1
sure which.
I hit a patch of snow on my way in here, and 2
there's no way I can stand for an hour.
3 (Slide.)
i 4
I'll try to describe the methodology we're going a
5 through, both in terms of what we're doing and how we're 6
doing the uncertainty analysis, and then proceed.
But I'll 7
take it in a somewhat disjoint fashion to meet the needs of 8
our people by starting with what we're doing on the 9
containment of entries and the source terms, and proceeding 10 then into a discussion of uncertainty as it applies to those i
11 features.
12 And then I'll come back and discuss the so-called 13 front end, the risk rebaselining of the accident sequence 14 frequency and go from there to what we're doing in the 15 consequence modeling.
So it's not necessarily the logical 16 flow of information.
17 May I have the next one now?
18 (Slide.)
19 On the containmant analysis, we're developing the 20 detailed containment event tree for the reference plant.
i 21 We're actually in this reduction program.
22 I'll show you an example of the type of questions 23 that are on these trees.
These are very large trees.
They l
24 have undergone peer review by a team put together by Mike 25 Cordini of the University of Wisconsin.
i ACE. FEDERAL REPORTERS, INC.
202-347 3700 Nationwide Coverase 800 33 H 6 4
5840 04 08 45 1
DAV/bc 1
The containment analysis considers the accident T~S 2
progression, the fission product release, migration and w-3 removal mechanisms and phenomenology uncertainties and 4
process-induced failures.
5 Implicit in developing the tree is the 6
consideration of the IDCOR and SASA issues.
As we get into 7I the uncertainty analysis, part of this...
8 (Slide.)
9
...this is a very abbreviated example of what a e
10 containment event tree looks like.
They're what we call 11l virtual trees.
They're done in a generic way that can apply 12 to any plant of a given type.
13 They begin with a number of questions on the tree p
kJ 14 that specify the conditions under which you are entering the 15l tree.
So, to a large extent, the first 20-30 questions on I
16j the tree have to do with the finding of boundary conditions 17lj in the analysis.
I 18j For this reason, the trees are very large.
You 19l have more than two possible choices and you have well over i
20l 30 or 40 questions.
It doesn't take long to realize that, i
21 too, that 30 or 40 is a rather large number.
22 The trees are not as complex as that sounds, so 23 many of them are definitions of boundary conditions.
The 24!
questions that are asked on the tree are questions like:
l 25 I the state of the plant.
Is the vessel depressurized?
Are 7-s t
)
w-ACE FEDERAL REPORTERS, INC.
202 347 3700 Nationwide Coverage 800 336 6646
5840 04 09 46 1
DAV/bc 1
the drywall sprays operated?
Is the containment acutuated?
2 That is, dry wells.
3 Some of these are questions on the performance of 4
the operators.
Some of these are defined by the basic input 5
conditions of the accident sequence hereon.
It's worth 6
noting that, in some cases, these conditions are not 7
accident sequence specific even, but rather cut set 8
specific.
There are different ways you can get into the 9
various accident sequences.
10 All that has to be factored into your analysis of 11 the event tree.
Basic questions of the event and timing of 12 the plant conditions and the various phenomena involved.
13 What is the containment failure pressure?
How do you handle 14 it?
What kind of delta P do you get due to hydrogen 15 burning, and that sort of thing, were all represented in I
16 this containment event tree.
17 The next one now.
18 (Slide.)
19 We've used a variety of sources of information 20 for this.
Certainly, a lot of the work that's coming out of 21 the source term code package and the various codes that 22 we've used were also obtaining information from the 23 containment loads working group and the containment 24 performance working group.
The various analyses that have 25 been done under SASA.
Assorted steam explosion review ACE FEDERAL REPORTERS, INC.
202 347 3700 Nationwide Coverage 800 336 M46
5840 04 10 47 1
DAV/bc 1
groups.
Generic issues studies.
2 Essenti'lly, everything that's been done that can 3
provide us information and understanding, how to work our 4
way through the containment event trees and what the 5
possible steps through those trees are is being factored in, 6
as I'll describe when I get to the uncertainty analysis.
7 The weights we put on various portions of the 8
tree and how we progress through are handled as we consider 9
the uncertainty in each of these basic elements, if you 10 will.
11 MR. BENDER:
Joe, are we still operating on the 12 basis that we're analyzing the plants as they are, or are 13 you analyzing the plants as they might be?
14 MR. MURPHY:
We are analyzing the plant as it is.
15 MR. BENDER:
What help is that?
16 MR. MURPHY:
There's one minor exception to that.
17 MR. BENDER:
What help is that going to be?
18 MR. MURPHY:
I think you have to know where 19 you're starting from.
One of our first steps as we do the l
20 accident sequence rebaselining, and I'll get to that as I 21 talk to that, is to send a team of people to a plant for a 22 week to find out what the as built drawings are, what the 23 actual procedures are, how does the plant implement the 24 procedures?
l 25 I don' t think you can really say you know what g
U ACE FEDERAL REPORTERS, INC.
202 347 3700 Nationwide Coverase 800 336-6M6
5840 04 11 48 1
DAV/bc 1
the current status of the plant is until you've done that.
(}
2 MR. BENDER:
I think that's what the operators 3
might ought to be doing.
It seems to me, expressing, 4
obviously, just an opinion, that what the NRC ought to be 5
doing is trying to find out how the answers might be 6
influenced by certain pieces of technological treatment that 7
could be included in the sequence.
To look to see whether a 8
plant is vented or not is one thing.
It might be better in.
9 my mind to say what could be the effect of the thing if it 10 were done?
11 MR. ERNST:
We'll have both.
12 MR. BENDER:
That's what I'm trying to find out.
13 What you've told me so far is you're going to see what the 14 plant is as it exists and do the second part, which is to 15 look at the sensitivity of it, the changes in design, that 16 seems to me to require some kind of criterion for judgment.
17 I don't know what the criteria are.
18 l Do you?
19 MR. MURPHY:
I'm not sure I understand your 20 question.
There's essentially two steps in the analysis.
21 One is to find out what the risk of the plant is today.
22 That's as best we can within the resources we have.
That's 23 the process I've been describing so far.
24 Then there's an evaluation of risk reduction 25 options.
What options could you take that would reduce the ACE FEDERAL REPORTERS, INC.
202 347 3700 Nationwide Coverese 800 3364M6
5840 04 12 49 1
DAV/bc 1
risk?
}
2l Now, some of these come...some of these can be
/
3 handled as sensitivity studies off the basic model.
Others 4
require changes in the model to implement.
What would you 5
do if you installed the new feature X -- perhaps it's a 6
bunkered RHR system, or something of that sort?
7; You could also take the models and preturb the 8l models and rerun the case and see what value you'd gain from I
9i various risk reduction options.
That is also being done.
10 DR. SAUNDERS:
Let me understand this.
If you 11 4 have a large fault tree or event tree that's specific to a 12 plant, you know the boundary conditions or the range of 13!
variation of every input.
You calculate then the associated
(~N l
sJ 14 risk through your probabilistic analysis.
i 6
15l Now you come into an analytic question which I
16l says:
What can I do for, say, a given amount of money which 17i will reduce the risk the most?
i 18l If your model is correct, that's an analytic 19l problem, what are you doing simulating things?
I don't 20!
understand that.
21 22-l 23l l
24l 25 7,
(
)
v ACE. FEDERAL REPORTERS, INC.
202 347 3700 Nationwide Coverage 800-336- % 5
5840 05 01 50 1
DAVbur 1
MR. MURPHY:
What you have to do is rely on the
(}
2 judgment of your analyst as well as the history of the sort 3
of things going on around the country, things that the 4
committee has suggested in the past, and go back and perturb 1
5 your models and see what would the risk be if I changed the 6
design of the plant to incorporate some option for risk 7
reduction.
8 That way you evaluate what you get if you 9
implemented that option.
10 DR. SAUNDERS:
Yes.
My question is:
why isn't 11 it an adequate procedure to try to say, given this amount of 12 money, what can I do to the plant to reduce the risk the 13 most?
14 MR. ERNST:
That is basically what we are doing.
15 We are using several things as criteria, one being the $1000 16 a person rem set forth as a safety goal.
17 The display hopefully will show whether or not 18 certain risk changes -- how they match up with the $1000 a 19 person rem, for example.
20 MR. MURPHY:
I will have a slide in a little i
21 while that discusses the way we are doing the risk reduction 22 and risk rebaselining.
That may help a little bit.
23 DR. SAUNDERS:
All right.
24 MR. DAVIS:
Joe, what do you do about changes 25 that have not been made at the plant but have been
[
i ACE. FEDERAL REPORTERS, INC.
202 347 3700 Nationwide Coverage 800 336 4646
5840 05 02 51 1
DAVbur 1
mandated by the NRC and are scheduled for implementation at
(}
2 the next refueling outage, for example, like the Mod 3A ATWS 3
situation?
i 4
MR. MURPHY:
With the exception of the ATWS 5
fixes, we are nodeling the plant as it is the time we visit 6
it.
With the ATWS fixes, we are including what they are 7
pennitted to do.
8 MR. DAVIS:
Are you going to acknowledge in your 9
report that other changes are coming?
You know, you may 10 conclude that the plant needs to have some fix put on it, 11 but with changes that are being mandated that conclusion may a
12 not be valid.
13 Do you see what I am saying?
I 14l MR. MURPIIY:
I see what,you are saying.
\\-
15 MR. DAVIS:
You may find out that the B sequence 16 dominates risk.
17 MR. MURPHY:
I can't promise that we are doing it 18 on each and every thing.
There have been so many things 19' that are being backfitted.
20 But for the major items that we are aware of from 21 our team going to the plant and talking to the plant as to 22 what they plan to do in the near future, many of those will 23 be treated as sensitivity studies.
24 But the base case is essentially based on the 25 plant as is, with the exception that the ATWS fixes are ACE-FEDERAL REPORTERS, INC, 202-347 3700 Nationwide Coverage 800-336-6646
-... ~.-
5840 05 03 52 1
DAVbut 1
considered to be installed and operated.
(}
2 MR. ERNST:
As is?
If you have something that is 3
going to be installed at the next shutdown, you have the 4
drawir.gs where you have the procedures that are going to be 5
implemented, and they have been instructed to consider that 6
as an accomplished fact.
7 If they don't have the drawings, don't have the 8
procedures, we don't want to guess how the thing is going to 9
be implemented.
10 MR. MURPHY:
We have essentially the final design 11 and the procedures to operate the system, but if we hear 12 there is going to be a redesign of System X two years from 13 now and all we can hear is general statements as to what the 14 purpose of that design is and we don't have PMIDs and 15 electrical drawings and operating procedures, we assume it 16 doesn' t exist.
17 DR. KERR:
Well, now under the current operating 18 procedures, if it is a backfit, then the calculations of 19 risk reduction have already had to be made, haven't they?
20 '
MR. DAVIS:
Not in all cases.
21 DR. KERR:
Well, if it is a backfit mandated 22 today, that would be the case, wouldn't it?
23 MR. MURPHY:
A backfit mandated today.
But 24 there's newer TMI fixes in the mill.
25 DR. KERR:
I have been told it is impossible to O
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-336-6646
u
~
5840 05 04 53 l
1 DAVbur 1
calculate risk reduction due to TMI fixes.
(}
2 DR. SAUNDERS:
By whom?
3 DR. KERR:
On several occasions by members of the 4
NRC staff.
5 MR. DAVIS:
I think this may be a bigger problem 6
than you may be aware of right now.
7 For example, B&W is putting an electcic driven 8
aux feed pump now into their Davis-Besse plant.
I don' t 9
know how you would consider that if you were choosing their 10 plant at this point in time.
But it would make a big 11 difference, I suspect, on the core melt problem.
12 DR. CATTON:
They are also planning a high power 13 vent of rather large size.
14 MR. DAVIS:
Plants always have these things they.
15
,are thinking about doing, and at least when I have talked to 16 them there is always -- I know you have got to draw the 17 line.
18 MR. MURPHY:
At some point in time you have to 19 fix the design and progress with your analysis or you never 20 get off of go.
21 MR. DAVIS:
It is a problem, and I don't know the 22 answer.
23 DR. KERR:
This is a living document, Pete.
24 DR. SAUNDERS:
What he says is reasonable.
25 (Slide.)
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800 336 6M6
l 5840 05 05 54 1
DAVbur 1
MR. MURPHY:
In terms of what we are doing with I'T 2
the source term, I think we have covered most of the V
3 material already.
The methodology is essentially the same 4
as that used in BMI-2104 and the N-0956, with some 5
improvements.
6 I think Ralph has already covered the points that 7,
I have here.
8 The Surry analysis will rely heavily on what is 9
presently N-0956.
There were only a few source term code 10 package runs to support the analysis of the Surry plant.
11 The other plants will rely primarily on new runs, 12 using the source term code package.
13 Basically, our tool for the source term 14 methodology is the source term code package.
)
15 MR. DAVIS:
Joe, on that Surry analysis, what do 16 you assume about whether or not the RHR pipes are underwater 17 for the feed sequence?
18 MR. MURPHY:
We are treating it 19 probabilistically, based on a judgment call, if you will, a 20 Bayesian analysis as they get to the uncertainty.
21 Basically the drawings were looked at by a team 22 of five experts, and looking-at the situation, a subjective 23 estimate was made as to whether it would be underwater or 24 above water.
That weight is factored into the uncertainty 25 analysis.
l ACE-FEDERAL REPORTERS, INC.
202-347 3700 Nationwide Coverage 800 336-6646
5840 05 06 55 1
DAVbur 1
As I recall -- I don't recall the number, but I
{~}
2 think they were leaning towards about a 75 percent chance 3
that it would be underwater and about a 25 percent that it 4
would be above water.
5 MR. DAVIS:
That is an important number, I 6
suspect.
7 MR. MURPHY:
I expect so, too.
8 (Slide.)
9 I think again Mal pretty well covered the binning 10 process.
11 Basically, the way we get from the front-end 12 analysis through to the exercise of the codes is by many 13 accident sequences into -- for want of a better word -- bins 14 or groups that have the same characteristics in terms of how 15 we would expect the accident phenomenology to be.
16 We start off by identifying the accident, the 17, dominant accident sequences associated with the rebaselining 1
18l effort to the estimate of the core melt frequency.
We take 1
19; the containment of entries that these sequences feed and 20 reduce those to the key containment failure modes and then 21 proceed on to say which of these situations have we had 22 source term code package for 0956 runs that describe this 23 process.
Are you aware that we need more analysis?
24 This is a judgment call because every sequence is 25 slightly different.
The binning process is an amalgamation ACE-FEDERAL REPORTERS. INC.
202-347-3700 Nationwide Coverage 800 336 6646
5840 05 07 56 1
DAVbur 1
that causes you some uncertainty, and we rely on the
()
2 judgment of a team of experts from Sandia, Battelle, 3
Brookhaven, Oak Ridge, as well from NRC.
4 Coming out of this. we get release 5
characteristics for each bin.
This means whether the 6
primary system is pressurized or not, whether the sprays are 7
operating or not, whether you would expect the core to melt 8
before the time of containment failure or after -- these 9
type of characteristics -- and then the source term code 10 package is run.
11 Those runs are deemed to be representative of all 12 the sequences which get into that bin.
Typically, we wind 13 up with something about 15 bins per plant.
(:)
14 DR. KERR:
Joe, what is meant by a dominant.
15 sequence?
Is that based on risk?
16 MR. MURPHY:
It is based primarily on core melt 17 probability, but recognizing that the sequences that 18 dominate core melt don't necessarily dominate risk, we look 19 at everything that is greater than a 1 percent contributor 20 to the core melt probability.
And therefore, I think it 21 does -- that discrimination is low enough that I think we 22 can pick up the risk as well.
23 (Slide.)
l 24 Proceeding onward, let me talk a little bit about 25 sensitivity and uncertainty analyses.
O ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800 336-6646
5840 05 08 57 1
DAVbur 1
Our objective here is to provide a good
/~T 2
engineering perspective on the modeling assumptions that are U
3 driving the analysis and to provide a reasonable envelope in 4
which the actual value would likely be found.
We do not 5
feel it necessary to express this in terms of a formal 6
statistical bound.
7 In other words, what we are trying to express is 8
an engineering judgment, calculated using statistical 9
techniques in many ways, but recognizing the impreciseness 10 of our input data.
We are not trying to assign these 11 results as prec~.se 5/95 percent bounds or anything of the 12 sort.
We don't believe they are defendable as such.
13 Rather, we are saying that we believe coming out
\\
14 of this process these are reasonable engineering bounds that 15 express a range in which we think we will find where the 16 outcome would lie.
17 If you have more confidence in the statistical 18 methods than I do, you could probably interpret these in a 19 statistical sense because they are calculated formally.
~20 I would caution against that, however.
I don't 21 think our input knowledge is based on a degree of belief 22 type of assessment.
I am far from convinced that if I asked 23 an engineer for his 90 percent degree of belief that I would 24 get a different answer than when I asked him for his 95 25 percent degree of belief.
qLJ ACE-FEDERAL REPORTERS, INC.
202 347-3700 Nationwide Coverage 800-336-6646
5840 05 09 58 1
DAVbur 1
What I can get from.him is what he thinks is a
()
2 reasonable engineering upper bound, and he doesn't know 3
whether it is 90 or 95, and I don't either.
4 So we are treating it in this general engineering 5
approximation sense rather than in a formal statistical 6
sense.
7 DR. KERR:
I think it is entirely reasonable.
I 8
don' t see how you can do anything else.
9 But what I would hope is that where you deem A
10 things to be fairly important, that you give a little bit of 11 discussion as to how this process might have developed.
12 It seems to me there will be some engineering 13 I
uncertainties on wheezier grounds than others.
It is
/^
l 14 helpful to people that are trying to judge that have some' 15 feeling for the degree of uneasiness that might accompany a 16 particular decision.
17 MR. MURPHY:
We are not at a point where I can 18 give you a specific example yet.
I have a backup slide that 19 gives sort of a theoretical example of the hypercube.
If 20 you want to take the time, I can show you how it works, but 21 we are not at the point where I can give you a specific 22 example yet in terms of our understanding of the uncertainty 23 analysis.
24 (Slide.)
25 What we are using to handle the uncertainties is ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 600-336-6646
5840 05 10 59 1
DAVbur 1
the limited Latin hypercube method.
This is a constrained
(~')
2 or stratified Monte Carlo sampling scheme.
w/
3 Now, feeding into it is not just, say, the source 4
term code package.
5 DR. KERR:
Excuse me.
Is this connected somehow 6
with the earlier statements about good engineering judgment?
I 7l MR. MURPHY:
Yes.
This is the way we are doing 8
it.
9 MR. ERNST:
You use your engineering judgment to 10 get at your ranges and degrees of belief.
That is what Joe 11 meant by then you take this and put it into a statistical 12 sampling process.
I 13!
MR. MURPHY:
The problem you get is that you have
'N 4
l 14 a very, very complex problem.
I 15l Now, I can give you one thing, and I could get a 1
16l team of experts together in the room and say let us look at 17 the bottom line number and everybody give me an estimate as 18 to what the uncertainty would be.
We could do that.
I 19 don't think it would be very meaningful.
20 But instead what we are doing is looking at the 21 major parts of this containment.
In particular, now we are 22 talking about the containment of the source term analyses, 23 and we are trying to develop 10 to 15 issues which we think 24l drive the uncertainty in our ability to predict what 25 happens for each one of those 10 to 15 issues, going through
>73!
.%d ACE-FEDERAL REPORTERS, INC.
I 202-347 3700 Nationwide Coverage 80 4 336-6646
5840 05 11 60 1
DAVbur 1
a process where we try to estimate what the probable
(}
2 possible range on this phenomenon is.
3 One of them might be direct heating.
That is an 4
issue in itself.
Mark will talk about the specific issues 5
later.
6 But we will take each of these issues and say 7
what is the range -- not only what does the source term code 8
package predict for this, but what is a reasonable range, 9
considering the work that has been done, on request, the 10 IDCOR results, everything else we know.
11 Bring people together in a room and try to 12 estimate potential values for these phenomena as they could 13 occur.
14 DR. CATTON:
Could I maybe have ycu give me an
.5 l example?
16 The CORCON/VANESA get tied together.
One thing 17 they found, that the temperatures that CORCON predicted were 18 too high.
This leads to a particular kind of a tack, and it 19 leads to some conclusions about the goodness or badness of 20 VANESA.
21 The KFK tests show that the temperatures are 22 lower.
The TV camera shows that you produce big bubbles.
23 Now, those are very different.
Wha t do you do 24 with that?
25 7-You get two experts in the roota, one of whom has V
ACE-FEDERAL REPORTERS, INC.
202 347 3700 Nationwide Coverage 800-336-6646
5040 05 12 61 1
DAVbur 1
tiever seen the KFK test and one who built the code.
Do you
(
2 vote?
What do you do?
3 HR. MURPHY:
In a case like that, rather than 4
trying to say this is my degree of belief; namely, I believe 5
one and this would be the other, given conflicting 6
information, neither of which I can discount, I would weigh 7
each one of them as 50/50, and I would go into my Latin 8
hypercube saying either one of them could occur; which is 9
the probability?
10 DR. CATTON:
There is another example.
That is 11 this in-vessel recirculation.
You fail some of the small 12 pipes on the hotleg.
You fail the surge line, or you fail 13 the steam generator tubes.
O
\\-
14 What do you do now?
Divide it into four and 15 various combinations?
16 DR. SAUNDERS:
Sure.
That is what he said.
o 17 DR. CATTON:
Is that what you would do?
18 DR. SAUNDERS:
I don' t know.
19 MR. MURPHY:
For each of these issues -- and we 20 are going to select about 15 of them.
I don't have the list 21 with me now.
Mark does -- we are taking these and we are 22 looking at the possible range of outcomes that comes out of 23 this expert panel that we put together to look at them.
24 The expert panel has a variety of different 25 degrees of belief.
Some of them, if you would, are very
(
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-336-6646
5840 05 13 62 1
DAVbur 1
conservative in their approach.
Some of them are very
()
2 optimistic in their approach.
3 DR. CATTON:
Are any of them deterministic or 4
physically based?
5 MR. MURPHY:
To the extent possible, they are 6
based on what calculations have been done.
But in a number 7
of these areas, we are at a point where we may have models 8
without good experimentation to validate them.
We are in 9
areas where there are different models, and the IDCOR 10 results differ widely from some of the calculations we 11 have.
12 13 O
14 l 15 16 17 18 19 20 21 l
22 23 l
24 l
25
()
l i
ACE-FEDERAL REPORTERS, INC.
l 202-347-3700 Nationwide Coverage 800-336-6646
'5840 06 01 63 1
DAVbw 1
DR. SAUNDERS:
The point is, I can understand how
(}
2 you can ask an engineer for an envelope of possible values.
3 Most of them will agree on limits.
Whenever you start 4
sampling within that range, the idea that you would take a 5
uniform distribution or any other distribution is what's 6
arbitrary and which governs, as you will admit, the 1
7, distribution of outcomes.
8' MR. MURPHY:
I fully agree it's arbitrary.
Not 9
everything is flat.
There are some cases where we could say 10 as we look at the technical community, a large body of the 11 technical community believes that the thing is somewhere int 12 the middle, but there are some people on both extremes that 13, say this could be a bad problem, or this is no problem at 14 all, and they have done some calculations to quantify this.
15 That will give a lower rate to their opinion, but we won't 16 reject it.
17 DR. KERR:
Then you're going to weight the 18, opinions of experts.
Expert 1 will get more weight maybe 19 than Expert 5.
20 MR. MURPHY:
Effectively, what we did is, Ralph 21 started it.
For the first plant, we had a meeting in 22 Albuquerque a couple of weeks ago, where we had essentially 23 the teams comprised of our review groups of the source term 24 and containment.
Each member -- the Sandia personnel 25 involved in the study gave their idea as to what the 73 G
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 80433MA46
l 5840 06 02 64 1
DAVbw 1
potential outcomes would be.
Their base case, if you would,
(^))
2 would be based on the source term code package run, but 3
then, recognizing some phenomenon that may not be in the 4
source term code package that somebody thinks is impo r tan t,
5 they look at various options as to else a phenomenon could 6
behave rather than the way our present model calculates it.
7 And the experts are being asked to give their opinion as to 8
what weighting factors would be appropriate to use in the 9
sensitivity analysis to represent the possibility that these 10 other things could exist.
11 These weights are roughly akin to a degree of 12 belief, but I suspect the large majority of the people who 13 are involved in establishing them would object to calling it Ol 14 a degree of belief, because their spread is wider than their 15 own individual degree of belief on it, but rather what they 16 think is appropriate to use as a spread.
17 Given that this happened, we asked each of the 18 chairmen of these groups to take the results of these 19 individual votes, sit down with their members and give us a 20 consensus opinion of what weights should be used in doing 21 the analysis.
22 So there's a distribution.
In some cases, the 23 distribution is flat.
In other cases, I might point out 24 that the nature of the process here is a discrete 25 probability value.
We're not inputting curves, we're l
ACE. FEDERAL REPORTERS, INC.
202 347-3700 Nationwide Coverage 80 4 336-6646
1 5840 06 03 65 1
DAVbw 1
inputting specific discrete values and assigning weights to
()
2 those values.
3 DR. SAUNDERS:
But they're all unimodal.
You 4
don't have engineering opinion on one hand, shifted to this 5
hand of the bound, another one which shifts the other end.
6 MR. MURPHY:
Only to the extent that since we're 7
using discrete values, we're saying, if you calculated using 8
method X, you get a point here, and if you calculated it 9
using method Y, you get a point here.
We don't allow for 10 anything in between.
11 So to that extent, you can call it dimodal, but 12 it's because we're using discrete values and assigning 13 weight to them.
If the team of experts felt that there was O
14 potential for ranges in the middle, then discrete values 15 would be used.
The technique we're using to handle 10 to 15 16 issues in 3 to 5 specific values for each issue with weights 17 assigned to it and then progressing through,this with the 18 constrained Monte Carlo technique of the limited Latin 19 hpercube, there will be 20 to 30 runs made samples through 20 the three.
The nature of the sampling process is that it's 21 constrained to sample from various parts of the 22 distribution.
So it's more effective in terms of the number 23 of runs that you made than the standard Monte Carlo.
24 The disadvantage of that is obviously, if I'm 25 making only 20 runs through, if I assign something a weight ACE-FEDERAL REPORTERS, INC.
202 347-3700 Nationwide Coverage 800-336-6646
5840 06 04 66 t
1 DAVbw 1
of 1 percent, that's the constraint put into the limited
()
2 Latin hypercube.
3 There are some disadvantages associated with it, 4
compared to an ordinary Monte Carlo, but we feel they're 5
well outweighed by the advantages and the fact we can make 6
many more runs and cover much more ground.
7 DR. SAUNDERS:
But of course, information always 8
behave like the square root of the number of trials or 9
sample size.
So the more you run, the more information 10 there is.
Presumably, the best thing would be to run all-11 possible values, but that's too expensive.
So what you do, 12 I'm trying to understand.
You take the range of values and 13 sample from those independently -- not independently.
You 14 select what you call a Latin square in several dimensions.
15 Then sample from that a fixed number of times.
16 MR. MURPHY:
Let me show you a fairly simple 17 example.
If I had seven issues with three potential values 18 for each issues --
19 (Slide.)
20
-- I'd wind up with different weights assigned to 21 the different levels on each issue.
On X 1, for instance, 22 40 percent of the samples, I've weighted at level 2, 40 23 percent.
So if I run 20 samples, I'll get 40 percent of 20 24 or 8 samples that involve level 1 of X 1 mixed up with some 25 of these other samples here, but I'm constrained to take O
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-336-6646
5840 06 05 67 1
DAVbw 1
40 percent of my samples from that value.
()
2 So if I get to a point where I have a 10 percent 3
weight, at least two of my runs would involve that 10 4
percent weight, but it could be involved with any of the 5
other parameters.
It's a way of assuring a sample space, in 6
terms of each of the individual parameters.
So it gives me 7
a more uniform sampling technique, but obviously, I can miss 8
things by the contrained nature of it.
9 MR. ERNST:
There might be a couple of points 10 that need to be made, and someone correct me if I'm 11 statistically incorrect in what I'm saying.
When you 12 finished the limited Latin hypercube, let's just, for 13 argument's sake or example's sake, say that there's 20 runs, b-14 and then we're going to, in most cases, go to a 595 on the 15 LLH runs and say that represents a reasonable range of mean 16 values.
That isn't the 595 percent of the number, but 17 that's a reasonable range of mean values.
18 DR. SAUNDERS:
The expected output has this 19 number, and then expected output will fluctuate with it.
20 MR. ERNST:
It isn't a full range of 21 uncertainty.
22 DR. SAUNDERS:
I understand that.
23 MR. ERNST:
It's very important to understand 24 that, and then Joe mentioned this 1 percent item that would 25 more than likely not be sampled, if that is an important b3 ACE-FEDERAL REPORTERS, INC.
202 347 3700 Nationwide Coverage 80 4 336-6646 L.
5840 06 06 68 2
DAVbw 1
item to risk, it will still be identified in the writeup, in p(,)
2 my view.
And-say it did not get picked up in the reasonable 3
range of means, but there is a possibility of -- and then 4
whatever the risk importance importance and the origin of 5
that uncertainty might be.
6 DR. SAUNDERS:
But the point is that when you say 7
mean, what you're doing is, you're averaging across the set 8
of opinions.
9 MR. ERNST:
Sure.
10 DR. SAUNDERS:
And it's just an expected value.
11 We get a measure of what the expected value will fluctuate.
12 MR. ERNST:
There'll be a range of expected 13 values.
(
14 DR. SAUNDERS:
Yes, I understand that, but again, 15 you say it does not give us any idea of the total range of 16 outcomes.
17 MR. ERNST:
We cannot advertise that as a 5.95 18 percent on the range; that's correct.
19 MR. MURPHY:
What we're going to do, the way 20 we're treating it is, we're going to take the 5.95 percent 21 limited Latin hypercube, which means that in the case that I 22 had 20, I'd eliminate the high and the low.
23 DR. SAUNDERS:
I understand that.
24 MR. MURPHY:
I just at that point treat that band
{}
I have as a reasonable engineering estimate as to where I 25 ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-336-6646
5840 06 07 69 1.
DAVbw 1
think the value may lie, with no statistical support, and so
[>l 2
on.
s, 3
DR. SAUNDERS:
These are not confidence 4
intervals, in the sense that the statistician uses the 5
word.
All he's doing is chopping the range.
He's chopping 6
the upper and lower ends of his distribution of expected 7
values.
8 MR. ERNST:
Absolutely correct.
We don't want to 9
advertise this at something that it isn't.
10 DR. CATTON:
But it's really averaging across a 11 set of opinions.
12 DR. SAUNDERS:
Of opinions.
13 DR. CATTON:
Times expert forcefulness.
14 MR. MURPHY:
It's taking the expert opinions.
15 DR..CATTON:
In a group of experts, one can 16 dominate just by sheer personality, whether right or wrong.
17 MR. MURPHY:
I hope we have enough people 18 involved that that's not happening.
19 DR. CATTON:
It does happen.
20 MR. MURPHY:
We're taking these values and 21 putting weight to them.
One thing I forgot to mention is 22 that the code does have a correlation between issues.
So 23 the fact that you've driven one item high, may, by its very 24 nature, drop something else low.
25 DR. CATTON:
I mean, one of your experts may be i
l l
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 80 4 336-6646
5840 06 08 70 1
DAVbw 1
a very strong-willed individual and everybody capitulates to
()
2 the wrong --
3 MR. MURPHY:
I think that's a problem anytime you 4
get a subjective judgment.
5 DR. CATTON:
As long as your chairman is 6
demanding a consensus, it's a problem.
If your chairman is 7
willing to use his own judgment, based on what he hears, you 8
can eliminate some of that.
9 DR. SIESS:
He has to be smart too.
10 MR. MURPHY:
There's another part to this process 11 that I ought to describe to you.
What I've described now is 12 what's going on in the laboratories.
What's going on is 13 part of the SARRP.
In terms of the issue papers, we're Q
14 developing, that have been alluded to a number of times, and 15 that Mark will be talking about, if I ever get of f, there we 16 are taking our own look at the issues and we're developing 17 an NRC position on these issues as to what NRC feels the l
18 proper range and the proper best value for each of these 19 issues is.
20 DR. CATTON:
Are these the white papers we've 21 been hearing about?
22 MR. ERNST:
Position papers.
23 MR. MURPHY:
These may or may not agree with what 24 has come out of our activities with the lab that I've just 25 described.
If they do, fine and dandy.
If they do not, ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-336-6646
5840 06 09 71 2
DAVbw 1
what's in 1150 will be the opinion that the Staff developed n()
2 out of the issue papers and we'll go back and rerun the 3
analysis through the Latin hypercube, using the Staff 4
opinion rather than the laboratory opinion.
5 The NUREG CRs that come out addressing the severe 6
risk reduction program will have the laboratory's version of 7
what they believe should be in there, but NUREG 1150 will 8
have Staff opinion on it, which may or may not agree with 9
the laboratory.
10 MR. ERNST:
Let me describe the process, which 11 might help a little bit.
The process that happened in the 12 meeting at Sandia about three weeks ago.
13 In any one set of issues, there may be five or O,
14 six people on this panel, but there's also an audience of 15 about 20 to 25 people, which would include other 16 contractors, NRC people and everything else.
And they're 17 free to chime in with their own opinions as information to 18 that panel.
Then as Joe said, we're still free to make up 19 our own minds.
If we want an additional range to be 20!
described in 1150, we will ask the contractor to do so.
21l MR. MURPHY:
I should mention that we do have one 22 check on the process, if you will, at least the process of 23 the limitations that are imposed by the limited Latin.
24 hypercube.
We're going to be looking at two sequences, 25 TC and TB on Peach Bottom.
There we'll forget the word O
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 804336-6646
5840 06 10 72 1
DAVbw 1
" limited," in terms of the limited Latin hypercube.
The
()
2 Latin hypercube analysis will be done, but at this point, 3
there will be no constraint on the number of issues to be 4
studied or the possible values for them, or whatever.
5 A thorough analysis will be made of those two 6
sequences, using this technique without the constraints 7
which go with the word " limited," and we'll get some idea as 8
to what we have gained or lost by the limited portion of the 9
analysis.
10 DR. SAUNDERS:
You have 15 variables, 5 possible 11 states.
That's 5 of the 15 different calculations you're 12 going to make?
That's what you told me a few minutes ago.
13 15 variables, 3 to 5 different states.
14 MR. MURPHY:
What I'm saying is, that we've
-15 constrained the number of variables to 15, as we start off, 16 by saying that these are the key elements.
There may be 17 50.
In this type analysis, we'll do 50.
18 DR. SAUNDERS:
So you'll have 5 of the 50 as 19 calculation.
That's getting up there, you know.
20 MR. MURPHY:
Within obvious constraints.
At some 21 point, you have to cry " tilt."
All I can say is, from what 22 the program's costing, they're doing a lot of calculations.
23 I don't remember the number of issues you're considering.
24 MR. CUNNINGHAM:
I don' t think it's been settled 25 yet, but it's much larger than the 10 to 15 we've been ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 80 4 336 4 646
1 5840 06 11 73 1
DAVbw 1
talking about.
(3 2
DR. KERR:
You're convinced this will give you j
3 better results, that if you just ask the experts to vote on 4
what they finally believe the risk to be?
5 MR. MURPHY:
It would be far more traceable, with 6
far better defense.
7 DR. KERR:
It would certainly be less 8
complicated.
This other looks more sophisticated.
I'm just 9
wondering if the results would be different, but I suppose 10 they would be.
11 DR. SIESS:
When you say "more traceable," you 12 mean more credible?
13 MR. ERNST:
No, I said it would be more 14 traceable, with a better explanation of what happened.
Le t 15 each individual make up his mind whether it's credible or 16 not, I guess.
17 MR. MURPHY:
A lot of us have different opinions 18 on this, but my own opinion is that if we had a team of 19 experts, and we were sure each of them went through the 20j logical thought process that's required by using asking the 21 15 questions, if we were sure they were going to go through 22 that logical process, their answer probably would be roughly 23 equivalent to what we're getting out of it, although we 24 couldn't trace what was driving the importance of it.
r3 25 DR. KERR:
This is what bothers me about the U
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 80 4 336-6646
I I
5840 06 12 74 1
DAVbw 1
process, Joe, because of I think an uneasy feeling about
_ ()
2 asking people for their opinion.
We're taking this opinion 3
and we're massaging it with a process that looks very 4
sophisticated, and then we're saying that massaging has made 5
the process more logical.
6 No,w maybe it has, but I'm reminded of the 7
apocryphal story about the early days of computers, when 8
CBS, let's say, had set up a computer that was supposed to 9
analyze election returns.
At the,last minute, they 10 discovered the program wouldn't work, so what they did was 11 take people and analyze them, and then they ran a short 12,
program that permitted the computer to print what you put 13 li into it.
They printed out the computer returns, and they O
\\-
14 looked much better, because they'd been printed out by a 15 computer.
16 I know this is not the same process, but for some 17 reason --
18 MR. ERNST:
I'm very sympathetic for what you 19 just said.
I think Joe and Mark will both agree I've been 20 campaigning long and hard to try and make sure that the l
l 21 statistical sampling process doesn't overpower the 22 engineering judgment.
The words in 1150 have to describe 23 what we're doing.
What we're doing is not basically a 24 statistical process.
It's basically a way of coming up with fs 25 what we believe are reasonable engineering bounds on the v
26 estimated mean values.
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800 336-6646
5840 07 01 75 1
DAV/bc 1
DR. KERR:
You've given it a lot of thought.
I'm
(~)T 2
willing to wait and see what happens.
3 DR. SAUNDERS:
But I don't understand why you 4
don't take the 5 to the 50 and do a sampling to whatever 5
degree you can afford, instead of going through this Latin 6
hypercube, or fractional, incomplete balanced block designs, 7
or whatever you want to call it.
8 I can make up some new designs, you see, and give 9
different answers perhaps than yours.
Why don't you just 10 sample from the big scheme?
11 MR. MURPHY:
The thought that went into it was it 12 was a way of covering the sample space with a relatively few 13 number of runs.
14 DR. SAUNDERS:
So you get sufficient information 15 at less cost?
Is that what you're telling me?
16 MR. MURPHY:
We thought it was a technique that 17 would give us the kind of information we needed to identify 18l this general type range in a manner that would maximize our 19 use of resources.
20 MR. ERNST:
We're actually doing it both ways.
{
21 We pick the issues we think are the dominant risk l
22 contributors, in which case, the others should be very small 23 contributors.
24 We're not that interested in the exact number of 25 the range because we're just advertising it as a reasonable O
ACE-FEDERAL REPORTERS, INC.
202 347-3700 Nationwide Coverage 800-336-6646
5840 07 02 76 1
DAV/bc 1
range.
So if it's a little bit off, the check is to see
{}
2 whether we're way off.
3 If we're way off, then we have to go back and 4
rethink the whole process again.
5 DR. SAUNDERS:
All right, I understand what 6
you're doing.
Thank's.
7 MR. MURPHY:
Could I have the next one, Mal?
8 (Slide.)
9 I think I've said this all already.
Let's go one 10 more and see if you have questions.
11 (Slide.)
12 My summary is a little out of place because it 13 assumes you've already heard the front end analysis part of f'l v
14 what I'm about to proceed to, but I'll give you the' summary 15 anyhow.
16 We're doing the rebaselining and the actual 17 frequencies using compressed PRA methods, which we believe 18 are highly cost-effective.
Really an intriguing tool that 19 has worked out very well.
20 We have the evaluation of the accident l
21 methodology again with the best tools that we have available 22 to us at the time.
That's basically the source term code 23 package.
24 We're doing the offsite consequence analysis 25 using an improved code max, which I'm going to talk about f.- S i
\\_)
l l
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-33MM6 L
5840 07 03 77 1
DAV/bc 1
later.
'i (G
2 And I've described the uncertainty analysis to 3
you, which we feel will give us reasonable engineering 4
perspectives on the possible range of outcomes with the 5
purpose of using it to focus attention on the important 6
engineering and phenomenological assumptions, on which we 7
have a lack of knowledge.
8 And as should be completely obvious from our last 9
20 minutes or so of discussion, our uncertainty estimates 10 are not rigorous, 11 Now I'll go back to the beginning of my talk.
12 (slide.)
13 Again, many of the comments here were answered by 14 questions that Mal answered as he was giving his general 15 overview.
In the interest of saving time, I will try to 16 gloss over areas that we covered before.
17 We are doing the risk profiles and perspectives 18 on the six reference plants in terms of the risk reduction 19 potentials that are being looked at.
20 On these technical items Dr. Catton asked about 21 earlier, the fixes were identified in 0900 earlier.
Several 22 years ago, the vented photo containment.
Add-on decay heat 23 removal.
Add-on containment heat removal features.
24 We also asked the analyst to look for plant 25 specific modifications that would be suggested by the ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800 336-6646 L
5840 07 04 78 1
DAV/bc 1
accident sequence rebaselining itself, and looking at the
()
2 analysis of the plant as it exists today, is there anything U
3 that strikes the analyst as sensible to do to try to reduce 4
risks?
5 I may be biased but I think the risk analyst, as 6
he has done his analysis, has a fairly good idea of what's 7
driving the risk to his plant and what type of modifications 8
might be cost-beneficial.
9 This is more than just getting importance 10 measures, say, coming out of fault tree evaluation, or 11 whatever.
Rather, he understands the system pretty well 12 from having done it, so he has an intuitive feel for where 13 he can gain advantages.
m 14 So he looks at it both ways in term's of what's 15 driving the risk and he gets that out of a standard 16 importance measure type of thing.
17 He also knows the plant pretty well after he's 18 been done with it.
He's been living with this thing for a 19 few months'.
He may well be able to suggest a way of 20 reducing risk that would be cost-beneficial.
I mean, a 21 specific example rather than review cut set X.
He has a 22 design modification in mind that can accomplish the 23 reduction of those cut sets.
But perhaps cut sets in.
24 several sequences at the same time, just because of his 25 integrated knowledge of the plant.
O ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-33MM6
~
l 5840 07 05 79 1
DAV/bc 1
We've asked him to look for that and use their
(}
2 engineering judgment in that type of process.
3 DR. KERR:
It seems to me at this point it would 4
be more likely to work if the people who were doing the 5
analysis also were good plant designers.
6 It is rare to find that combination in the same 7
person.
8 MR. MURPHY:
I agree.
9 DR. KERR:
Do you have people who are that kind 10 of people?
11 MR. MURPHY:
We have some pretty good people but 12 I don't think too many of them have design experience, in 13 the sense you're talking about.
Most of them in this part O
14 of the analysis tend to be PRA types.
15 DR. KERR:
That's sort of what I thought.
16 MR. MURPHY:
There are a couple of exceptions to 17 that but, in the main, that's the case.
18 Again, we talked about the display of the 19 important uncertainties and the insights in the use of risk 20 important information.
21 (Slide.)
22 This I think everybody knows.
I put it up and 23 described the sequence in which I was going to talk about 24 things, instead of talking about them in almost exactly the 25 opposite sequence.
We've talked about the two in the ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 80 4336-6646
5840 07 06 80 1
DAV/bc 1
middle.
I'm now back on the first one and I'll come back to
()
2 the last one.
3 MR. ERNST:
That sort of comports with how we've 4
been doing this project anyway.
5 (Slide.)
6 MR. MURPHY:
In terms of the activities that we 7
have in process, I think we covered everything on this.
8 Let's proceed with the next one.
9 (Slide.)
10 Getting to teh approach we're using for the first 11 four of the rebaselining activities -- Surry, Peach Bottom, 12 Sequoia and Grand Gulf.
And I will address later the 13 process that we use in these other plants.
O k
14 We start of f in the preparation phase where we 15 essentially learned everything we could about the plants, 16 based on records, where the information was already on hand 17 for the analyst.
If there was a previous PRA on hand, they 18 reviewed the PRA.
If the accident sequence evaluation 19 program had done generic modeling of the plant systems to 20 understand what was done there.
To study the FSAR and what-21 ever other information was on file with the NRC.
22 Ha"itt # knowledge of the plant from the 23 documents, the teas.nen went to the plant for roughly _a week 24 for a process, as I say, to obtain as built drawings to gs 25 assume the as built condition of the plant, comporting with d
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-336-6646
5840 07 07 81 1
DAv/bc 1
their understanding of the plant.
To get the procedures,
(')
2 the current procedures, and then to progress from those on 3
to interviews with the operators to understand how the 4
operator implemented the procedures.
5 At this point, they went back home and did more 6
or less standard event tree and full tree modeling.
I'll 7
talk a little bit more about the scopes of those in the next 8
slide, I think.
9 From there, we went to base case quantification l'0 and then, through uncertainty and-sensitivity analyses.
11 DR. KERR:
Why is Zion not on the list?
12 MR. MURPHY:
Zion we' re handling somewhat 13 differently.
Zion, the analysis, and the same thing is true O
14 of Lasalle.
In Zion,'the analysis is going to be based 15 mainly on an updating of the Zion PRA rather than starting 16 again drawing trees in this manner.
17 Basically, what they're going to do, the basic 18 document they'll be working on is on the front end, is the 19 Sandia review of the Zion PRA.
And then update that 20 information and go on forward with that.
21 Similarly, on Lasalle, Lasalle will be based on 22 the models that exist already in the Omni program, which are 23 to a significant greater level of depth in some areas.than 24 what we're doing in this rebaselining activity.
r-25 We will use those where they exist and we will
(_3/
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-336-6646
5840 07 08 82 1
DAV/bc 1
use the simplified method where they don't exist.
()
2 They are a somewhat different process.
At the 3
end of it, before they document it, they go back to the 4
plant and go over the conclusions with them to make sure 5
that the plant performs in the way they have assessed, and 6
revise the document.
7 We have detailed technical OA at each phase.
To 8
give you an idea of the scope of the OA, we have a OA team 9
at.Sandia that I believe is four people long and they're 10 spending 50 percent of their time OA-ing these four 11 analyses.
So the people doing the analysis have an expert 12 looking at what they're doing, not only from the standpoint 13 of finding errors, but making sure that they're using the 14 system assumptions and boundary conditions plant to plant so
,15 we come up with something and we're not comparing apples 16 and oranges.
17 MR. DAVIS:
Joe, all four of these plants have 18 previous PRA's?
19 MR. MURPHY:
Yes.
20 MR. DAVIS:
You're using that as the baseline to 21 start into this?
22 MR. MURPHY:
That's one of the places we're 23 starting, but you get something like Surry and WASH-1400 24 that was published 10 years ago -- actually, the design was 25 frozen about 12 years ago -- that plant doesn't look really O
ACE FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-336-6646
5840 07 09 83 1
DAV/bc 1
too much like it did back in the days of WASH-1400 in some
(}
2 key features.
3 MR. DAVIS:
And it may not 10 years from now.
4 MR. MURPHY:
Well, we're going to be closer 5
timewise, at least.
You know, I'll agree, it's a moving 6
target.
7 (Slide.)
8 To give you an idea without going through all the 9
gory details of what we're doing in each of these methods, 10 let me just show what we're doing to give you a comparison 11 of what we're doing versus WASH-1400.
The earlier RSSMEP 12 stuff, the IREP stuff we did a couple of years ago.
And now 13 our RMIEP program, which is now in progress.
14 This gives you a good idea 'of what the scope of 15 te analysis is.
16 MR. DAVIS:
I'm surprised your human factors 17 detail is not any better than WASH-1400.
18 MR. MURPHY:
There should be an asterisk on that 19 one, Pete.
The basic human error probabilities associated 20 with leaving components in the wrong position are basically 21 based on the same type of analysis done in WASH-1400, 22 although the procedures for doing it are significantly i
23 better now.
They're documented and they've been tried; 24 where, in WASH-1400, they were not documented in areas where 25 human factors appeared to be very important.
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-336-6646
5840 07 10 84 l
1 DAV/bc 1
The main one is TC on Peach Bottom, the Mach I.
()
2 There we're doing a detailed state of the art human factors 3
analysis on that, as good as anything I'm aware of having 4
been done in any PRA.
5 MR. DAVIS:
I got that impression.
That's why I 6
was surprised to see that.
7 MR. MURPHY:
On that one case where it appears to 8
be really the dominant thing in the accident sequence, we're 9
spending a lot of money and we're getting what appears to be 10 a very good analysis of the impact.
11 We have a limited search of the accuracy 12 available to us, the data.
We're doing limited data 13 analysis on key components the analysts believe are b
14 important, plant-specific data.
In some cases, there's a 15 l lot.
In some cases, there's not too much.
It varies plant 16 to plant.
17 I guess, rather than go through each one of these 18 l In detail, the next viewgraphs, I believe, highlight the 19 ones where what we're doing is somewhat less than IREP.
I 20 can address those directly.
If you want to discuss any of 21 the other ones, we'll see what we can do.
22 MR. WARD:
What plants are being done under the 23 RMIEP program?
24 MR. MURPHY:
RMIEP is one plant, Lasalle.
(Slide.)
25 l
l l
l l
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-336-6646
1 l
5840 07 11 85
)
i 1
DAV/bc 1
Let me just talk about actuation and control.
In
()
2 terms of the pressurized water reactors, the actuation and ss 3
control circuitry is treated as a black box.
What I mean by 4
that is they are taking the actuation circuitry for a given 5
system and they're developing a fault tree model for the 6
actuation part of this once.
And then they're assuming they 7
can treat that as a black box with that reliability value 8
forever, for every other system of the plant.
9 On the boilers, where the nature of the control 10 system is such that many relays and coils feed different 11 systems, it wasn't possible to do that,,so there is detailed 12 modeling of the actuation circuitry on the BWR's but not on 13 the PWR's.
(~)
14 In general, we're modeling the TC power interface 15 directly, explicitly.
For AC mode of control power, we 16 generally that's transformed from the motive power source so 17 we don' t model the control power except we assume that it's 18 taken off the motive power.
19 The local control circuits are generally 20 considered to be part of the component.
I'm talking about 21 the local control circuit at the component itself.
The lead 22 switch on a Limitorque valve is considered part of the valve 23 and included in the valve failure rate rather than modeled 24 explicitly.
25 Of course, in developing the valve failure rate O
l l
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-336-6646
5840 07 12 86 1
DAV/bc 1
model, you have to have that in mind, too.
Apples and
(~}
2 oranges don't mix.
w-3 (Slide.)
4 In terms of the electrical systems, we generally 5
modeled them at teh 4160 level provided that there's firm 6
indication that we have independence at the lower buses.
7 Again, in general, they're looking at bus faults, 8
transformer faults as the main source of faults.
9 We haven't done...if something is sitting on a l
10 480 bus, we haven't done the detailed fault tree modeling 11 that takes you from the 4160 down to the 480 for every 12 circuit breaker that exists on that train.
13 Experience has shown that the kind of things that b
14 get you in trouble in a PRA tend.to be faults at the higher 15 level, in the 4160 range, or the 125 volt DC cange.
Those 16 are the ones that are explicitly modeled.
17 In terms of subtle interactions, the kinds of 18 things you find from a very detailed investigation of 19 control circuitry or the very fine details of the plant.
20 DR. SAUNDERS:
I don't know what an interaction 21 is.
22 MR. MURPHY:
A dependent failure.
I can give you 23 an example of the kind of things we found in the past.
But 24 what we had the team do as their first step is, basically, 25 the OA team, as a matter of fact, was to go through all the Ci ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800 336-6646
i
.O 5840 07 13 87 1
DAV/bc 1
PRA's they were familiar with and we have a team that 2
essentially encompasses almost the entire PRA community, who 3
have done these PRA's.
And to find these subtle i
4 interactions that were very easy to miss.
5 DR. SAUNDERS:
What's a subtle dependent failure?
6 MR. MURPHY:
Let me give you an example.
7 DR. SAUNDERS:
All right.
8 9
A 10 11 12 13 14 15 16, 1
17 18 19 20 21 22 23 24 25 O
ACE-FEDERAL REPORTERS, INC.
3)2-347-3700 Nationwide Coverage 800 33H646
5840 08 01 88 1
DAVbur 1
MR. MURPHY:
When we did our PRA on Millstone, we
()
2 found it was possible to get into a situation looking at 3
load shedding relays, that there was redundancy to assure 4
that the load was shed so that the relays would open -- two 5
relays would open, either one of which would shed the load 6
from all diesels.
7 If any one of these things stuck open, you were 8
getting a signal that said when you got your diesels back 9
on, you started the diesels up, there would be a signal to 10 all the engineering safety features, both buses from each 11 relay that said don't load.
12 So a single failure of a relay contact would 13 -
disable electric power to all trains of safety equipment.
( )_
14 This was found -- it was found actually by fault 15 tree analysis that was being done by Northeast Utility 16 personnel at the time.
It was the first LER that I ever saw 17 that started off at 4:00 p.m. while doing fault tree 18 analysis that we found.
19 DR. SAUNDERS:
Why is that subtle, though?
l l
20 l MR. MURPHY:
Because it requires you to go in 21 your modeling all the way down to the load shedding relays, 22 which takes you through train after train of relays and 23 coils as you go down through.
24 DR. KERR:
" Subtle" means nobody has seen it 25 before?
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 80433Mi646 L
5840 08 02 89 1
DAVbur 1
MR. MURPHY:
Yes, that is what it means.
( }
2 DR. SAUNDERS:
That is fine.
3 MR. MURPHY:
It is that you have to get to such a 4
great level of detail to get it that it goes into your 5
modeling activity.
Since we knew we weren't going into that 6
level of activity, we asked people what they had observed in 7
the past, and we had a 20-or-30-page list of essentially all 8
the strange things that PRAs have found -- I won't say all, 9
but most of the strange things that PRAs have found.
10 We have taken those and then given explicit 11 instructions to the teams to go deeper into this area and 12l look for these type of interactions.
But it is more an 13 inductive process than a deductive. process.
(s~T I
/
14 Ordinarily, you go through a fault tree to do it, 15 and there is a deductive logic that allows you to find it.
16 This time it is an inductive logic based on having seen it 17 in the past.
18 In general, we modeled instrument air as single 19 failure rather than going into the details of the f ault 20 trees and the air compressors and all the various ways you 21 can lose air.
Taking generic data where po.ssible, we have 22 looked at the history of instrument air losses and 23 essentially treated the whole system as a black box.
24 MR. BENDER:
Joe, there is a big environmental 25 qualification program.
What influence does it have on the bs ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-336-6646
5840 08 03 90 1
DAVbur 1
PRA program?
()
2 MR. MURPHY:
At the present time I think the 3
right answer is not enough.
4 MR. BENDER:
In what way?
5 MR. MURPHY:
If I was talking in the Army program 6
rather than what we are doing now, I could talk in much more 7
detail about what we had last week.
8 But in fact we are not looking at what we call 9
external events in this program.
We are not looking in this 10 rebaselining effort at things like fires and floods.
11 Flooding, in particular, is an interesting one 12 because you may be spraying hot water rather than just 13 having the water come out.
In that kind of information, Okl 14 information coming out of the EQ program, we give you some 15 idea of how the failure rate varied as you increased 16 temperatures and pressures.
That would be very helpful.
17 Unfortunately, most of the equipment 18 qualification programs are not designed with PRA in mind.
19 They are designed with other objectives.
20 The fact that I have taken a pump and put it in 21 an autoclave and run it for 30 days under a given set of 22 conditions and I tested it when it came out and it worked is 23 enough to meet the equipment qualification requirements that 24 are under the regulations, but it doesn't tell me anything 25 on how the failure rate varies now that I have exposed it l
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800 336-6646
5840 08 04 91 1
DAVbur 1
to this harsh environment.
f~T 2
So the ability to take that information and feed V
3 it back in the PRA requires that I have a lot of data.
4 Right now I don' t think I have enough to do it.
It is 5
something we are looking into very seriously in some of our 6
research programs, but it is very difficult because of that 7
problem.
8 MR. BENDER:
It is a two-way street.
If you had l
9 the information, then, it might change the failure cate that 10 you are using as data for the analysis.
If you are not 11 taking any credit for it, though, in one way or the other it 12 is hard to understand why there is any significance attached 13 to doing it.
14 MR. MURPHY:
Wha t we are doing in our later PRAs 15 is to try to get a handle on the potential importance, to 16 identify where environmental stresses can affect the 17 component.
18 This is done by putting, say, location 19 identifiers on the component as we are in the trees and then l
20l identifying what locations will be affected by a given 21 stress, be it high pressure, temperature, high humidity, or 22 whatever.
23 Then we can see if there is one area where we 24 have a number of components all subjected to a high 25 temperature / pressure environment, say, and to assess whether ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800 336-6646
5840 08 05 92 1
DAVbur 1
these are of critical impor tance.
()
2 If they appear in the dominant cut sets that 3
require accident sequences, it suggests that this is an area 4
that needs further work.
If it turns out upping the failure 5
rate by a factor of 10 doesn't make any difference, then 6
maybe it is not.
7 MR. BENDER:
I think that is a good way of doing 8
it, but you didn't say you were doing that.
You were just 9
saying you could do it.
10 MR. MURPHY:
We are at the process now on our La 11 Salle PRA where we are putting these identifiers on -- the 12 location id,entifiers on the trees.
We are putting the 13 environmental hazards on the trees, but that is something
~
14 coming out of that we can look at.
I can't look at it until 15 it is done.
16 (Slide.)
17 Now, we get to the consequence model, as I shift 18 around up here.
19 As part of the MELCOR Code development system, we 20 have developed a new consequence model called MACCS.
It has 21 essentially seven new features that are different from CRAC 22 2, which is the tool we have been using in the past.
23 We have the ability now to treat time dependent 24 releases.
CRAC is pretty close to the back, for those of 25 you who are looking for the viewgraph.
CRAC was a puff
{}
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800 336-6646
5840 08 06 93 1
DAVbur 1
release model.
This is a multiple puff release.
So you can
/~T 2
model the time dependent failure of the release rather than b
3 assuming it all goes in one puff.
4
-This has some substantial influences.
For 5
example, if you are looking at dose to people who are 6
evacuating and you spread the time demand of the release 7
over two or three hours, it may be instead of getting the 8
whole thing, which they do if you had expressed it at a puff 9
at time zero, since they now assume they are going to be 10 some distance before the road before the last piece comes 11 out, there is a redrction in the influence of it.
12 There is improved atmospheric dispersion, 13 including crosswind distribution, better particle size 14 distribubion within the plume, and up-to-date plume rise 15 models.
16 They are adding additional exposure pathways --
17 the food pathways, deposition on the skin, improved 18 dosimetry.
This is basically work coming out of Harvard.
19 In terms of emergency response, there is ability l
20 to consider graded response.
There is ability to consider l
21 the road network evacuation, where in the past it was a 22 straight line evacuation, at least in the CRAC 2 modeling, 23 delocation of sheltered individuals and relocation outside l
24 the EPZ.
25 In CRAC 2, once you were sheltered on site, you
,-(m)g ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-336-6646 s
5840 08 07
~
94 1
DAVbur 1
never left.
( ).
2 In terms of the long-term effects, it now 3
includes long-term ground shine and direct deposition on 4
crops.
I don't know what the OR there means.
5 And we have the improved health effects models 6
with hazard functions used for early effect, additional 7
injuries and latent effects.
8 DR. KERR:
What does improved dosimetry mean?
9 MR. MURPHY:
I don't have the guy that works on 10 that with me now, and I can't answer you, Bill.
11 DR. MARK:
Does it include the option of 12 sheltering?
13 MR. MURPHY:
Yes.
It allows you to handle rd 14 sheltering much better than in the past and allows you -- if 15 somebody is sheltered and it is a prudent thing not to be 16 sheltered, it allows them to move later.
17 The decision in the earlier CRAC 2 model was once 18 you were, sheltered, you were sheltered forever.
19 MR. DAVIS:
Joe, what do you anticipate the 20 composite effect of all these changes will be versus the 21 CRAC 2 results?
Will the risks go down or up given the same 22 source term?
23 MR. MURPHY:
In what I have seen to date, it 24 varies a little, but it looks like it is going up.
25 MR. DAVIS:
That surprises me because one of the s
ACE-FEDERAL REPORTERS, INC.
202 347-3700 Nationwide Coverage 800-336-6646
3 5840 08 08 95 1
DAVbur 1
problems I had with CRAC is that unless you tell it
(~')
2 otherwise it assumes peop1.e are relocated and their lifetime
%J 3
commitment would be 25 rem.
About 90 percent of the latent cancer fatalities were based on that avoidable dose.
4 5
Now, I would have assumed something different 6
'would be done in a revised code like that.
7 MR. MURPHY:
I believe there is something 8
different.
If I understand the improved point model, it 9
really has the effect of increasing things.
10 Al and Mark may know more about this than I do.
11 MR. ERNST:
There is a fair amount of increase in l
12 the health effects model, but we have got the wrong guy here 13, to answer the specifics.
14 MR. BENDER:
Is there a report which tells in 15 more technical detail what has really been done?
16 DR. KERR:
On MACCS you mean or what?
17 MR. BENDER:
On MACCS.
18 MR. ERNST:
I am not sure there is documentation i
19 yet.
20 MR. MURPHY:
There is some form of 21 documentation.
I don't think the final documentation is out 22 yet, but we are at the point where we are putting it out for 23 trial use for the national labs.
So there has to be some 24 sort of definition as to what it is.
25 MR. BENDER:
I just wondered if there was some l
l l
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-336 6646
5840 08 09 96 1
DAVbur 1
discussion as to what was being improved ar.d why.
~
2 MR. MURPHY:
I am sure we could give you a 3
relatively short description of what that is.
We are 4
talking three, four or five pages.
I feel reasonably sure 5
that exists somewhere already.
6 If you are talking about a couple of hundred --
7 DR. KERR:
Why don' t you send a copy to Dean?
8 MR. BENDER:
I am not looking for a couple of 9
hundred pages report; I am looking for a few page report 10 that has some substance to it.
11 MR. MURPHY:
The model has been presented to the 12 CS&I Committee and to assorted technical meetings.
I am 13 sure there are papers that describe it in detail that I am im 4
(
I 14 not familiar with.
15 MR. BENDER:
If I could just know what is being 16 done about improved atmospheric dispersion, which I think is 17 the most important technical matter to be dealt with.
That 18 by itself would be very useful.
19 DR. CATTON:
For example, the crosswinds 20 distribution -- how do you use it?
What do you gain?
21 DR. KERR:
We are going to get all that later.
22 MR. MURPHY:
An obvious problem with a lot of 23 these things, you could do a lot better modeling if you had 24 better data.
25 DR. CATTON:
It should be a balanced approach, I U
ACE-FEDERAL REPORTERS, INC.
202 347-3700 Nationwide Coverage 800-336-6646
1 5840 08 10 97 l
1 DAVbur 1
would think.
(}
2 MR. MURPHY:
I hope we have achieved that.
3 (Slide.)
4 Now, I am back to where I started, the 5
sensitivity analysis.
So I think this means that we have 6
covered all my viewgraphs in a rather disjointed fashion.
7 I do want to emphasize -- oh, I am sorry.
As we 8
are going through the front end, we are doing something 9
slightly different, and I hope that is where we are now.
10 (Slide.)
11 In terms of the accident frequency rebaselining 12 for the base case, we are doing essentially what is 13 traditional in PRA.
Data is being treated as having a 14 lognormal distribution and that uncertainty associated with 15 that distribution is being propagated through the event 16 trees and the fault trees the same way it was on the IREP.
17 Then we are doing sensitivity analysis on the key 18 assumptions that entered into that calculation.
We are 19 asking the analyst where his assumptions on success 20 criteria, in some cases the.effect of using generic common 21 cause factors as we do the analyses, various assumptions 22 they may make as to whether component cooling is needed for 23 a given component, whatever.
It varies from plant to 24 plant.
25 But typically, they have identified 14 or 15 l
l t
ACE-FEDERAL REPORTERS, INC.
202 347-3700 Nationwide Coverage 80 4 336-6646
~
5840 08 11 98 1
DAVbur 1
issues that the analysts feel are important enough to merit 2
sensitivity studies, and most of these have to do with 3
either modeling or boundary condition assumptions.
4 They are going back and doing the sensitivity 5
study on that, repropagating the data through the tree for 6
each of these issues, and then ef fectively associating a 7
weight or a degree of belief and which way they believe the 8
sensitivity should go.
9 Again, it vill be combined in one of two ways.
I 10 have one way illustrated on the next viewgraph.
There is 11 another way we are,still considering using.
12 (Slide.)
l 13 l What I have here is the way we had initially O-14 planned to do it and still do, for that matter.
This shows 15 I the results of the base case.
This is the standard upper 16 and lower bound with a mean associated with our estimate of 17 our base case.
18 Then we run various sensitivity issues, each one 19 of which now has a new mean and a new upper and lower bound, 20 and we plan to present the results in a format like this 21 which says that for these sensitivity cases this is the 22 range of the means.
23 These are the extremes that we have observed, and 24 this is the 5 and 95 percent bounds that we originally 25 associated with our base case.
So we have this box and ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-336-6646
5840 08 12 99 1
DAVbur 1
whisker type presentation, the range of the means associated f) 2 with this and the upper and lower case.
s/
3 We have a code that we are considering using, 4
called TMAC, which is essentially the limited Latin 5
hypercube.
That would allow us to take each of these and 6
put a distribution on them, put a weighting function or a 7
weighting factor on to what extent the analyst thought there 8
was a chance that the sensitivity analysis, the sensitivity 9
issue could be the way it was represented -- in other words, 10 two pumps needed instead of one -- and then it would be 11 possible to fold that through and generate something similar 12 in concept in terms of what came out at the end.
13 DR. SAUNDERS:
Box A is the result of your 14 limited sampling scheme from each of the cases or from the 15 base case?
16 MR. MURPHY:
I have done the base case, and I 17 have done all my other sensitivity issues.
I have 18' propagated through using sets in rather a thorough way.
19 DR. SAUNDERS:
I understand that, but what data 20 gives you the box?
21 MR. MURPHY:
The box comes from the range of the 22 means of these sensitivities.
It indicates a spread of the 23 means of the various sensitivity studies.
24 DR. SAUNDERS:
\\ll right.
So you have how many 25 of these means all together that go into calculating the ACE. FEDERAL REPORTERS, INC.
202 347 3700 Nationwide Coverage 800 336-6646 I
5840 08 13 100 1
DAVbur 1
box?
O 2
MR. MURPHY:
Typically, on the order of 10 to
-J 3
15.
4 5
6 7
8 9
A lo 11 12 13 14 15 l
16 17 i
18 l
19 20 21 22 23 24 25 ACE-FEDERAL REPORTERS, INC.
2 2-347 3700 Nationwide Coverage 800 336 4646
5840 09'01 101
()DAVbur 1
DR. SAUNDERS:
All right.
Then you sample from 2
these at different levels, is that right?
How is it you 3
have done it?
4 MR. MURPHY:
We have done each one of these 1
5 calculations separately.
All we have done is to say that 6
the range of the means here in the three cases I have got on 7
here -- the highest one is here, which defines this one; the 8
lowest is here, which defines the lower part of the box.
9 DR. SAUNDERS:
I see.
No calculation was made?
10 MR. MURPHY:
No, it just describes the bounds on 11 each of these sensitivity studies.
What we are considering 12 but haven't gotten to yet is going one step further and 13 attaching weights or degrees of belief associated with each' 14 one of these and folding that in.
In that case it would be 15 weighted, and you wouldn't get the direct correlation that 16 shows in this picture.
17 DR. SAUNDERS:
Just one other thing, one other 18 question:
you say it is standard to assume that everything 19 has a lognormal distribution, and therefore if these things 20 are affected as product, as most of them are, then you can 21 use the convolution and get the direct distribution of 22 consequence of several variables, right?
23 Suppose -- of course, when you do that, you have 24 the folding together and cancellation of errors that always 1
25 results -- suppose you take a fat-tailed distribution and ACE-FEDERAL REPORTERS, INC.
202-347 3700 Nationwide Coverage 800-336-6646
5840 09 02 102 1
DAvbur 1
did the same analysis.
/~
2 Do you have any idea what it would look like?
()T 3
MR. MURPHY:
One of the sensitivities we are 4
exploring right now is to change the distribution, change 5
from a lognormal assumption to something else.
A beta or 6
gamma essentially would do the same thing.
7 DR. SAUNDERS:
That is my point.
A fat-tailed 8
distribution is going to be such that the sample mean has 9
the same dispersion as each individual mean.
You don't get 10 the concentration that you get with the normal.
That is the 11 worst case probabilistically.
12 I wondered if anybody -- if you had ever thought 13 about doing that or tried that.
n' 14
'MR.
MURPHY:
I haven't thought of it.
I can't x-15 say nobody on our staff has.
16 DR. SAUNDERS:
Well, you know what the Koshi 17 distribution is?
18 MR. MURPHY:
Yes.
19 DR. SAUNDERS:
A sample mean has the same 20 distribution as each observation, the same scatter that will 21 prevent this cancellation through this, and of course it 22 will be not a more difficult analysis but it might scare me 23 if we did it.
24 But you ought to think about that.
25 MR. MURPHY:
We will see if we can input ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 8043364646
5840 09 03 103 1
DAVbur 1
something like that into our code.
To the extent we can do
()
2 it, I think it is something we definitely need to consider.
i 3
MR. BENDER:
'as t the' fact that you are doing a 4
sensitivity analysis has value.
Any way you do it will be 5
useful.
6 But the problem really is to decide what part of 7
the analysis needs to be considered.
In'some cases it won't 8
make a damn bit of difference whether you use the 9
lognormal.
In other cases it may have great significance.
10 But a selection of things to do in order to do 11 the sensitivity analysis is kind of critical.
What data 12 should be examined?
You can't examine them all because 13 there is too big a chunk of information.
14 Have you given thought to what data is going to 15 be considered?
16 MR. MURPHY:
As I said, most of what the 17 sensitivity analysis is based on now is based primarily on 18 the modeling assumptions, critical modeling assumptions, and 19 they come out from the analyst's knowledge that he got to 20 System X -- I will give you an example that I had happen to 21 me at least three separate times.
22 You take an assumption and you look at the 23 numbers that are in the FSAR, and they say there's 4000 gpm 24 should be flowing to the heat exchanger.
You assume some 25 fault that is causing a diversion of flow out of the system, O
i i
r i
ACE-FEDERAL REPORTERS, INC.
202 347-3700 Nationwide Coverage 800-33M446
5840 09 04 104 1
DAVbur 1
and you do your calculation.
You find out there is 3500 gpm 2
going through that heat exchanger, and the fault tree being a
3 a dichotomous technique, you have got to say that is either 4
success or failure.
5 Your first reaction as an analyst is it doesn't j
I 6l meet the success criteria that is defined in the PSAR, but I 7
know if it isn't enough flow it is awful bloody close to 8
it.
Therefore, temperatures will increase very, very 9
slowly.
10 We still have to finally answer the question:
do 11 I model it as one pump or two pumps?
12 l So you make an assumption, and if you are a very 13 conservative guy you say two pumps and if you are a very t'~^
14 ll realistic guy you probably say one pump makes sense because s
15 there is got to be conservatism in the FSAR assumption.
16l Whichever way you go, you are not sure, and that 17 h is the kind of thing we are treating in the uncertainty.
18 I have seen them go both ways.
19 MR. BENDER:
That is a very good example.
If you l
20 '
did enough of those kinds, I think you could improve the 21 knowledge of how much you can do.
That is to some degree an 22 issue.
23 MR. MURPHY:
You run into a lot of these where 24 you just don' t have the time and resources to go back and do 25 the whole thermal hydraulic analysis of the plant to find 7
U ACE-FEDERAL REPORTERS, INC, 202 347-3700 Nationwide Coverage 800 336-6646
5840 09 05 105 1
DAVbur 1
out what a true success criterion should be for the
(}
2 situation.
3 MR. BENDER:
If I could use an example that is 4
closer to my heart, the probability of a power failure, 5
which maybe 10 to the minus 2 per year or 10 to the minus 3, 6
I think examining the sensitivity of that kind of thing with 7
a probability distribution might tell you a lot, and it is 8
those things which have a fairly high probability of 9
happening that we need the most depth down in the low 10 probability range.
11 It probably doesn' t make much dif ference what the 12 distribution is, but if it is a high probability range it 13 does.
14 MR. MURPHY:
I tend to agree with you.
15 DR. KERR:
Further questions?
16 (No response.)
17 MR. MURPHY:
That completes it, I guess.
18 Mark will go on.
19 DR. KERR:
Before Mark gets started, I am going 20 to suggest a 10-minute break.
21 (Recess.)
22 DR. KERR:
Before Mr. Cunningham gets started, i
23 have you come to some sort of agreement on the appropriate 24 frequency for the accident initiators?
25 It seems to me I remember earlier there was some ACE FEDERAL REPORTERS, INC.
202 347 3700 Nationwide Coverage 800 336-66 4
5840 09 06 106 1
DAvbur 1
uncertainty about the initiator frequency you were going to 2
use.
3 MR. MURPHY:
What we are doing for the initiator 4
frequency, we had a study done by Idaho about two years ago, 5
I think.
They published a report on the frequency of 6
accident initiators by plant on an average.
That is the 7
basic resource.
8 Then the team also looks at the plant history, 9
where there is a lot of history.
On something like La 10 Salle, there is effectively no operating history to judge on 11 the generic values you used.
On something like Surry --
12 DR. KERR:
Have you arrived at what you connider 13 to be a satisfactory value now for those initiators?
14 MR. MURPHY:
The answer there is "yes."
15 DR. KERR:
Thank you.
16 (Slide.)
17 MR. CUNNINGHAM:
I am here to talk this af ternoon 18 a little bit about the work that is going on within the 19 staf f to develop staff positions on the risk importance of 20 particular severe accident issues.
21 DR. KERR:
Does risk importance mean risk now and 22 not core melt?
23 (Slide.)
24 MR. CUNNINGHAM:
In gener{al, most of the issues 25 we are talking abor t are phenomenological issues.
So we are ACE FEDERAL REPORTERS, INC.
202 347 3700 Nationwide Coverage 800 33 4 646
5840 09 07 107 1
DAVbur 1
talking risk.
In some cases we will also discuss their
(}
2 impact on core melt frequency.
Generally, it is risk.
3 We are doing this for two reasons.
4 The first one is that for NUREG-1150 that we 5
discussed here and for the IDCOR process that the staff is 6
going through to resolve and close out the IDCOR 7
interchange, we need to develop staff positions on specific 8
issues that could be important for the risk of a plant.
9 The issues we are dealing with are related to 10 issues that have come out of the technical interchange 11 meetings that the staff has had with IDCOR.
They have come 12 out of the specific instances perhaps of work that has 13; resulted in the ASEP and SARRP analyses to date, and they 14 are issues that have been identified in BMI-2104 or the 15 draft of NUREG-0956.
16 A second reason for doing this is that, as Joe 17 described before, SARRP is performing an uncertainty 18 analysis for itself and then eventually for NUREG-ll50, and l
19!
we wanted to have staff interaction on the ranges that were I
20 being used by SARRP and in a sense concurrence on them.
If 21 we didn't have concurrence on them, then we would develop 22 our own ranges and have the SARRP people do the calculation 23 again with what we accept as the position on these issues.
24 (Slide.)
25 There are a series of steps that are probably O
ACE FEDERAL REPORTERS, INC.
202 347 3700 Nationwide Coverage 800 336 6646
4 5840 59 08 108 1
DAVbur 1
pretty straightforward.
The important part is that the work 2
that is being done is being done jointly with the NRR staff 3
and the Research staff and our contractors.
4 For each issue that has been identified, there is 5
a staff person in Research and a staff person at NRR 6
respor-ible for seeing this process through.
I 7
So again the process itself is pretty 8
straightforward.
We have identified a set of issues and 9
identified people who will be involved from the staff to 10 develop these position papers.
11 In addition, once these teams get going, they 12 will identify their own supporting contractors, et cetera.
13 The first step they are going through is a l'
)
=
/
14 compilation and interpretation of the existing data.
This 15 gets into something Joe was talking about before, in 16 addition to data sources from RES and NRR considering the 17 IDCOR reports, Special Technical Report 85.2, I guess it is, 18 which is IDCOR's analysis of the NRC technical issues and 19 other sorts of data as well.
20 Given that, there is an interpretation phase 21 where we try to make comparisons between the data that is 22 available here and then use that to develop the types of 23 ranges and subjective weights that Joe described a little 24 while ago in his uncertainty analysis.
25 (silde.)
7ms l
J v
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800 336-6646
5840 09 09 109 1
DAVbur 1
In support of the development of these ranges of f
2 weights, we have time to permit the identification and J
3 performance of some supplemental calculations.
The amount 4
of work that has to be done will vary a good bit from issue 5
to issue.
In some cases there will be essentially no new 6
work.
In others we are almost starting fecm scratch.
7I This will all be folded together, the short term 8
work and the existing data, to develop the position paper.
9 This whole process then is going to be overseen by senior 10 staff and then the management will focus on that.
11 (Slide.)
12l In general, we have tried to define the issues in terms of our major headings or intermediate products in the 13 t
i
'x
/
14 SARRP analysis; that is, in terms of accident sequence 15 frequencies, source terms, or the various parts of the 16 containment, either loads, structural response, or the 17 failure mode probabilities, so that you don't end up in a 18 situation where your issue result, for example, is a 19 variation in the mass of thorium expected to exit the 20 vessel.
21 That, in itself, is useful information, but 22 doesn't translate into something -- in itself, it doesn't 23 provide you much risk information.
Only when that is 24 translated into a containment load source term or something 25 like that can you really bring it into the risk picture.
,s I
I
'J ACE-FEDERAL REPORTERS, INC.
202 347-3700 Nationwide Coverage 804336-6666
5840 09 10 110 1
DAVbur 1
DR. KERR:
Do you have a good definition of what
(}
2 you mean by containment failure?
3 MR. CUNNINGHAM:
Jim Costello just left, 4
unfortunately.
He is the one who is dealing with this.
5 DR. KERR:
Does Jim Costello have a good 6
definition of what is meant by containment failure?
7 MR. CUNNINGHAM:
He has a good definition of it.
8 I can't tell you what it is.
9 DR. KERR:
As long as somebody has one.
10 DR. MARK:
And everybody uses it.
11 DR. KERR:
Eventually, other people ought to know 12 about it, but for now --
13 (Laughter.)
14 (Slide.)
15 MR. CUNNINGHAM:
I just want to go through the 16 issues that the staff is considering, and in places where 17 they correspond to IDCOR issues I have identified the 18 corresponding IDCOR issue.
19 An issue that we brought up a little.while ago is 20 the issue of core concrete interactions.
In this issue we 21 consider both the thermal hydraulics and the CORCON aspect 22 of it, if you will, and the VANESA aspect, the radionuclide 23 releases.
We have amalgamated that into one issue.
24 Direct heating is one.
We have included in here 25 something else that Ivan brought up, which is the potential g()s ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 80 4 336-6646
~
5840 09 11 111 1
DAVbur I
for reactor coolant system failure, which would preclude the
{}
2 direct heat expulsion that would result from direct 3
heating.
That is tied up in this issue.
4 In-vessel conditions and dispersal X vessel, 5
hydrogen, both the generation aspect and the burning aspect, 6
the chemical form of iodine within the reactor coolant 7
system and containment.
8 MR. BENDER:
Excuse me.
Before you take that 9
off, are you going to try to develop a probabilistic 10 discussion of those issues, or is this merely a technical 11 discussion?
12 MR. CUNNINGHAM:
Is it technical or is it 13 probabilistic?
Is that it?
-)
s>
14l MR. BENDER:
I am asking whether it is one or the 1
15 other.
Maybe I should have asked is it going to be both?
16 Up to now it has never been probabilistic.
17 MR. CUNNINGHAM:
The intent is to try to make it 18 probabilistic.
There is an engineering aspect of it, if you 19 will, but by the weights that Joe talked about and the 20j degrees of belief, the intent is to try to make it 21 probabilistic as well.
22 MR. BENDER:
The way I heard it discussed before 23 was divide the statistics up in accord with the distribution 24
.of the number of technical experts that had opinions.
25 Is that the way you are going to do it?
,s
(_)
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800 336-6646
5840 09 12 112 1
DAVbur 1
MR. CUNNINGHAM:
It is the same process, yes.
It
,(
')
2 is a judgmental, probabilistic analysis.
'. J 3
MR. BENDER:
It is not all that satisfying, I 4
might say.
5 DR. CATTON:
That is an excellent example, the 6
example I chose for a reason, because it is not 7
probabilistic.
It is deterministic, and somebody has 8
developed a model that hasn't included all the things that 9
ought to be in it.
You can't treat it probabilistically.
10 It is just not right.
You should fix it.
11 Maybe that is a foolish belief.
12 DR. KERR:
Mr. Meyer has already told you that 13 just because some expert disagrees with the NRC they are not I
/
i/
14 necessarily going to change their mind.
15 DR. CATTON:
What about more than one?
16 (Laughter.)
17 DR. KERR:
We have only_ identified one so far.
18 MR. MURPHY:
If I can interject, it is not quite 19 as bad as you are making it sound.
20 What happens basically is -- let's say we have a 21 model of known efficiencies and our base case is based on 22 that model.
People will try to make bounding calculations, 23 hand calculations, based on their understanding of the 24 deficiency and whatever else that has been done.
In large 25 part what has been done that they are relying on are the 7x YJ ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 80 4 336-6686
5840 09 13 113 1
DAVbur 1
questions to get some idea of how large the release could be 3
2 in this phase if they had misrepresented, if the model is (G
3 inaccurate.
4 And if essentially the team of experts agrees 5
that this deficiency is in the code and that this bounding 6
analysis is a reasonable analysis, then I would expect a 7
high weight on the bounding analysis and a low weight on the 8
code.
9 10 11 12 13 m(.)
14 15 16 17 18
. 19 20 21 22 23 24 25 i
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverase 800 336-6646
i 5840 10 01 114 1
1 DAV/bc 1
DR. CATTON:
Somehow, that doesn' t sound very
~')
2 satisfying.
Why don't you fix the code?
(V 3
MR. MURPHY:
Presumably, if we come to the 4
conclusion that there's major deficiencies in the code, the 5
code's going to get fixed.
But, in the interim, we're going 6
to handle it that way.
I I
7 The code fix may take another two years.
8 DR. KERR:
Let the record stipulate that 9
Mr. Catton is unhappy with the VANESA code.
10 DR. CATTON:
I just chose VANESA as an example.
11 I wouldn't want Dave Powers mad at me.
I could pick 12 others.
13 MR. CUNNINGHAM:
As Joe was saying, the process 14 that was gone through a couple of weeks ago in Albuquerque, 15 the discretized set of parameter values that were chosen, 16 for example, for something like VANESA, one of four 17 typically was the VANESA calculation from the source term i
18 l code package calculation, or whatever; and there were others 19 based on quest analyses and things like this.
20 i I don' t recall exactly how they weighted them.
21 But, in many cases, I think they weighted other calculations 22 almost equal in value in their opinions to a VANESA 23 calculation, given that we recognized the uncertainties in i
t 24 VANESA.
25 DR. KERR:
I think what Mr. Catton is saying, I'm 7
(
l ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800 336 6646
5840 10 02 115 1
DAV/bc 1
reluctant to speak for him myself, he's so articulate, but
(~)
2 the f act that you weight against something which is utterly V
3 wrong along with other things that are right doesn't strike 4
him as being a very logical approach.
5 Perhaps you disagree with him that VANESA is 6
utterly wrong.
I'm not trying to take sides on this.
I 7
don't know enough to do so.
But I think I get his point, 8
and I think you do, too.
9 MR. CUNNINGHAM:
Clearly, if something we believe 10 is utterly wrong --
11 MR. BENDER:
Whether it's right or wrong or not 12 will be important only if there is some difference in the 13 way in which the NRC handles its regulatory requirements.
I']'
14l
\\
I guess I would have to use that as the test:
15 What regulatory actions am I likely to consider concerning 16 this technical issue and how might they be influenced by a 17 change in the judgment?
18 That may not be totally probabilistic, but 19 probabilities will enter into it.
20 MR. ERNST:
That's why I think it's very 21 imp.ortant to have this so-called range of expected means, or 22 whatever you want to call the range.
Then, if you 23 understand what drives that range, depending on the 24 regulatory application you choose to make -- you may decide 25 to be at the high end, or you may decide it doesn't make a O
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-336-6646
5840 10 03 116 1
DAV/bc 1
lot of dif ference where you are -- but you can make some I) 2 judgments of where you want to be for a regulatory problem.
's' 3
MR. BENDER:
I like your approach, Mal, but I'm 4
inclined to say that the technical discussion will have to 5
include something that says here are the regulatory 6
decisions that will be made in connection with this matter.
7 Without that, the technical discussions aren't 8
much help.
9 MR. ERNST:
I think you're getting into Chapter 10 4,
that says:
Given all this information, how might it 11 apply in certain kinds of regulatory situations?
12 MR. BENDER:
There's a tendency, when you put the 13 discussion in Chapter 3 and the application in Chapter 4, to
(
)
As' 14 fail to correlate.
15 DR. KERR:
Back to the process.
I hate to say 16 something that's so obvious, but I don' t need to very much.
17 It certainly must have occurred to you as you've done this 18 that you can influence the outcome of this process very much 19 by the experts you choose and by the instructions you give I
20 them.
21 That is a very important part of the total 22 process.
I certainly can't tell you how to do it.
But it 23 will have a strong influence on what you finally did.
24 MR. ERNST:
That's exactly right.
25 MR. CUNNINGHAM:
Going on with the other issues
(
\\._,/
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-336-6646 L-
5840 10 04 117 1
DAV/bc 1
that we're considering in the staff...
(
2 (Slide.)
3
- )
3
...considering the issue of reactor coolant 4
system natural circulation.
This is one case where the 5
issue, in a sense, one of the statements I made earlier that 6
it's a direct input in a sense to the SARRP analysis.
The 7
natural circulation issue is in fact an input to three other 8
issues -- the direct heating issue, that part of it which 9
deals with the potential for failure of the reactor coolant 10 system; this issue, the revaporization of radionuclides; and 11 hydrogen generation.
12 Going on, we have again rovaporization of 13, radionuclides in the reactor coolant system before the
(~ 's i
(/
14 vessel breach and after.
15 DR. CATTON:
Mark, under the first one, do you 16 include the possibility of steam generator tube rupture, and 17 everything else?
When you say " localized", you mean 18 anywhere in the RCS?
19 MR. CUNNINGHAM:
Yes.
Again, the containment 20 response issue, which deals with, again, we have some sort 21 of failure criteria which I've described as gross failure, 22 leakage before that failure, requirements of pressure and 23, temperature that cause that failure and the size of the i
24 resulting hole in the containment.
25!
DR. CATTON:
On this last one, do you include 7s Y.]
ACE-FEDERAL REPORTERS, INC.
202-347 3700 Nationwide Coverage 800-336-6646
5840 10 05 118 1
DAV/bc 1
location of the leak as well?
It seems to me, if you're
(}
2 putting in an enhanced plume rise model here, you ought to 3
be looking at elevation that the release takes place from; 4
or else there's no sense because of the augmentation of 5
atmospheric dispursion.
6 MR. CUNNINGHAM:
Implicit in this is the 7
penetration is low in the building or the spring line.
8 (Slide.)
9 Finishing out the issues are the performance of 10 the secondary containment buildings and the PWR's, the 11 influence of containment venting in boilers in terms of the 12 )
human reliability aspect and the engineering feasibility, 13 the likelihood of having an alpha failure mode steam 14 explosion, and something that I'll call common.cause 15 failures, which is actually one front end or accident 16 frequency-oriented issue here.
17 This is probably going to act as a surrogate for 18 several types of common cause failures.
For the moment, we 19 haven't concluded exactly what would be in that issue.
20 DR. CATTON:
Didn't the steam explosion review 21 group resolve IDCOR No. 7?
22 MR. CUNNINGHAM:
For all intents and purposes, 23 the position paper we see here will reflect the surge work, 24 unless something dramatic happens.
25 MR. ERNST:
That is the intent.
The position ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800 336-6646
5840 10 06 119 1
DAV/bc 1
paper on steam explosions, there is a position.
It will be
()
2 characterized for completeness in NUREG 1150.
V 3
The other issues all have some working to do.
4 MR. DAVIS:
Mark, let me ask you about one I 5
didn't see on the list.
That's pump seal failures.
Is that 6
one you're going to consider?
7' MR. CUNNINGHAM:
At the moment, I believe that's 8
being considered as an issue, not as an issue in this sense, 9
but gleaned from the sensitivity studies as part of the ASEP 10 work.
11 MR. DAVIS:
Let me ask you why.
I'm worcied 12 about it.
Two of the most recent PRA's -- Seabrook and 13 Millstone III -- assume that you do get pump seal LOCA.
But 14 there's a lot of dispute over when it occurs and what the 15 flow rate is.
16 Now, the NRC, in their A-45 decay heat removal 17 program, assumes that it never occurs.
The assumption is 18 being made that it's going to be fixed.
I'm curious as to 19 how you're going to handle this problem if you have both 20 ends of the spectrum being addressed in the NRC and the 21 industry.
22 MR. CUNNINGHAM:
Again, that specific issue is 23 being addressed, as far as I know, in ASEP, so that the 24 accident frequencies predicted for the various plants and 25 the core melt frequency for the various plants will reflect gJg ACE-FEDERAL REPORTERS, INC, 202-347-3700 Nationwide Coverage 800-336-6646
5840 10 07 120 1
DAV/bc 1
a sensitivity if you accept pump seals as always occurring
()
2 or never occurring.
You'll see them on the sensitivity.
In 3
a sense, if need be, that kind of work could be elevated to 4
become an input into a position by staff on that issue.
5 MR. MURPHY:
Pete, I don't have a direct answer 6
for what I know you want.
In the ASEP study, I believe 7
they're making an assumption that if component cooling water 8
to the seal is lost for X-minutes, you get seal failure of 9
size Y.
I can't tell you what the X and Y are.
A 10 MR. DAVIS:
It looks like NRC is going to have 11 two conflicting positions, perhaps, between this program and 12 A-45 on whether it happens or not, and what perhaps the 13 consequences would be.
14 The other one that I'm curious about is the V 15 sequence; the IDCOR analysis for Zion indicates that it 16 won't occur.
Yet, in the Zion PRA done by PLG, it does 17 occur and is a dominant risk contributors for internal 18 events.
19 Now, you're going to have to resolve that 20 somehow, I would guess.
And I don' t know which analysis is 21 correct.
22 MR. MURPHY:
That one will be based primarily on 23 the Sandia review of the PLGd report, where I believe, and 24 I'm going back a couple of years in memory now, I believe 25 they found it was more probable than PLG did.
l l
l i
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-336-6646
5840 10 08 121 1
DAV/bc 1
MR. DAVIS:
But IDCOR says it won't happen.
()
2 MR. MURPHY:
I understand that.
It will be based 3
on the best estimate.
4 MR. DAVIS:
I don't think PLG did any stress 5
analysis for that accident.
They did the same thing that 6
was done in WASH-1400.
They assumed it would occur if the 7
check valves failed.
But IDCOR took it one step further, 8
actually looked at the pressurization of the RHR system and 9
determined that only the pump seal would fail and you'd get 10 a totally different accident that's relatively beneign.
11 MR. MURPHY:
What's happening here is part of 12 these sensitivity issues they are considering event V, the 13 event V consideration that I mentioned earlier is whether o
I 14 it's wet or dry.
Not so much on the PWR's as on the 15 boilers.
There's a real question as to whether the stress 16 on the pipe is such that it would cause a failure.
17 The question exists no the PWR's but on the 18 boilers, the delta P at 1,000 psi less.
But, on the 19 boilers, it is definitely phased into it.
I think it is on 20 the PWR's but I'm not positive.
21 MR. CUNNINGHAM:
This is a somewhat different set 22 of issues than the SARRP program is analyzing.
I think, for 23 Surry and SARRP they are considering the issue of, one, is 24 it in fact a large enough rupture of the system to lead to g-25 core melt?
Two, is it above water or below water?
V)
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-336-6646
1 5840 10 09 122 i
1 DAV/bc 1
I think both issues may be handled as inputs to
(}
2 their uncertainty analysis.
3 MR. MURPHY:
I think that's right.
4 (Slide.)
5 MR. CUNNINGHAM:
Just to show these same issues 6
again, these slides do a couple of things.
One, it just 7
simply indicates what organizations are responsible for 8
developing these position papers.
And it's just a mapping 9
of issues to the six reference plants.
10 What it breaks down to is that there are a number 11 of the issues that are specific to boilers.
Others are 12 specific to PWR's.
There are some that go across the board.
13 DR. CATTON:
What do those initials stand for --
(~)
\\/
14 DSRO?
15 MR. CUNNINGHAM:
That's Dr. Spiess' division at 16 NRR.
The other ones are DAE, Division of Accident 17 Evaluation in Research; DET, the Division of Engineering 18 Technology in Research; ourselves, DRAO.
19 DR. CATTON:
So this is all in NRR then?
20 MR. ERNST:
Most of it is in Research.
21 MR. CUNNINGHAM:
One division within NRR is 22 responsible for their side of these positions.
The 23 separation on the other side goes among the three 24 divisions.
25 (Slide.)
O.
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800-336-6646
5840 10 10 123 1
DAV/bc 1
The product in the sense that we generate here is
/)
(
2 a series of position papers that will end up as-an appendix 3
to NUREG 1150.
The format of the issue papers is very 4
similar to what IDCOR has put together in technical report 5
85.2.
It's pretty straightforward.
Definition of the 6
issue, relevant plants and sequences.
What we think the 7
cange of the issue is.
The importance of issues.
We ig h ts.
8 Comparisons and conclusions on the risk importance of this 9
issue.
Long-term work that needs to be done.
10 Let me make it clear that going from step three 11 to step four is not obvious.
These sets of ranges for the 12 individual issue papers have to be put together and 13 integeated together to come up with the risk to the plant,
(' \\
l
'l 14 and uncertainty of risk to the plant.
And, in fact, the 15 output of that risk is the measure of the importance of 16 individual issues.
17 (Slide.)
18 Just to give you an idea of the schedule, for all 19 intents and purposes, the positions papers are going to be 20 developed and finalized over the next four months.
This is 21 the timeframe that 1150 will actually be written, so they're 22 just in time to get into 1150.
23 And that's all I have to say.
24 DR. KERR:
Are there questions?
25 MR. DAVIS:
What's the RVW oversight?
i l
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 800 336-6646
5840 10 11 124 1
DAV/bc 1
MR. CUNNINGHAM:
I'm sorry.
It should be review
()
2 and oversight.
Data compilation and interpretation.
3 Position paper development.
4 MR. DAVIS:
Thank you.
5 DR. KERR:
Other questions or comments?
I 6
(No response.)
7 DR. KERR:
Thank you very much, gentlemen.
8 I would like your written comments at some 9
reasonable time sent to Mr. Houston.
In the meantime, if 10 you have any specific thoughts you want to make, it seems to 11 me that the approach is a reasonable one.
I look forward to 12 hearing more about it.
It's a tough problem.
I would
/~)
13l encourage that, insofar as you can, that you use not only
~#
14' Latin hypercube sampling but English to find out where the 15 difficulties still lie.
I 16l (Laughter.)
17 DR. KERR:
The meeting is adjourned.
18 (Whereupon, at 4:10 p.m.,
the meeting was 19 adjourned.)
20 21 22 23 24 25 O
ACE-FEDERAL REPORTERS, INC.
202-347-3700 Nationwide Coverage 80 4 336-6646
CERTIFICATE OF OFFICIAL REPORTER
^'
(
%))
This is to certify that the attached proceedings before the UNITED STATES NUCLEAR REGULATORY COMMISSION in the 4
matter of:
NAME OF PROCEEDING:
ADVISORY COMMITTEE ON REACTOR SAFEGUARDS SUBCOMMITTEE ON CLASS 9 (SEVERE) ACCIDENTS DOCKET NO.:
1 PLACE:
WASHINGTON, D.
C.
DATE:
WEDNESbAY,' JANUARY 29, 1986 were held as herein appears, and that this is the original transcript thereof for the file of the United States Nuclear Regulatory Commission.
)
.l Q i
( sigt ) /1_ m ed M (TYPED)
DAVID L.
HOFFMAN Official Reporter ACE-FEDERAL REPORTERS, INC.
Reporter's Affiliation 4
4
-m-.
m
. _ _ -,.. _ ~. -, _ _..
'e l
NUREG-1150 STATUS J. A. MURPHY JANUARY 29, 1986 DRA0-86026 1/27 l
- +e.ua.i.9pqqm.pv.3 =rp".w amippnv.w e... -
1r, T
e i
o RISK PERSPECTIVES AND PROFILES ON SIX REFERENCE PLANTS:
SURRY i
PEACH BOTTOM '
6
' s, g
SEQUOYAH l
GRAND GUI.F ZION -
I c
I-LASALLE o
RISK REDUCTION POTENTIAL l ;
GENERIC NUREG-0900 FIXES (E.G., FILTERED VENTED l
CONTAINMENT, ADD-ON DECAY HEAT REMOVAL, ADDITIONAL CONTAINMENT HEAT REMOVAL ETC.)
PLANT-SPECIFIC MODIFICATIONS SUGGESTED BY ACCIDENT SEQUENCE REBASELINING AND CONTAINMENT TREE INSIGHTS o
DISPLAY AND CONSIDERATION OF lMPORTANT UNCERTAINTIES o
INSIGHTS ON USE OF RISK INFORMATION FOR PLANT-SPECIFIC AND GENERIC REGULATORY APPLICATIONS I
DRAO-85080 11/C1/ K a.
.. w.-
e o,
RISK E00ATION
- EE No N,y CONS W,17 )
RISK g
g g
g I
J SYMBOL DEFINITION SOURCE FRE0 FREQUENCY OF CORE MELT ACCIDENT SEQUENCE I ASEP. IN COORDINATION WITH SARRP 1
CRMP,y PROBABILITY OF CONTAINMENT RELEASE SARRP CONTAIW1ENT EVENT ANALYSIS y
MODE 7. GIVEN ACCIDENT SE00ENCE I l
FISSION PRODUCT SOURCE TERM FOR BIN TO BCL STCP CALCULATIONS BINNING FP,17 l
{
WHICH SEOUENCE I WITH RELEASE MODE 7 MEETINGS. AND SARRP ANALYSES IS ASSIGNED CONS, p,
)
M NEWNNEK.NN MWMNRNWUN I
l SOURCE TERM FP,17 i
9 O
e
2 t
ig I
i
.i ACTIVITIES IN PROGRESS o
UPDATING AND REBASELINING SIX REFERENCE PLANT PRAS o
CONTINUING CONTAINMENT EVENT TREE ANALYSIS AND RISK REDUCTION ANALYSES FOR REFERENCE PLANTS o
SUBSTANTIALLY SUPPLEMENTING THE SEQlJENCES INVESTIGATED IN BMI-2104 TO INCLUDE ADDITIONAL POTENTIALLY RISK-DOMINANT SEQUENCES o
PREPARING RISK REFERENCE DOCUMENT SUMMARIZING THE RISK ASSOCIATED WITH THE SIX REFERENCE PLANTS AND INCLUDING A CHARACTERIZATION OF THE ASSOCIATED UNCERTAINTIES I
A.
j O
O O
9 G EN ERA A3 3ROACF (Surry, Peach
- Bottom, Sequoyah, Grand Gulf) l 1.
Preparation Phase 2.
Plant Visit Event and Fault Tree Modeling 3.
4.
Base Case Quantification 5.
Uncertainty and Sensitivity Analyses 6.
Confirmatory Plant Visit 7.
Revision and Documentation
- Technical QC Review at Each Phase
- Senior Consultant Group Review of Approach and Results
- Continuous NRC-DRAO Interactions O
e
e SE I
D.
'U TS REH P_
e s
s s
s s
s T
E s
m s
s s
n s
s O
M e
o e
e e
e e
e s
I R
L s
L L
L L
L L
OT e
e e
e s
s s
r e
m m
m s
s s
o P
r D
E o
o o
o e
e e
E R
3 M
s s
S L
L L
M I
RA P_
M M
e e
e e
e e
e e
P A
S r
r r
r r
r r
r r
O S
0 o
o o
o o
o o
o C
R M
M M
M M
M M
M M
L 0_
0 IA 4
T 1
E e
e e
e e
e s
m D
H e
e m
r r
r s
o o
S m
r r
o e
o O
W S
M M
S M
L S
M M
F A
o o
o o
L e
E o
s s
t s
f t
r r
u V
n n
o o
a E
e t
p C
I L
f v
e
/
p o
E o
u n
l F
y n
as S
s 0
y g
r ol o
io cm m
m 5
c n
n e
t i
r e r
t a
v ar t
e a
i et m
1 E
r o
u n t
ct 1
U u
a o
m c
t s
h s t y o
u e
co ey t
c t
t n
H R
S l
c a
i S
A D
G I
.I ER UN i,
i';1
I l
ACTUATION AND CONTROL O
ACTUATION AND CONTROL CIRCUITRY TREATED AS A " BLACK BOX" O
DC POWER INTERFACE IS EXPLICITLY MODELED 0
AC MOV CONTROL POWER GENERALLY ASSUMED TRANSFORMED FROM MOTIVE POWER SOURCE 0
LOCAL CONTROL CIRCUITS GENERALLY CONSIDERED PART OF CbdPONENT O
SS 4
DRA0-86020D 1
/86
s t
0
'e
,r a
O ELECTRICAL SYSTEMS GENERALLY MODELED AT ll160V LEVEL PROVIDED INDEPENDENCE EXISTS AT LOWER VOLTAGES l
0 SUBTLE INTERACTIONS BASED ON SURVEY OF PLANT FOR INTERACTIONS OBSERVED IN PREVIOUS PRAS OR THE HISTORICAL RECORD 0
INSTR 0 MENT AIR GENERALLY MODELED AS SINGLE FAILURE l
l I
DRA0-86021D 1/
/86
k mi. : 44F(4'QS%6p3mTyympg. gs..,, e n.
o
- I
.z.
,1 i CONTAINMENT ANALYSIS n\\
s DETAILED TREES FOR. REFERENCE," PLAdTS BE NG PREPARED UNDER o
l}
SARRP l
p PEER REVIEW OF PLANT-SPECIFIC TREES COkiPLETED BY UNIVERSITY o
i OF WISCONSIN -
'l PROGRAM CONSIDERS ACCIDENT PROGRESSION,' FISSION PRODUCT o
RELEASE, MIGRATION AND REMOVAL MECHANISMS, AND PHENOMENOLOGY UNCERTAINTIES AND PROCESS-INDUCED FAILURES
.i l
o CONSIDERATION OF IDCOR AND APS lSSUES IN UNCERTAINTY l
ANALYSIS
^
lj
.i I
l i
I,e s
DRAO-85056 1/27/86 1
G G
G
GENERALIZED CONTAINMENT EVENT TREE (SCHEMATIC)
)IMPUTS OUTPUTS PLANT DAMAGE 1.
CORE MELT OR SUCCESSFUL STATE REC 0VERY (LINKED TO ASEP)
I
- 3.. CONTAINMENT RELEASE MODE
~
DEFINITION FREQUENCIES FREQUENCY 2.
SOURCE TERM BIN FREQUENCIES QUESTIONS ASKED ABOUT
- OPERATOR ACTIONS, E.G.
- IS THE VESSEL DEPRESSURIZED?
- ARE DRYWELL SPRAYS ACTUATED?
- IS THE CONTAINMENT VENTED?
i
() EVENTS & TIMING, E.G.
- CORE DAMAGE BEFORE CONTAINMENT FAILURE?
- CONTAINMENT FAILURE BEFORE, AT, OR AFTER VESSEL BREACH?
- PLANT CONDITIONS, E.G.
- WHAT IS THE CONTAINMENT LEAKAGE LEVEL?
f
- WHAT IS THE POOL BYPASS FLOW LEVEL?
- PHENOM'ENA, E.G.
- CONTAINMENT FAILURE PRESSURE?
- CONTAINMENT PRESSURE RISE DUE TO H2 BEFORE VESSEL MELTTHROUGH
- HYDROGEN BURNS IN REACTOR BUILDIN6?
O
,t'
~ ~
\\
()
SOURCES OF INFORMATION MRC-SPONSORED STUDIFS CONTAINMENT LOADS WORKING GROUP (CLWG) o CONTAINMENT PERFORMANCE WORKING GROUP (CPWG) o SEVERE ACCIDENT SEQUENCE ANALYSIS (SASA) PROGRAM o
BATTELLE ANALYSES FOR ACCIDENT SOURCE TERM PROJECT OFFICE (BMI-2104) o REACTOR SAFETY STUDY (RSS) AND SUBSEQUENT RISK ASSESSMENTS o
STEAM EXPLOSION REVIEW GROUP (SERG) o SENERIC SAFETY ISSUE STUDIES (TAP A-43, A-44, A-45, ATWS) o OTHER STUDIES BY THE NATIONAL LABORATORIES (SNL, BNL, PNL) o
[])
REPORTED IN NUREG DOCUMENTS i
UTILITY
~
INDUSTRY-SPONSORFD STUDIES o
FULL SCOPE PRAs INDUSTRY DEGRADED CORE (IDCOR) PROGRAM o
FINAL SAFETY ANALYSIS REPORTS (FSARS) o PLANT-SPECIFIC EMERGENCY OPERATING PROCEDURES o
STATION BLACK 0UT DATA AND ANALYSIS (EPRI) o (2) i
~
SOURCE TERM METHODOLOGY METHODOLOGY IS ESSENTIALLY THE SAME AS USED IN BMI-2104 WITH SOME IMPROVEMENTS l
SOURCE TERM CODE PACKAGE--INTERIM VERSION I
MARCH 3 j
CORSOR INTEGRATED INTO MARCH I
CONSISTENT TREATMENT OF,FP RELEASE FROM FUEL l
CORCON REPLACES INTER IN' MARCH l
CONSISTENT TREATMENT OF CORE-CONCRETE ATTACK l-TRAP-MERGE COUPLED ANALYSIS PERMITS' TREATMENT OF REVAPORIZATION IN-VESSEL i
VANESA, NAUA, SPARC, ICEDF SAME AS BMI-2104 EXCEPT INTERFACES REQUIRE LESS USER INPUT SURRY ANALYSIS RELIES HEAVILY ON NUREG-0956 O
O DRAO-86032 1/28 6
g l
l BINNING PROCESS 0
DOMINANT SEQUENCES IDENTIFIED BY REBASELINING ANALYSIS 0
CONTAINMENT EVENT TREE REDUCED TO KEY FAILURE MODES O
SPECIFIC NEED FOR STCP RUNS DETERMINED BY JUDGEMENT OF ANALYSIS TEAM AND SOURCE TERM EXPERTS (SNL, BCL, BNL, ORNL, NRC) 0 RELEASE CHARACTERISTICS GENERATED FOR EACH BIN DRAO-86022 1/d 86
0 o
4 NEW FEATURES IN MACCS 1.
TREATMENT OF TIME DEPENDENT RELEASES 2.
IMPROVED ATMOSPHERIC DISPERSION CROSS WIND DISTRIBUTION, PARTICLE SIZE DISTRIBUTION, UP TO DATE PLUME RISE MODEL 3.
ADDITIONAL EXPOSURE PATHWAYS (FOOD PATHWAYS, DEPOSITION TO SKIN, EARLY RESUSPENSION) (ADULT DOSE CONVERSION FACTORS, SKIN DOSE, POPULATION DOSE) 4.
IMPROVED DOSIMETRY 5.
EMERGENCY RESPONSE (GRADED RESPONSE, ROAD NETWORK EVACUATION, DELOCATION OF SHELTERED INDIVIDUALS, RELOCATION OUTSIDE EPZ) 6.
LONG-TERM PROTECTION (LONG-TERM GROUNDSHINE, DIRECTOR DEPOSITION ON CROPS) 7.
HEALTH EFFECTS (REVIEW AND UPDATE WASH-1400, HAZARD FUNCTIONS USED FOR EARLY EFFECTS, ADDITIONAL INJURIES, LATENT EFFECTS--BEIR III, UPDATED GENETIC EFFECTS)
/
DRA0-86033 1/2 6
i-------
.f-
}
O NUREG-1150 SENSITIVITY / UNCERTAINTY ANALYSES e
i OBJECTIVES:
PROVIDE A GOOD ENGINEERING PERSPECTIVE ON MODELING 1.
ASSUMPTIONS THAT DRIVE THE ANALYSIS (SENSITIVITIES PROVIDE A " REASONABLE" ENVELOPE IN WHICH THE ACT 2.
VALUE WOULD LIKELY BE FOUND AND WHICH NEED NOT BE EXPRESSED IN TERMS OF FORMAL STATISTICAL BOUNDS
.(UNCERTAINTY)
O i
f l
I' O' '
d.,.
~ ~ - - ~ ~
O aVREG-1150 SENSITIVITY / UNCERTAINTY ANALYSES OBJECTIVES:
1.
PROVIDE A GOOD ENGINEERING PERSPECTIVE ON MODELING ASSUMPTIONS THAT DRIVE THE ANALYSIS (SENSITIVITIES) 2.
PROVIDE A " REASONABLE" ENVELOPE IN WHICH THE ACTUAL
{
VALUE WOULD LIKELY BE FOUND AND WHICH NEED NOT BE EXPRESSED IN TERMS OF FORMAL STATISTICAL BOUNDS (UNCERTAINTY) nU o
i O
y f
3 i
ACCIDENT FREQUENCY REBASELINING 0
DATA--USE LOG-NORMAL DISTRIBUTIONS AND PROPAGATE'THROUGH FAULT TREES AND EVENT TREES l
0 MODELING ASSUMPTIONS--PERFORM SENSITIVITY ANALYSES ON KEY ASSUMPTIONS a
O e
DRA0-86023 1/ F/86
~
s O
CORE MELT FREQUENCY DISPLAY y
i L
(
~
C B
A<
~
MOST PESSIMISTIC SENSITIVITY
~
O
~ '-
BASE CASE
~~
MOST OPTIMISTIC SENSITIVITY CASE A.
RANGE OF MEANS B.
"5/95" - BASE CASE ONLY C.
"5/95" - EXTREMES OF SENSITIVITY CASES e.
i i
O l
'O
+
-~-, Q.
,e.
i UNCERTAINTIES IN MODELING OF SOURCE TERM AND CONTAINMENT PHENOMENOLOGY PERFORMED USING LIMITEDLATINHYPERCUBEMEtHODS 0
CONSTRAINED (' STRATIFIED) MONTE CARLO SAMPLING SCHEME (NUREG/CR-3624) 0 REQUIRES SUBJECTIVE ESTIMATES OF POSSIBLE RANGES ON PARAMETERS BASED ON EXPERT JUDGEMENT
'l, UTILIZES 3-5 SPECIFIC VALUES WITH WEIGHTS TO EACH VALUE
.l WEIGHTS ROUGHLY AKIN TO DEGREES OF BELIEF 1
SHOULD BE INTERPRETED AS ENGINEERING JUDGEMENT, NOT PRECISE STATISTICAL lf PARAMETERS l'
l 0
CONSTRAINED TO 10-15 IMPORTANT ISSUES AND 20-30 RUNS PER PLANT BY AVAILABLE RESOURCES O
CORRELATIONS BETWEEN ISSUES EXPLICITLY CONSIDERED DRAO-86024 1/
/86
c..
LIMITED. APPLICATION OF LATIN HYPERCUBE SAMPLING DISADVANTAGES:
1.
DISCRETIZATION AND WEIGHTING 0F INPUT PARAMETER SPACE REQUIRES ENGINEERING JUDGEMENT, 2.
RESULTS ARE CONDITIONAL UPON A,
DISCRETIZATION AND WEIGHTING OF INPUT PARAMETER SPACE B.
LIMITED NUMBER OF ISSUES CONSIDERED 3.
SEPARATE CALCULATIONS BASED ON RESULTS OF EXISTING STCP RUNS WILL BE REQUIRED T0 QUANTIFY SOME ISSUES, O
O O
~/
DRA0-86034 1/28/86
i l
SUMMARY
i l
l REBASELINING USING COMPRESSED PRA METHODS WHICH APPEAR HIGHLY COST EFFECTIVE EVALUATION OF ACCIDENT PHENOMENOLOGY WITH BEST TOOLS AVAILABLE 0FF-SITE CONSEQUENCE ANALYSIS USING IMPROVED CODE UNCERTAINTY ANALYSIS DESIGNED TO YIELD REASONABLE ENGINEERING PERSPECTIVES ON POSSIBLE RANGE OF OUTCOMES CAPABLE OF FOCUSING ATTENTION ON IMPORTANT ENGINEERING AND PHENOMEN0 LOGICAL ASSUMPTIONS OR LACK OF KNOWLEDGE UNCERTAINTY ESTIMATES NOT RIG 0ROUS If/86 DRA0-86025
m
~
S A--
-OR NUR-G
50 l'
i MARK A.
CUNNINGHAM i
l OFFICE OF N UC LEAR R EGU LATO RY RESEARCH DIVISION OF RISK ANALYSIS AND O P ERATIO N S 1
I JANUARY 29, 1986 i
B1 1-28-86 O
O O
m TI-l E PURPOSE OF THE NRC WORK ON POSITION PAPERS IS TWOFOLD:
T O DE VE LO P. FOR NUREG-1150 AND THE IDCOR RESOLLITION PROCESS, STAFF POSITIONS ON THE RISK IMPORTANCE OF A SET OF MAJOR ISSUES, DRAWN FROM:
IDCOR-NRC TECl-INICAL ISSUES f
ASEP-SARRP REFERENCE PLANT STUDlES N U REG-0956/B M I-2104 TO SUPPORT THE SARRP AND NUREG--1150 REFERENCE PLANT UNCERTAINTY / SENSITIVITY ANALYSIS WITH BEST--AVAILABL.E INFORMATION i
S.
- - ~
..,,.. +.
NRR AND RES STAFF AND CONTRACTORS ARE SUPPORTING TI-lE DEVELOPM ENT OF POSITION PAPERS THROUGH A SERIES OF STEPS
\\
l l
STEPS 1
ISSUE IDENTIFICATION 2.
SELECTION OF ISSUE TEAMS j
3.
C( >MPIL ATION AND INTERPRETATION OF EXISTING DATA i
DATA GOU RCES RES NRR IDCOR OTHER I NTERPRE TATION COMPARISONS DEVELOPMENT OF RANGES, WEIGH ~is
._ __ _ _-- Q- -
Q v,.. :,
fif EI"51 CONTINUED),
4.
IDF NilFICATION AND PERFORMANCE OF SI-lORT-TERM ANALYSFS 5.
DEVEl OPMENT OF POSITION PAPER G.
COORDINATION / REVIEW BY STAFF / MANAGEMENT SENIOR STAFF COORDINATING GROUP MANAGEMENT OVERSIGi-IT/ CONCURRENCE e
, _..... e e
,h i
i l
l 1
IN
- GENERAL, THE NRC ISSUES ARE DEFINED IN TERMS OF MAJO R SARRP INTER M EDIATE PRODUCTS:
ACCIDENT SEQUENCE FREQUENCIES SOURCE TERMS CONTAIN M ENT
.j L.O A D S (PRESSURE / TEM PERATURE)
STRUCTURAL RESPONSE FAILURE MODE PROBABil' TIES og
ISSUES J
CORE-CONCRETE INTERACTIONS THERMAL-HYDRAULICS (IDCOR #10)
RADIONUCLIDE RELEASES (IDCOR #9)
DIRECT HEATING (IDCOR #8)
POTENTIAL FOR INDUCED RCS FAILURE IN-VESSEL CONDITIONS l
EX-VESSEL DISPERSAL HYDROGEN GENERATION (IDCOR #5)
BURNING (IDCOR //17)
,i CHEMICAL FORM OF LODINE I
l WITHIN RCS i
1 CONTAINMENT RS 1 --2tb86 O
O O
lSSUES (CONTINUED) j RCS NATURAL CIRCULATION (IDCOR #2)
LOCALIZED OVERTEMPERATURE FAILURE OF RCS REVAPORIZATION OF RADIONUCLIDES HYDROGEN GEN ERATION REVAPORIZATION (IOCOR #11)
PRE-VESSEL BREACH POST-VESSEL BREACH CONTAINMENT RESPONSE (IDCOR #15)
LEAKAGE AREA BEFORE " GROSS" FAILURE PRESSURE / TEMPERATURE LOADS FOR " GROSS" FAILU RE SIZE OF RESULTING HOLE 86 1 -2a --80 0
0 O
- - - - ~ -
ISSUES (CONTINUED) 1 SECONDARY BUILDING PERFORMANCE (IDCOR //16)
VENTING i
i^
HUMAN RELIABILITY ENGINEERING FEASIBILITY l
STEAM EXPLOSIONS (IDCOR #7) i i
PROBABILITY OF e4 FAILURE MODE i
l COMMON CAUSE FAILURES D7 1 8G O
O O
i F
l ii x
x xxxxxx t';
I st 9
b i
j I
k, X
X X X X X X i
e i
l i
i l@j i
I X
X X XXX XX X xxxxxx xx l
l l[
(O W
x x
xxxxxx l
[9 i
7/
U X
XXXX XX m
r-Z<
n n
a w
O fq O
^
n^
O y
n O~Ob s,
O^m O O< C N W O
g
-d 4
z-
^
C< m E
- tp C s i,, 0 t_
U O OEO C WdO W
Z x
W4 C
K w<
O 44 vo 4 4 4 400%
y ov c L_
eOw ww v
W 0
W m Z < w < LJ <ea<Ov"oq0 g
Z o o
zM F-Ov gOW Qv c
^
v E
Z p
v_
v M O Wk CO@
r L O L OJ 2 E C C 2 9" vi eZw8f7
_J 4 --
1 wW 5 r ua et
- Q
- v. 2 o<t
,x
.; o, d Ld I
- N o <. N
.v tru i
2 rv wi v m
j CH 2
% v pZ t ZvC"J Z @< 3 I
6 I
l M,9y n
l Cg
%-U l
i r,
y L-( r
>u I
QE
~j -
e i
e: UWVe,0 z
l' t
i 2
.w n
1 2 -
t i
g vm 2
w e
l b
2 l
O O
l U
U l
1 i
t h
I i
i
+
t
i i
THE POSITION PAPER FO'RIUAT IS SIMILAR iO THAT USED IN IDCOR TECHNICAL REPORT 85.2 i
1.
DEFINITION OF ISSUE 2.
RFl_EVANT PLANTS AND SEQUENCES 3.
ISSUE RANGE l
l A.
IMPORTANT SUBISSUES
- i DEFIN ITION RAN G ES WEIG HTS j
COMPARISONS (BMI-2104, SARRP, IDCOR, ETC.)
8.
METHOD FOR DEVELOPING ISSUE RANGE FROM SUDISSUE RANGES CODE CALCULATIONS C.
RESU LTS FOR ISSUE RANGE WITH COMPARISONS (BMI-2104, SARRP, IDCOR, ETC.)
4.
CONCLUSIONS ON RISK IMPORTANCE 5.
LONG-TERM WORK 6.
REFERENCES D11 1-28-06 O
O O
a POSITION PAPER SCHEDULE LEGEND DATA COMP /INTERP x
POS. PAP ER DEV.
o t
RVW/ OVERSIGHT o
a NUREG 1150 3
x t
l
=
a i
=
=
=
a n
JAN FEB MARAPR MAYJU N JUL AUG B12 1 -- 2 8-8 G O
O O
m
1.
i ACRS SUBCOMMITTEE MEETING NUREG-1150 JANUARY 29, 1986 i
1.
OVERVIEW (PURPOSE, CONTENT, SCHEDULES, ETC.)
M. L. ERNST 2.
METHODOLOGY AND UNCERTAINTIES J. A. MURPHY 3.
NRC POSITIONS ON MAJOR ISSUES M. A. CUNNINGHAM DRAO-86027D 1/27/86 O
1 *.
i l
PURPOSE OF NUREG-1150 PROVIDE GREATER UNDERSTANDING 0F CURRENT RISKS FROM SEVERE ACCIDENTS AT o
l POWER PLANTS INTERNAL EVENTS AT SIX REFERENCE PLANTS DIFFERENT CONTAINMENT DESIGNS TMI FIXES IMPROVED SOURCE TERM INFORMATION ASSESS USEFULNESS OF SUCH INFORMATION T0:
o EXTRAPOLATE TO OTHER PLANTS MAKE PLANT-SPECIFIC AND GENERIC REGULATORY DECISIONS PRIORITIZE AGENCY RESOURCES l
DRA0-86028D 1/27/86 llI
. --w
])
C RN
(
O
{
i 'k, o
[RP k
{R[t P
N i
e N
)P T
M C
R_
A A
S_
o 7S YL A
tkN A\\
g 7
G eo N
\\
)R; e
P U
)
)
M P
E S
P R
F G
R
\\W 1
O V E
T T U
C C
MW N
R e
A N A N
C t
\\E T M R S
No IS E
A AIS (P U
M T
O 1
L
(
g"8 CnN O
B TS Q
(
N H
RY
{Y S
A Ru C
EL SL K B o
N A CA A
ECNN hN WR N E
E R
PTUA
)Q'A ss r
r r
kS I
)
L_
R P
C R
hs,E E
O S
8 R
R kD A
k A
hU
(_
(
1 h(3,U (T
hS A
N S
T I
E R QR T
T k' 'OD QU, hP MF hN RE UO C
A O4 L
D t
S L
+
~
.h S
/%
L NU N8$
E 8 R
E2/
4
'L tt LL e
'(C_f 0
G t
'H 8 Q 1
,N
'C 2
I
' N A
I
" 'C e
A M
.D T
B 8
[l 0-r
').
M$s)E
\\P.
M P
L O~
R E
E T
N E(
R T
S T~
A S
A T
N R'
S Y
0'Gu(I
(
8Ns T
(
6 S'
H'I A
L a
kTi C'T kNE A
D T_N A N N$
l Q
L'T Q
v f
CE a P'S i.[
O
e
)C N
R 0 O N
5I
(
1 T 1 A u
u R
GA EP RE UR NP M
A ET ie T
S N
T R
T E
C R
R M
A A
E S
p S S
P S
MI
~
X
(
E S
E I/
S S
E Y E
S U L W
E A
L A E M R
s AN V A T
e I
m V A E
T E
E T N
T I
t R
A E
r L O V
E E
C A
R N e T
U l
N O
R A E
1
)P T e M
9 E
)
/
)
)
P R
X s A
O C
S f
M P
E S
G M
R E
1 O Y E
C C
N R
E A
M A
N N A T
C T
T U
T N R N
C I
t L
s O O S t
s
(
c o
E S (P US M
L
(
I A
O AI I
B KT A
C
(
E T
B TS QIS S
CISC j
H RY E Y A
ERU h
C EL SL K S L
D A CA NA S E EFE ECNN ON IRR SOR PTUA CA 2
O-r NE.
)
y
?
O P
S PE M
C R
E S
R S
R R
(D A
L E S A
k S
HM
(
ASC 4 {)t
[
A T N S
A N
N U E T.
D'S O E R
%Q i
C Y
IT RE l
OUD VQA f
DOO 9NN D
A k
1
=
{-
<l
^
S
/O
^
A.Ni Ld N
a[S U
R
/L L M 4
0 R
,.O
~
1 m
2 C
~
B
- d T( 5 MS I
%'E\\)$k$
I i
M P
O k T%)P E
E T
R MEENR N
\\
I N
A EMNS Y
GS
(
k'T i h I
S R P h'T S
N O
a
,E Y
l T L TL h N D\\k hNE NA itau EN EV IsLT vA VE QED N e P S; 27-
1/28/86 l '.
(REVISED)
SCHEDULE FOR NUREG-1150 AND NRR SUPPORT ACTIVITY SURRY PEACH BOTTOM SEQUOYAH GRAND GULF ZION LASALLE 1.
ACCIDENT SEQUENCE C
C C
C C
2/13/86 INITIAL INPUT 2.
SOURCE TERM C
C C
C C
2/13/86 BINNING 3.
NUMBER OF SOURCE 4
6 8
3 4
3-6 TERM CODE RUNS 4.
SOURCE TERM CODE C
C C
C 1/31/86 3/27/86 RUNS 5.
RELEASE CHARACTER-C C
C 1/30/86 1/31/86 4/21/86 ISTICS 6.
CONSEQUENCE CALC.
C C
2/3/86 2/13/86 2/28/86 5/5/86 7.
REFINE ACCIDENT C
C C
C C
3/12/86 SEQUENCES 8.
CONTAINMENT TREES C
2/7/86 C
3/3/86 3/31/86 6/4/86 DRAFT REPORT 9.
BASELINE RISK C
2/12/86 2/17/86 3/6/86 3/14/86 6/4/86 CALCULATION
- 10. RISK / RISK RED.
C 2/28/86 3/10/86 4/15/86**
3/28/86**
6/16/86**
TABLES
- 11. RISK / RISK RED DRAFT 4/2/86**
5/5/86**
5/29/86**
5/13/86 4/30/86 7/15/86 DETAILED RPT.
- 12. RISK / RISK RED. FINAL 5/29/86 7/1/86 7/24/86 7/10/86 6/30/86 9/10/86 DETAILED RPT.
INCLd!!bFINALSENSITIVITYANALYSESANDUNCERTAEITYRANGES lll
'}
CONTENTS OF NUREG-1150 1.
SUMMARY
OF RESULTS AND CONCLUSIONS 2.
INTRODUCTION 2.1.
PURPOSE 2.2.
OBJECTIVE 2.3.
PLANNED USE 2.4.
BACKGROUND I
2.5.
SCOPE AND METHODOLOGY 2.6.
TREATMENT OF UNCERTAINTIES 2.7.
PRESENTATION OF MATERIAL 2.8.
FUTURE WORK 3.
MAJOR FINDINGS 3.1.
GLOBAL PERSPECTIVES ON CORE-MELT FREQUENCY AND REACTOR RISK 3.2.
CORE MELT FREQUENCY ESTIMATION 3.3.
CONTAINMENT PERFORMANCE 3.4.
SOURCE TERMS 3.5.
OFFSITE CONSEQUENCES 3.6.
RISK-REDUCTION POTENTIAL 3.7.
EXTRAPOLATION TO OTHER PLANTS I!I DRA0-86029D 1/27/86
1 *.
i NUREG-1150 CONTENTS (CONTINUED) 4.
REGULATORY USES 4.1.
INTROD0CTION 4.2.
USE OF PRA-BASED INFORMATION IN DECISIONMAKING 4.3.
PLANT-SPECIFIC REGULATORY APPLICATIONS 4.4.
GENERIC REGULATORY USES 4.5.
PRIORITIZATION OF AGENCY REQUIREMENTS 4.6.
REGULATORY REFORM APPENDIXES:
A.
METHODOLOGY B-G. PLANT-SPECIFIC RESULTS (SURRY, PEACH BOTTOM LASALLE, GRAND GULF, SE000YAH, ZION)
H.
NRC POSITIONS ON RISK-IMPORTANT ISSUES I.
UNCERTAINTY ANALYSIS J.
COMPARISON OF RESULTS K.
TECHNICAL ANALYSES OF SPECIFIC REGULATORY ISSUES O
llI DRA0-86030D 1/27/86 1lI
l'.
i i
j FUTURE SCHEDULE t
i o
PUBLISH DRAFT NUREG-1150 AUGUST 1986 o
PUBLISH FINAL TECHNICAL REPORTS MAY-SEPTEMBER 1986 o
PUBLIC COMMENT SEPTEMBER-NOVEMBER 1986 OCTOBER 1986 o
WORKSHOP o
FINAL NUREG-1150 SPRING 1987 o
ANALYSIS OF B8W AND CE PLANTS SUMMER 1987 o
ASSESSMENT OF NEW RESEARCH RESULTS FY 1987 o
UPDATE OF NUREG-1150 AS WARRANTED DRA0-86031D 1/27/86 O
_