ML20010C432
| ML20010C432 | |
| Person / Time | |
|---|---|
| Issue date: | 07/23/1981 |
| From: | NRC |
| To: | |
| Shared Package | |
| ML20010C401 | List: |
| References | |
| NUDOCS 8108200074 | |
| Download: ML20010C432 (86) | |
Text
~ _g l
1 i
e l,
1 UNITED STATES OF AMERICA i
2i NUCLEAR REGULATORY COMMISSION i
4I PUBLIC MEETING I
5; SAFETY GOAL WORKSHOP e
e i
k%$
6[
PLENARY SESSIOt 2 L'5 7
Cliffside Inn sj 8
Harpers Ferry, West Virginia d
Ci 9
"'hursday, July 2 3, 1981 i
i 10 i The metting was convened, pursuant to notice, at 5
h 11 !
3:10 p.m.,
with Dennis Rathbun, Office of Policy Evaluation, 1
5
\\
f 12 l presiding.
E.
13 +
PRESENT:
=
l 14 !
DENNIS RATHBUN, Chairman EDWIN ZEBROSKI t:
I 2
IS l HERBERT J.'C.
KOUTS EDWARD O'DONNELL 6
i j
so.
LESTER LAVE GEORGE SEGE
.6 d'
17 PAUL SLOVIC NORMAN BRADBURN E
18,
THOMAS COCHRAN CHRISTOPHER WHIPPLE f
19,
STEPHEN L. DERBY R.
TALBOT PAGE a
i 20 !
GORDON COREY MR. MALSCH 21,
VOJIN JOKSIMOVIC DOUGLAS E.
MAC LEAN l
l 22 ALLAN C. MAZUR 23]
DAVID OKRENT
]
24 CHARLES B.
PERROW 25 !
ELLYN WEISS I
8108200074 810724 ALDERSON REPORTING COMPANY. INC.
u Q
f 2
l 1
EEEEEEElEEE l
l l
2 MR. RATHBUN:
This afternoon, between 3:00 and probably 3
ten of 4:00, we were going to have panel reports from the 4
various chairmen, and first I would turn to Dr. Kouts for his 5!
report on the quantitative elements panel.
e A
\\,
n 3
6!
MR. KOUTS: We spent all our time talking about the e
i E
i g
7 details of the five stated goals, the five stated criteria,.of 8
the two quantitative goals and the three -- sorry, the two d
=
9 qualitative goals and the three quantitative criteria that were I
E 10 in the draft document, spending not very much time on the end, f
E 11 because time was running out, but in doing this discussion, we
<3 d
12 l-probably went over most of the agenda we had established in the z
5 I
s 13 l first place of seeing how this document stacks up against the
=
i E
14 concepts that were gen erated at the Palo Alto meeting.
du 15 !
I am only going to touch on some of the highlights of Y
J 16 l the conclusions.
We were far from unanimous on just about i
h' 17, anything we talked about, but t here were some -- there were some 0
=
\\
18 j points that generated relative unanimity or represented at le. st
=
u t
I 19 i a major part of the area under a distribution curve, shall we A
i 20 say, and that's about the best we could do on a lot of things.
21f In the first place, on the first qualitative statement i
22[ of the safety goal -- let me see if I can find it --
which says l
23 that members of the public should be provided a level of 4
1 24 protection such that no individual bears a significant risk to a
i 25 l life and health, this general statement was pretty well supported !
3 l
h I
ALDERSON REPORTING COMPANY. INC.
I
3 i*
l I
1 1 3 by the panel with t he proviso that somehow a better guidance i
i 2j be provided as to what the word "significant" means here.
There 3.
ought to be some expansion of the statement of the goal, or j
l 4l perhaps a preamble to it, or an expansion on it to make it clear e
5 what "significant" really means here.
i E
l 8
6l And it is doubtful t hat one could tie down the word 4
M 2
7' "significant" too far without becoming quantitative about it, s
1 8
8.
but an attempt should be made to do that.
a d
d 9
It was concluded that we actually do need more f
a E
10l qualitative safetyvgoals than are stated here.
There should be 3
5 I
i 5
11 some-safety goals which address the likelihood of large i
B l
d 12 :
accidents with large consequences to the public, or large z
5 d
13 l accidents with appreciable consequences to the public, something E
14 l of the sort.
And a statement that there is no significant I
s t
=
E 15 l probability of these or words to that effect.
I don't have a 4'
5 l
j 16 !
good statement of what is desired here, but the general -- the 3
s
)
17 I general character of such a goal I think is pretty clear.
1 6
E 18 People would like assurance that in fact the i
E I
i t
likelihood of accidents that affect the public in a large health A
19 l l
20l or fatality sense is going to be very small, negligibly small if 21 you can accept words like that.
22 ;
Furthermore, there seems to be a fairly strong view 23 that there should be a safety goal which makes it clear that not i
24 only preventive measures are used to achieve this objective, but 25
- mitigation is used also, and this does not necessarily mean that i
i i
l ALDERSON REPORTINC COMPANY. INC.
l l
I
-.. - - =.. _ = _. - -
l 4
I 1
l l
1 i
l 1 !
you have a goal piled on top of a goal.
I mean, that you have, l
l t
i i
2' say, separate probabilities for mitigation and for prevention, l
l 3;
but that somehow it be made clear that mitigation has to be r
I I
4 relied on in order to achieve the goal.
l 5 l!
MR. COREY:
Is that an implication of more mitigation l
g N
j 6l or --
4 g
(
7l MR. KOUTS:
Not necessarily more mitigation, but don't i
j 8!
try to do it all by prevention, that kind of thing.
And maybe d
9 9
even have a comfortable margin for mitigation included in --
E E
10 MR. CO REY :
What do you mean by that, Herb?
4 i
11 g
a j
12 MR. KOUTS:
Sorry?
=
i g
13 !
MR. COREY:
What do you mean by that?
I don't i
m s
3 14 :
understand. What is mitigation?'
E 15 MR. KOUTS:
Well, to try to make the reactor so safe x=
y 16 '
that you don 't need to put a containment around it, that kind of w
i d
17 thing.
Don't make the probability of an accident so small that w
=
l 5
18 i
-- don't feel that you could make the probability of an accident 2
19 f so small that you don't need to put a containment there or an 5
20 l ECCS or hydrogen emission systems, or something of the sort, 21 I whatever is called for.
l 22,
MR. COREY:
I see.
23 MR. KOUTS:
Concerning the quantitative goals, what 24)thequalitative--aretheycalledgoals?
Yes, they are called 1
t 25 l goals.
Concerning the first of these, there was a general view l
I i
l I
ALDERSON REPORTING COMPANY. INC.
l l
l
. =.. -....
P 5
o I
that the range which is stated of between 5 and 10 times 10~
is l
2l probably too large, and there were two -- two statements made 3
that bore on the fact that that is probably too large a number.
l 4!
The first is that there is a general feeling by the public at I
g 5j large that something is not small enough unless it is scmething E
i j
6l like one in a million or something like that, and this is -- this s
g 7'
5 in a million to 10 in a million is probably too large to be
~
7 l
8 considered small in a public appreciation sense.
I d
i i
9i The second is that if you combine latent and early i
o y
10 l fatalities 'in the sense in which this criterion contemplates, z
=
j 11l then you are probably imposing too large a burden on certain B
i f
12 l subsets of the populace for whom this would be considered an i
E i
j y
13 '
abnormally large probability of fatality against their background l
l 14 l probability.
w 2
M i
2 15 Now, this is a point which had been considered in the w
1
=
l y
16 !
ACRS construction of safety goals, where they actually concluded j
I j
d 17 that young girls -- I mean, they recognized the fact that young i
a
=
5 18 i girls have a relatively low probability of death from all causes, i
{
19 l something like one in 10,000 per year, including autom biles and o
5 i
20 !
smoking -- that is why it is young girls and not young boys --
l 21 and this 5 to 10 times 10-6 would be a large number, would be i
22 considered not negligibly small at any rate, compared to that I
23 background number.
24 So this was another source of some concern that that l
I i
l
-6 l
25 l 5 to 10 times 10 1s too large a number in a criterion or the t f
i l
I ALDERSON REPORTING COMPANY, INC.
l y
l l
6 l
i.
i 1
sort.
2i There was a general feeling that it's okay to add 3,
early fatalities and latent fatalities algebraically in a l
l 4l criterion of this sort.
You don't really need to, although there e
5 were some -- I won't say that was unanimous by any means.
In P.
l 6;
criteria of this sort, it should be made completely clear what a
7' accidents or what events are included in the criterion. It is not 5
g 8;
at all clear the way the criteria are stated whether we are d
c 9
considering accidents in a classical sense that reactor safety i
l c
10 considers them.
Are we considering sabotage as well?
Other z
j 11 sources wherebi cffects of this sort might be brought in, too.
3 j
12 No particular answer was suggested to whether you 5
13 i shoula Include, say, sabotage in or out except a realization of j
14 l the fact that it is almost impossible to calculate it, and maybe 2
15 !
if you try to put it in it would be almost impossible to de so, 5
f 16 l but if you don't include it, make it clear that you are not I
d 17 including it and why.
a=
18 l It was pointed out that the qualitative criteria, the E
I 3
19 l second qualitative criterion has in it a reference to an ALARA i
20 l type concept and the quantitative criteria do not have any ALARA 21!
type criteria.
This is a discrepancy which we 411 need to 22 !
resolve, which should be resolved in any -- in any further i
l 23 version of this document.
l j
24 It was also pointed out that -- it was also pointed out 1
i 25 l that the reference to other technologies, which is in the i
ALDERSON REPORTING COMPANY. INC.
l l
l
~. -. -.
. ~.. -.
~...
7 i e i
l 1
l 1 ) second qualitative criterion, is not very well reflected in i
I 2i the way the quantitative criteria are stated, and even though it 3
may be that the numbers in the quantitative criteria were A,
l j
4I selected bearing in mind the -- the effects of other competing i
1 s
5'. technologies for producing electricity, that point was not made E
j 6
clear in the discussion, and if this is going to be done, if this i
R
)
Q 7
concept is going to be retained, that link should be made more s
5 8
clear.
I d
i c
9 On to the last quantitative criterion, there was a Y
10 belief that -- which was not unanimous by any means, but it 3
11 looked as if it were the larger of the two, that 10-4 per year l
j 3
l d
12 for core melt is too high a rate to put in the criterion.
If a i
z i
=
a 1
E 13 l rate is going to be -- if there is going to be a criterion on 1
=
j j
14 core melt alone, it probably should be less frequent than that.
i
+
i 4
m 15 '
As I say, this was not unanimous by any means, a nd there was a i
2 a
4 x
-4 l
J 16 strong view that 10 was probably quite reasonable.
It keeps z
i l
17 l the probability of core melt down to -- I am stating the other 4
3 l
}
18 view -- keeps the probability down to something acceptable over I
=
l a
i I
19 the history of water reactors, and you may be able -- you may not 5
i j
20 i be able to count on calculations of lower numbers anyhow.
}
j 21 I There was an interesting suggestion that thi 7 -- a i
i d
i
]
22 core melt criterion of this sort be tied in somehow in an i
23 implementation plan to a requirement that -- you have a require-24 [ ment like this, and if you run into trouble with this criterion, l
l 25 l this triggers a full-scale risk assessment to determine whether i
l
- l 1
a ALDERSON REPORTING COMPANY, INC.
a
___,_..,_._,_-_..,__.._,,_.__,____,_.m.
,,_.-,,-mm.
8 o
i
'l i
l 1
or not the risk itself is still in hand.
i j
2' Two things, I think, were pervasive in the discussion.
3 There was a general-feeling, I think;almost entire, anyhow, that I
4i retention of defense in depth and deterministic methods is --
i l
l g
5j really has to be included.
I mean, this is what we are all 1
g j
j 6j talking about.
And we are not talking about risk assessment as g
i
{
R 7
taking the place of anything &lse, or adding to anything else j
j 8;
in the regulatory process, but risk assecoment is an adjunct to j
4 i
2, 9j be used perhaps ir the sense that the draft document calls for.
c I
i O
i g
10 I hope I have covered everything of importance.
4 z
e r
j j
11 MR. RATHBUN:
Thank you very much, Herb. I would like 3
i j
"j 12 to turn now to Dr. Lave, who will report on the qualitative 1
3 g
13 elements of the discussion session.
=
i 14 l MR. LAVE:
We started off with vast expressions of s
i J
E 4
I 15 unhappiness with the qualitative safety goals that were in the a
a=
}
j 16 document and lots of other things, and so I am instructed to i
M l
l d
17 '
present youwwith some of our initial -- remarks.
~
l f
18 '
First of all, we wanted to emphasize something that I
~
j 3
19,
we had said in Palo Alto, which is that it is very difficult to 5
j 20 l try and pick out one aspect of this problem and look at it, that 21 j
instead one has to emphasize the overall aspects of the problem, j
22 !
and in particular what we had in mind is that it is very j
23 difficult to say something about qualitative safety goals without 1
I i
24 simultaneously talking about verification and implementation, l
i 25 I that you need a coherent statement of qualitative safety l
1 l
ALDERSON REPORTING COMPANY. INC.
l
9 l
o x
I i
1 l g als, quantitative safety goals, if possible, and implementation l.
2 all at the same time, that it doesn't make any sense to talk i
i 3,
about qualitative safety goals without talking about how they are i
i 4l implemented in the same breath, and similarly.with implementation 5l and the various goals.
e E
I n
6!
And so, you will see later on that we tried to do all e
i E
g 7
this.
1 1
l 8{
The second point is that you need different goals for n
d i
i d
9 different purposes.
We took the viewpoint as if we were 1
?-
i E
10 l advising Congress on revising the Atomic Energy statute, or i
a i
i I
11 defining a set of Commission goals.
We did not take the
<a d
12 !
perspective of what would be the goals that would be used in 1
z i
i
=
l 5
13 l licensing individual reactors, for example, and we want to E
i s
14 stress that there are very different goals that are used for 4
a
'c 15 different purposes, and you will s ee how that is going to cut 5
16 !
in just a minute.
i l
V i
j W
I i
g 17 l The third is that we want to emphasize that the j
5 i
j E
18 l documents we were given said relatively little about implementa-i 4
=
I 19 l tion, about verificationnand implementation, and Panel B thought xn i
l 20 l that that was an almost unforgiveable omission, and we are i
21 all pounding on the table at the appropriate time and saying 22 l that we hope that will not occur again.
i 23 f Having said those things now, let me go on and talk i
24,
about qualitative safety goals a bit.
The first point there is I
l 25 j the dif ference between absolute versus balancing kinds of goals.
s i
i ALDERSON REPORTING COMPANY,INC.
l 1
l 10 i
i 1
Absolute goals are ones that say that the goal is zero risk or 2j small risk or something of that sort; balancing goals are ones l
3 that talk about the good that comes from having something around 4
as well as the risks.
And we all agreed that you needed to have s
5 balancing rather than absolute goals.
That may not sound like i
n 8
6:
much, but I think that was a major point.
a
(
7 Second is that we looked at the differences between M
3 8
focusing on events versus focusing on consequences, and decided
)
d d
9 that the relevant r :ter was consequences, not events.
That ia, 3.
i E
10 it should not be the number of whatever it is you describe as i_
l 11 reactor accidents, but the number of adverse and the phrase 3
d 12 that we continually used was non-trivial somatic and genetic z
4
=
j.
13 ! effects should be the focus of what you are doing.
4
=
i j
E 14 !
And the last part of all this is, we thought that 4
i
=
l l
E 15 a good idea for the future would be an annual report by the 3
j 16 l Nuclear Regulatory Commission on the reactor accidents in the i
g 17 '
previous year, reporting to the public on what had been done as a w
I
=
18 : way of, in one sense, informing the public of what were the good
=
E 19 l and bad things that had gone on in the previous year, and secondly 5
i l
20 '
crying to relate those to their regulating practices.
21 Given those, we defined then five qualitative safety I
i i
I 22 i goals, and I will just read them to you briefly.
I might note 23 before I read all these that before we starred the discussion i
24 )i people thought qualitative safety goals listed in the document l
i 25 j were terrible, but then you can listen to these and decide how l
I d
- l ALDERSON REPORTING COMPANY. INC.
i i
l 11 l
l l
1 they strike you.
l 2i (General laughter.)
l 2i MR. LAVE:
The first one was that the goal was to 4!,
minimize non-trivial adverse somatic and genetic effects in 1
I j
g 5l supplying the nation's desire for energy.
The second possibility E
j 6l was, no significant probability of non -trivial somatic and R
7:
genetic effects from reactor
- p. cidents.
The third goal was
-n j
Sl no non-trivial adverse somatic and cenetic effects from radiation d
2, 9l released from nuclear plant accidents.
F ourth was, rarely will 0
c I
g 10 l nuclear reactor accidents cause non-trivial adverse somatic and z
i
=
i j
j 11 i genetic effects.
And the fifth was that reactor accidents will a
I f
12 l not add more than a trivial increase in the probability of non-5 I
13 i trivial adverse somatic and gcnetic effects.
w l
3 14 (General laughter.)
b 15 l MR. LAVE:
We love the phrase.
- ~
5
,i j
j 16 i MS. WEISS:
Aren't you omitting -- and marks?
i i
d 17 '
MR. LAVE:
No, that'sowhat I was getting from, that w
=
a 5
18 i rarely will nuclear reactor accidents occur.
I think I've got i
P 1
3 19 l it.
Steve Derby, I don't see where you are. Is that okay?
5 h
20ll MR. DERBY:
Why don't you finish --
i t
j 21l MR. LAVE:
Okay.
22 I MR. DERBY:
-- and if there is any reason to bring it 1
i 23 up again, I will.
l l
24j MR. LAVE:
Okay.
We also thought that for each of i
25 i these you might want to add one or more of the following ALDERSON REPORTING COMPANY. INC.
9 12 o
l i
I constraints if you are particularly worried about them, and that i
2i is not to indicate that we were not.
The first is to look at 3,
high consequence accidents and say that, for example, nothing in 4l the above should be construed to tolerate a non-trivial i
e 5i probability of a high consequence accident.
The second one was E
I n
+
3 6i the risk to particular individuals, and again, you could have R
R 7
some phrasing like nothing in the above should be constrained s
E 8 ',
to allow any non-trivial risk to any individual.
n d
d 9l And the third was to make the point that whenever there
\\
10 l is uncertainty associated with some technology, that racionality z
i
=
i 5
11 demands that there be some trade-off between your' best estimate
<3 d
12 l of consequences and the amount of uncertainty. That is, the more z~
i d
13,
uncertainty, the more you ought to lower the tolerable estinate E
l 14 ' of an accident, so that the resolution of uncertainty would be w
i I
E 15 l allowed to increase the probability of adverse occurrences.
W l
=
\\
g 16 We then turned our attention to implementation, and d
17 here I will read what Ed Zebroski drafted which, to your 5
(
5 18 ;
surprise, you will know that we all cheered at.
The hour was 5
late.
C 19 A
20 (General laughter.)
I 21li MR. LAVE:
It says, any safety goal, qualitative or l
22 quantitative, is meaningless without the context ofz.a sat of 23 implementation processes.
These are required to cover the 24 differences between a reactor operation viewed as a stat i'
i 25 { object and a scatic regulatory and technical environment versus 1,
1
- I ALDERSON REPORTING COMPANY, INC.
l
j 13 1
i I
1: the actual situation in which all of these are dynamic.
l i
2 For example, a desirable implementation should have I
4 3
the following attriisutes.
One, a rigorously applied and 4
enforced process of evaluating the relative safety importance of i
1 e
5 regulatory actions and deployment of regulatory actions and i,
Rn i
6 efforts without exception in proportion to the evaluated e
i
+
R
\\
]
g 7
importance.
I N
J B
8 Two, a defined and rigorous -- I'm sorry, a defined, a
n d
I
=
9 rigorous, and continuous process of taking account of actral 4
1 3
10 operation, -- actual experience in plant operation together with 4
E 5
11 systematic updated weighting of their safety implications, with i
5 l
d 12 ' thorough and timely feedback of these lessons learned in plant 3
=
i s
13 operation, training, and management practices, and the i
E j
E 14 associated regulatory requirements.
a l
2 15 :
Okay.
So that was our attempt at looking at wz 16 implementation, and then finally we tried to say something about
_3
,1 a
i j
d 17 l luantitative safety goals, and agreed upon the following statement, i
i 5
18 !
namely, quantitative safety goals are not useful if there are i
i I
19,
not agreed upon methods for 'alculation and verification, and at 5
i 20 l present there are not such agreed upon methods.
I l
21l We then went in to say that probabilistic risk i
t
]
22 ;
assessment is useful for priority setting and should be used 4
l 23 ' more systematically and effectively in doing so.
So, we are not 4
24 being negative about PRA.
We are saying, though, that it should i
i
]
25 i not be used fo r setting quantitative safety goals at this point, I
i s
i d
l l
l ALDERSON REPORTING COMPANY. INC.
l
i 14 i
I I.
1' but rather should be used for setting priorities for looking 2'
at various kinds of events and should be used with more power in 3
doing that, more authority than it has in the past.
4 (Pause.)
e 5
MR. RATHBUN:
Thank you, Lester.
9 3
6, Now, for the last panel chairman report, we turn to j
g i
8 7'
Paul Slovic to give us a report on the economic, ethical and M
j 8
socio-political issues panel.
d.
d 9
MR. SLOVIC:
I think there was general agreement that i
Oh 10 the present draft statement did not adequately deal with the z
I g
11 j socio-political, ethical concerns that were discussed at the f
12 l first workshop.
We went around the table and quickly generated
~
5 13 i a list of about 18 different concerns or issues that would be
=
l l
14 worthy of discussion in our three hours, and pared these down to E
E 15 about ten items, which I think we managed to cover at too a
i
=
j 16 superficial a level, really, but I will kind of go through tha
^
l p
17 gist of these.
s 18 First there was a general concern with the logic and
~
l 3"
19,
coherence in the draft statement, and an expression of dis-A 20 I appointment, and a feeling that there wasn't a strong logical 21 structure to the document.
In more specific ways, this was 22 ! discussed, for example, with regard to the treatment of, you know, 23 cost benefit notions and the ALARA criterion, and it was pointed 24 out that really in a theoretical base that there were some very l
25 important distinctions here, aui what this document entailed was l
ALDERSON REPORTING COMPANY. INC.
l
15 1
I a kind of a hybrid of the approaches, and it wasn't at all clear 2l just what the logic of this was, whether it was or was not 3
defensible, you know.
It could be a very brilliant integration 4,
on the one hand, or it could lead to some, you know, 5l inconsistencies and implications that would really look foolish, s
R I
3 6'
and perhaps this should be analyzed in a little more detail.
e R
s 7
Also with regard to the concerns a bout the logical a
8 8 -l coherence of the document, there was concern that the link n
d I
=
9j between qualitative principles and quantitative goals was not at i
1 eb 10 i all clear, that the so-called self-evident premises were not i
s 11 1 necessarily self-evident, and that there needed to be a 1
3 d
12 *;
distinction made and emphasized between what was called normative z
5 i
d 13 :
considerations and empirical or descriptive considerations.
E A
14 i normative considerations being the sort of logical, analytic cs E
15 -
considerations, and empirical sort of deriving from what people 6
s J
16 i might want or how they would react to various goals.
m f
17 We also spent -- in turning away from the logical 5
i M
18 l side, although that pervades everything here, we spent most of
=
1
}
19 the morning discussing the probabilistic risk assessment and a
20 l its implication for goal setting.
I think this was stimulated i
21l to a great extent by the talk that Bob Denero gave, and the l
l 22 question was posed, you know; in the sehse that will the house i
23 of cards built out of PRA collapse under the weight of the l
1 24 accumulated uncertainties that were described this morning. When l
l 1
1 25 i do we say that, you know, this is a method that is imperfect, t
ti a
ALDERSON REPORTING COMPANY. INC.
i 16 l
l 1
it has uncertainties, but it is still usable, or when do we say, 2;
wait a minute, these uncertainties are so great we shouldn 't --
l 3!
you know, we just have to abandon it, or you know, what are the I
i 4i implications here, and we discussed the middle ground approach, 5l which I think is evident in the present statement, that -- that g
E 6
PRA has to be used in sort of a soft way, that it can be useful l
h" 7j as an aid to structuring thinking in the process of making a s
I j
8' lot of decisions that need to be made about certain aspects of I
d 1
0; 9l the technology and management.
3 10 We also discussed the problems.of verifiability of E
E 11 I these goals, and talked about the f act that even if, you know, l
3 d
12 l you can 't demonstrate verifiability, that nevertheless the goals z
E l
d 13 i could be important because they sort of set out basic principles 5
i j
14 i that are important.
I mean, you could keep it at a quantitative 2
l 5
15 l level, the whole goal setting, qualitative level, and sort of a
i 2
f y
16,
view it as, you know, like setting a constitution and sort of A
i g
17 defining what should be important in the NRC decisions, you know, w=
t 5
18 l what should be on the agenda as relevant social and political
=
l r
I E
19 i and value considerations here, and these could be set forth as
.I.
20 !
guiding principles pretty much independently of verification.
21 f IIere I see we seem to have been on a track that was 22 very different from what Panel B was concluding, that,you just l
I 23 ll can't divorce the two.
We felt that maybe you could.
Maybe you 1
24j could set forth general principles that really did have meaning, i
i 25 i that would help auide the learninc and development process over ii i
I 1
ALDERSON REPORTING COMPANY, INC, l
17 l
l I
I time without necessarily having to go right into the 1
2l specifics of implementation.
I 3l Now, here I think one can compare it to the process 1
4 of setting a constitution for a government without knowing I
l l
g 5'
exactly what the specific forms of implementation are going ro be.
9
]
3 6;
In the afternoon, we moved away from the probability i
7 assessment side to the issue of value assessment or value i
j 8l analysis, and here we discussed the notion that just as we can i
q
=;
9 talk about PRA, probabilistic risk assessment, we can talk about z
l 0
10 value assessment, and value assessment can be viewed as every 5
{
11 l bit as complex as PRA, and every bit as relevant, if not more j
3 g
12 relevant to goal setting, because isn't that what goals are 5
\\
i j
13 l about, you know, trying to develop goals which will tell us
=
1 I
j 14 i how to protect the health and safety of the public.
j I
E 15 l Nevertheless, value assessment is in a lot more I
l j
g 16 l primitive state than PRA is, and if you think the uncertainties w
i 17,
are great with PRA, the uncertainties with, you know, the treat-a h
5 18 ment of value questions regarding equity and risk aversion,.you i
l 5
l 19 l know, weighting of genetic versus somatic effects, and so forth, A
20 !
are every bit as unsettled, if not more so, in need of further
]
21 I
- analysis, i
l i
j 22 '
This led to a discussion of the fact that the i
i i
23 quantitative goals that were presented in the current document t
24 really have implicit in them a number of assumptions about these i
1 25j value issues which are not fully spelled out.
For example, the i
i i
i ALDERSON REPORTING COMPANY. INC.
I
18 4
I lumping together of latent and early fatalities has an j
2 implication about discount rates of negative consequences.
There
{
3 are sona assumptions about risk aversion in thgre that are very j
4 strcng asstmptions.
Also, the assumption that if you make the, i
J i
e 5
you know, the goals low enough you C.on't have to worry about I
N
]
6 equity concerns, is another.
i 6
7 I think there was an agreement that really these J
3 j
j 8
sorts of questions are in need of analysis, and that there are 0
q 9
sort of two ways to go here.
One is to at least be honest or 4
i 2
h 10 straightforward about what assumptions are being made here in 1
?
11 the statement of the' quantitative goals, or else to hold off i
I B
l
{
12 on the quantification side in order -- you know, until we 1
4 l
g 13 have a more thorough analysis of what the implications of these i
z g
14 considerations are for quantitative goals.
w k
2 15 Related to this was a discussion, for example, of the l
j 16 notions of comparative assessments that are implicit in some of 1
A 1
17 '
the goals, you know, that it is okay if the risk from nuclear b
l i
M 18 l power is like 1 percent of the total accident risk that people ch 19 g
accept, and what are the implicdtions here, you know, that we j,
5 i
20 are making comparisons with -- or comparisons with regard to 21 risk from other energy technologies.
These are based on com-i 22 paring very different distributions of risk.
They don't take J
23 into account the problems of defined -- definitional problems.
24 For example, you could take this 1 percent criterion and apply I
25
.it to reactor accidents.
You also apply it to waste management
.I f
ALDERSON REPORTING COMPANY. INC.
l
- I
i 19 1-risks, and you can kind of by definitional kinds of things you l
1 2'
can aaply it to a lot of different technologies, but who is l
3 watching the cumulative total, and you know, what does this add i
4 up to?
5 g
So there's a lot of logical questions regarding the 3
6 link between qualitative principles and quantitative principles a
7 that really do involve socio-political issues.
A j
j 8
I think what this led to is a feeling that really the d
9 qualitative goals need to be strengthened, and need to emphasize 3
I h
10 various socio-political concerns, and that one ought to consider 3
h 11 whether -- just how far to go towards.quantification at this a
I 12 point.
It may be a significant contribution, just to kind of 5
a j!
j 13 define the agenda of what is relevant for NRC safety decisions,
=
1 m
i 5
14 and then leave to a further evolution and iteration to go into I
$j 15 more specific implementation concerns.
x j
j 16 That's all I have.
t
^
f d
17 I MR. RATHBUN:
Okay.
Thank you very much," Paul.
1 E
{
18 Let me see.
At this point, we are scheduled to have 1
E j
19 Plenary Session 3 -- core evaluation.
However, before we go to a
i 20 l t t.at, I sat in on the quantitative safety goal elements, and I 8
i 21 would just like to clarify, understand a little bit better the 1
1 22 other two panels.
1 i
23 '
Is the sense of it, what I hear, that quantitative j
24 goal.
ay be good, but we are really -- we are not ready for them 25 now, or at least this paper doesn't lay the basis for them?
Are ALDERSON REPORTING COMPANY. INC.
t
20 I
i I
I we saying that there could be developed a basis, and something 4
1 I
l 1
2l significantly thove and beyond what we have here is required?
3 Or is the message that we just aren't there yet?
And let me ask k
j.
4, Paul first, and then go back to Lester.
1 1
?
g 5,
MR. SLOVIC:
Well, I don't think I can speak really I
O j
6!
for the panel.
I have the feeling that our panel felt that we 4
n j
R 7
are not there yet with regard to the ability to state quantitative s
]
j 8j goals that are logically defensible.
O I
=}
9 We could look at -- step back and look at first i
3 f
10 principles, i
Ej J
11 MR. RATHBUN:
Okay, fair enough.
Lester?
k j
d 12 MR. LAVE:
We thought there were two cifficulties. The 4
m i
3 l
5 13 l one that we talked about briefly but came to no resolution on was l
E 14 the difficulty of stating such a goal, 10-5, for example.
We i
M j
2 15 just didn't get enough into that to know what those numbers meant l
5 j
16l or how it is you do them, but we were skeptical about just g
17 i setting out such a number and saying that is clearly trivial, 5
i 5
18 l since it was not clear to all of us that any number you named was
=
{
b 19 I trivial.
i s
i n
20,
But the real difficulty had to do with the fact that r
21 l at least it was our understanding that the way in which you would a
_4 j
22 i say that you met any goal like 10 for the probability of 1
23,
core melt was via PRA, and the fact that there is not agreement 24 from different groups regarding the PRA estimates you get from l
l 25 an analysis, and that being so, and there being no independent I
i ALDERSON REPORTING COMPANY. INC.
s
..r..
,m--
21 i
4.
1 4
1 I mechanism for resolving those kinds of differences, what you are
{
l 2'
then reduced to is either stating a wide range, given different I
3 kinds of PRA, and then what is your criteria, that the most i
4 conservative PRA meet this, which doesn 't seem to be terribly i
e 5
logical.
It certainly is a way of doing it, but it didn't seem 6
3 6,
to have a lot to offer.
R R
7 It just seemed to us that until you have a fairly b,
3
]
[
8 general agreement on the way in which you are going to calculate d
j c
9 whether you are meeting that goal, that it is meaningless to i
o j
g 10 state a goal, and that's part o'f what we were talking about when i
z l
=
g 11 we were talking about the necessary interconnectedness between i
d 12 the
-- and the verification-implementation procedures.
Setting 5
=l 13 out a number didn't make any sense unless you talked about m
)
{
14 precisely how you could get some agreement that that indeed was 2
15 a method which was capable of giving you an answer that people 1
x j
j 16 would believe.
1
^
d 17 MR. RATHBUN:
Okay.
This is really, of course, Herb's 5
M 18 session, but I don't know.
I think we should ask for comment on
=
H f
{
19,
this very subject, whether or not we are ready.
Why don't we i
n i
j 20 l just solicit reaction from within the group as a whole?
i 21 Yes, sir, Dave Okrent.
1 i
22 MR. OKRENT:
I would like to ask Lave and Slovic I
j 23 '
whether in their negative point of view toward the potential for I
j 24 setting quantitative goals now, they include any of the following l
]
25,
considerations.
First, the staff and the -- the NRC staff and i
ALDERSON REPORTING COMPANY. INC.
-.--wr x
y..,-
,m-1
.___-_--+-_,__,.__-,...,..,-.---.._--mm--
~,._-.--,wy
_--- ~.
22 l
1, industry, in fact, are doing risk analyses and reliability I
2 studies, they get numbers, and they base decisions in part on the l
3 numbers they are getting, so de facto there is in place the system 4
not only of analysis but of goals, because they say' this number g
5 looks okay for a long time, or this number is large enough, we 9
6 should do something.
1 R
5 7
The second is that it may be possible to set a goal i
k k
8 with the recognition that there are uncertainties, and that in i
d f
some areas you may be able to identify a quantification that d
9 i
zo j
g 10 is commensurate with your need, and in other cases know that you E
=
4 II will have a wide band of uncertainties, and say, okay, I have to a
g 12 somehow allow for it, and it was suggested, I think, by Lave that 5
a 13 j
g when uncertainties are large you compensate 6r that, and that 4
=
m 3
14 could be part of what you call the implementation.
j g
15 Given that kind of situation, do you feel really i
z g
comfortable with sort of suggesting what we need, I will put it 16 i
s j
b.
17 in tne -- a lot more research and study before we can set t
4 l
1 i
w l
3 18,
quantitative goals.
Cb I9 By the way, I am equally willing to argue on the other j
g i
5
}
20 l side that we don't have a method, you know, of satisfying this, 1
j 21 I but since you have come up with the position, I would like to 1
i 22 l see why it is a sound position in view of the state of affairs.
j 23,'
MR. LAVE:
Well, let me try it first.
I think that i
24 first of all, we have heard this morning and have heard more 1
l 25 generally that there is a lot of agreement on the relative risks, I
l ALDERSON REPORTING COMPANY, INC.
,wr-e,-
y g
e-e---
YP mw-mi-m--
---w---m
+=
e
,,r
.4
--.,--n
-u.-
-~we-.,
-m
23 1
risk estimates you get out of PRA, and not very much agreement on 2
the absolute risk estimates you get out of PRA, aid so therefore 3
unless you have some way of calibrating it, that suggests that 4
you ought to be using PRA for priority setting, not for some 5
absolute measure, and that was the nature of the recommendation a
E N
8 6,
we came out with.
e 7
More generally, the answer to the question, I think, is A
i 5
8 not one of whether quantitative risk -- quantitative safety goals n
i d
e 9
would have no use, since it is quite clear, as you said, that 1
i 10 you can use these to bound your problem.
The issue has much more j
E I
=
11 to do with the usefulness of that.
That is, if you use the most i
i 3
d 12 conservative way of doing your PRA to get an absolute risk Ea j
j d
13 !
estimate, that everybody would agree is certainly no smaller than 5
l E
d 14l'
-- sorry, no greater than what you want, then the question is, is u
E 15 I that going to add anything meaningful, and I think that without E
16 doing a lot of discussing of that, we thought that that number 3
3 p
17 )
was likely to be sufficiently large, larger than your best l
w h
13 estimate that it wouldn't add materially at this point.
i 5
19 i Now, there were some people on the panel who didn't 2
j n
j 20l want to use quantitative safety goals at all, but we didn't have i
j 21 i enough time to discuss that in general.
That certainly is not my l
4 22 i personal feeling. I would like to get quantitative safety goals i
as soon as possible, since I think there is a discipline in 23 i
24 there that I think is welcome.
h I
25 MR. SLOVIC:
I think there was a recognition in our ALDERSON REPORTING COMPANY. INC.
i I
I 24 i
l 1 !
panel that quantitative goals can have an important use, that PRA l
2' and quantitative goals can be important in specific decisions, and 3
these decisions have to be made, and you have got implicit in those 4'
decisions one way or another some criteria, and it would be good e
5 to spell this out.
N 6l I think there was concern that the logic behind the R
l 7i goals as stated in this document was not fully spelled out, that s
j 8,
these gehls have not been analyzed as carefully as they should be d
i c[
9 with regard to the implicit assumptions that underly them, and
?
10 that this needs to be done, and whether or not such analysis 3
h 11 '
would lead to rough satisfaction with them or not, I don't know, l
3 Y
12 l but there'was concern 4 at least, you know, the assumption should 4
I g
13 !
be made explicit, there should be more careful analysis of them,
=
z 14 and then take it from there.
5 5
{
15 MR. JOKSIMOVIC:
Are you both saying that you are
=
g 16,
f aulting the document rather than the sa fety goals and the PRA, l
i 17 i or what?
a E
5 18 MR. COCHRAN:
I would fault the people that use PRA.
=
H 19 ;
I think it strikes a lot of similarity with my recollection of M
i 20 l people that played war games using large computer models, where 21l the abuses of the discipline were far more prevalent than any l
22 '
use that came out of it, and that one perhaps should set some 23 safety goals to ensure that the discipline of probabilistic risk 24j assessment is kept in check, and not abused, and perhaps we
{
25 l need some coals that sort of define the limitations of where this !
i d
ll ALDERSON REPORTING COMPANY. INC.
I
25 1,
particular modeling approach can be applied, and where it cannot I.
2 be applied, and get some sort of discipline in the process.
I 3
see the whole undercurrent of the efforts in this direction of t
i 4l
-- as being to try to eliminate some things about reactor safety 5li I
that I find essential, and that is more reliance on defense in j
g E
i I
j 61 depth to mitigate the consequences of accidents rather than less.
1 7
MR. SLOVIC:
Just to elaborate, we made the distinction
~
j 8
between goals and standards, and.we were concerned about, you 4
d 1
n 9
know, these quantitative targets used as general goals, although 2
c l
10 I think that from qualitative principles one can derive, you 3
{
j 11 know, standards to use at the moment for the decisions that have i
s' l
j 12 to be made at the moment, i
5 d
13 l MR. O'DONNELL:
I was on Panel C, and I think I feel, i
E l
l 14 I Paul, that you somewhat negatively stated or summarized our l
j 2
15 l discussion on the value of the quantitative goals, and I will w=
y 16 ask the other panelists to make their comments, but I thought the e
i l
d 17 l tenor of the discussion was that it is very difficult to set these 4
l
}
18 l l
goals, but that it is worthwhile attempting to do so, and that C
{
19 one should do it at least on an interim basis, and to set forth 1
.e j
20 both qualitative and quantitative goals that can be used in the j
i i
21 development of looking at PRA from -- looking at safety from a 1
22 PRA standpoint.
l 23 '
I think the criticism of the report was more in its 24 l structuring and lack of apparent underlying rationale for the J
l 25 qualitative and quantitative goals that were in there, not ALDERSON REPORTING COMPANY. INC.
-- ~.._ -.
l 26 i
)
j 1
necessarily that they were deficient.
1 1
2 I With respect to the linkage between the quantitative i
1
{
3 goals'and PRA, I think I would say this, that one can certainly
)
4
-- I can envision a regime where you would set que'titative i
g 5
safety goals, and these could be set independently of the E
l 3
6l implementation.
That is, these would be standards or expressions l
R 5
7 of acceptable risk.
Whether you can verify them or not is a u
j g
8 separate question.
One can use PRA, I think, cautiously to do d
9 that, so I could envision a regime where you would have 1
2 O
i 10 quantitative safety goals, and you may or may not have a methdd 3
j 11 of verifying them, but --
i 3
l tj 12 MR. COCIIRAN:
W uld you give one example where you could o
5 13 l
~
l verify one of the goals that has been set forth in the last
=
14 'i J
m j
session with PRA?
i 2
15 MR. O'DONNELL:
Well, I think, depending on how you m
j 16 !
treat the uncertainties, you may or may not have confidence in A
r I
j Q,
17 t the results of the PRA, but the point is this, that I can't w
envision a regime where you are using PRA but you don't have i
3 18 1'
s 19 a
safety goals, and that is what we are in.
Essentially we are 5
i j
20 !
using PRA now in the Zion and Indian Point cases where these 21 things are being considered or being ordered by the NRC staff l
22 and are going to be litigated in a hearing, and we have no way j
23 '
of determining whether the results of those analyses are 4!
I 24l acceptable or not, so I cannot envision -- if we are not going to I
25 have safety goals, quantitative safety goals, I don't see how we i
i i
l ALDERSON REPORTING COMPANY. INC.
1 t
r,,
-,,-w.--y,----------r-y m.-
,-w
27 1
can use PRA.
l l
2 MS. WEISS:
Is the way to address the fact you may be 1
3 using PRA prematurely to set safety goals prematurely or to say
)
4 we are using PRA prematurely and we ought not to be?
e 5
MR. O'DONNELL:
Well, I think the safety goals should h
i j
6 be set independently of the PRA.
They should be set from first R
l g
7 principles that are logical, rational, and consistent, and how k
3 l
8 you verify them, you may never be able to verify them.
i d
]
c 9
MS. WEISS:
I can't envision a regime where you l
N j
10 set a quantitative safety goal knowing in advance that th'ere is E
j 7
11 very little likelihood that you can agree on whether it has been R
I d
12 met or not.
It is an improvement over what we've got now.
$=
=
13 i MR. OKRENT:
The EPA sets safety goals of a sort when
]
E
{
]~
14 they review some pesticide in water.
They do a calculation which l
2 15 is subject to considerable uncertainty about the lifetime risk N
i j
16l of cancer at certain concentrations in the water, and then they l
17 '
go ahead and set a concentration on that basis, pretty much, if W
=
M 18 you read what is done.
=
C 9
19,
MS. WEISS:
Do you think that PRA, which is the only 1
5 l
20 l way that we know about now to verify -- to determine whether you i
i 21 need or don't need a quantitative goal, whether that can be done 22 with sufficient precision so that, A,
the technical community 23 can agree that the result::is a reliable one, and B, that the 24 l public can agree?
25]
MR. OKRENT:
I think that if the approach is such li d
0
- i ALDERSON REPORTING COMPANY. INC.
t
,----,--r--
l 28 i
I i
l 1:
that you don't require too much in a single step, for example, in i
i 1
2i prevention alone, that it would be possible to prepare analyses l
3l by these methods where the mean value meets the numbers of the i,
I I
4I sort we are talking about, and where a group of people who are, j
s 5
let's say, not politically or emotionally biased on the thing, 8
f
{
6 either way, might agree that you have or have not met it, with R
j d
7l some, you know, degree of confidence.
a j
g 8,
MR. COCHRAN:
They wouldn 't agree tha t the mean value e
i CL 9
has any meaning though, any -- that fou have any real confidence 1
z I
Og 10 that the mean value has any real significance.
z j
j 11 MR. OKRENT:
I don ' t know.
You have to trie it out.
n
{
12 l j
But --
i
=
\\
13 !
MR. COCHRAN: We tried it out in WASH 1400.
=
l l
14 l MR. OKRENT:
Well, again, I think it is conceivable l
1
)
2 15l that if the targets are not too small, that a group of independent
>l g
16 j experts might feel that after allowing for the wide uncertainties d
I 4
i b.
17 there is sufficient margin that they feel comfortable in saying aC 3
18 you have met these levels.
This is speculation on my part, but 4
5 E
19 ! I am trying to answer the question as directly and honestly as n
1
{
20 I can, and in fact in part that was a factor in setting the 21 numbers that were put in NUREG-0739.
We tried to put numbers in i
6 22 I that we thought it would not be impossible to at least discuss 1
23 ] in exactly that framework.
24 :
MR. RATHBUN:
Ed has a comment.
I 1
25j MR. ZEBROSKI:
Dave, I think we had a preamble to that A
i ALDERSON REPORTING COMPANY. INC.
l 29 I,
Il discussion which hangs on one word in the discussion paper, that I
2' the safety goal may add to consistency and coherency in 3
regulation.
If you read that as Mr. Rathbun did in the opening 4
statement, that a safety goal should: contribute to coherency and g
5 consistency of the regulation, then it follows as the night the 8
1 3
6!
day that you must have an agreed on verfiication process, and R
7 absent that it will add chaos instead of consistency.
s j
8!
MR. OKRENT:
Well, I agree it could add chaos.
I have e
[
9 to agree.
And it could also be abused by all sides.
z O
y 10 l MR. ZEBROSKI:
But the thing that you said was in place 3
l
=
i 11 l is the thing that I think there was wide agreement on, that the B
1 l
12 l use as a prioritization technique, which is what NASA has used 5
I g
13 i for 20 years to make things work better, is here and now and
=
i n
5 14 should be used and given credit.
]r 15 '
MR. OKRENT:
What I was saying is, it is being used
=
y 16 l for licensing kinds of decisions, not for prioritization, but w
I d
17 !
how important is something that would have to be fixed, that is x
l
=
5 18 '
quite different, and I would just like to note that in fact you
=
w h
19 ;
can use PRA and arrive at prioritization decisions which are 5
20 !
wrong.
In other words, pressure vessel integrity is a good 1
21 i example, the valuations where this was a very low probability 22 situation, but of course they always had the caveat, except for 23 transients, which are not part of the ASME code, and of course 24 ) when you look at transients you have to become interested.
i 25 j MR. ZEBROSKI:
But even there, the judgment is that ii I
)
ALDERSON REPORTING COMPANY. INC.
{
l 30 l
I I
j {
the problem hits you in about ten years, and so the. instant l
2j reaction issue is what you are talking about.
That is a prioritization' matter.
3l l
4 MR. OKRENT:
I guess if I were going to advance a 5
p sition, it would be that it would be useful to develop e
5 I
i 8
6 qualitative and quantitative safety goals,it would be useful to e
7 develop as good a way of doing PRA as we know how, with all the 8
kinds of checks and safeguards that would almost make Tom N
Cochran happy, and then having done this, don 't use it mi the 9
i h
10 sole input but as an input.
3 5
gj What bothers me is right now it is being used d
12 piecemeal without the kind of peer review you want.
There are E=d 13 de facto criteria around, et cetera, and so I would rather see 5
l D
34 the situation be improved.
15 MR. ZEBROSKI:
You know, we have been thinking cbout 5
16 ' there are several committees now working on this, the verification 3
A 6
17 of PRA, and there are just two ways you can go.
You can either 5
5 18 go as a process attribute, process verification, or you can go as
=
k 19 : product verification.
Product verification means that you would 5
20 l repeat each one of 20,000 or 30,000 calculations involving ten 21 or fifteen man years.
The process verification means you write 22 an incredibly detailed process and hope that everybody 23 executes it just right, also not feasible, i
24 l So I think the business of getting the PRA as a l
l 25 licensing decision process, it will work well for easy problems, i
b ll ALDERSON REPORTING COMPANY. INC.
l
0 31 l
I that is, where something is real bad, it will tell you that, the 1
i 2l outliers, and if something is real good, it will tell you that, 3
but what we are talking about is the great vast majority of gray 4l problems where a quantitative disagreement of a factor of ten is I
1 g
SI fatal on verification, md if somebcdy sees a way out of that, 0
l j
6i you know, I am a practitioner of PRA myself, and a supporter of "Q
i 7
the art, and I have a good part of my staff working at it,but
- j 8
I don't see a way out of that dilemma. I wish I could.
O d
9l MR. COCHRAN:
Let me ask you something there.
Two out I
I E
10 of four breeders have had core melts.
Would you argue that that E
l 5
11l actuarial experience is su;2icient to eliminate that technology 2
I d
12 '
on the basis of frequency of core melt goal that we have here?
E=
d 13 Or would you say that that doesn't apply, that we should do some E
j 14 '
sort of PRA analysis of some future design,and that that is much 2
15 j better than this actuarial experience?
w I
I J
16 i MR. ZEBROSKI:
There is an interesting paper on this E
g 17 i subject in the February issue of Nuclear Technology, and the 5
i 18 !
issue is that in both of these cases, EBR 1 melt, the reactivity
=
I
{
19 l contribution from Rod Bowing, and Firmy 1 local fuel melting, 5
20l the small, the ability of a small piece of metal to block several 21 i channels, that that -- the lesson learned there from a desien and I
22 l operating standpoint was monumental, and has been incorporated in 23 '
all future designs.
24 So those particular statistics, you can say that kind l
25 l of population for those particular problems doesn't exist, so I
ALDERSON REPORTING COMPANY. INC.
..=.
32 there is a learning process that has applied, but yes, I happen 1
\\
2l to think that the learning process statistics has to take into i
4 3
account those events.
However, you get a rather favorable answer 1
l 4
if you do make that analysis.
The mean time between major I
g 5
overhauls or major failures of equipment, as you know, has some E
j j
6, relationship to the total experience that you have had if you use i
R j
2 7
the experience diligently, and that is why our panel discussed a
{.
8, lot the need for defining the implementation environment if the i
l d
l c
9 safety goal is to have meaning.
i C
g 10 MR. PERROW:
Ed, are you saying that it was a i
3 1
5 11 monumental learning in the case of Firmy to know that you should
<3 i
d 12 put strong bolts on a piece of metal that goes in after the z
E I
d 13 I design is completed, and that you should mark it on the prints 1
E i
14 so that when something goes wrong -- I niean, that is the kind y
6 15 of monumental learning that takes place?
i
)
16 l MR. ZEBROSKI:
No, there's a more fundamental one.
j w
g 17 ; The inlet of the fuel assembly was designed so that a relatively t
w E{
18 small flat piece of metal would block it.
All fuel assemblies i
19 now are designed so that that is not possible.
They have a l
M l
j 20 l multichambered thing that you could -- if you tried to block it 4
i l
21 with many pieces of metal, you would have a heck of a time i
l 22 accomplishing that fact, so tilat if you go to the conclusion that i
)
23 t you should use strong bolts, you wouldn 't have learned very much, e
24 because there are many other sources of loose pieces of metal that i
I I
6 25 ' could block it, so you have learned a more fundamental lesson, I
i ALDERSON REPORTING COMPANY,INC.
I i
1
33 l
1 i
1!
you design the fuel bundle so that no small piece of metal can l
2l block it, in fact, many small pieces of metal can'.t block it.
1 3
MR. PERROW:
That still surprises me as a lesson that 4
had toibe learned.
1 y
5 MR. ZEBROSKI:
More seriously, it was a lack of S
j j
6 prioritization, if you will, because it was a patch insisted on i
R l
7 the last minute in the design without due -- without time for due 3
l 8
consideration, and the operator under the -- victimized by the 4
i d
- [
9 threat of a long shutdown caved it and put it in against his
?
i 10 better judgment, hastily, and as it turns out not very well.
J h
11 l MR. PERROW:
But the major lesson -- I don't build 1
3 g
12 these things, fortunately, but I am surprised that people didn't i
g 13 know that about blocking fuel bundles before they built such an 4
i m
i g
14 enormously dangerous and expensive engine.
If that is the kind l
{
15 of learning we've got to depend on in the nuclear industry, I am y
16,
more worried than I was when I came here.
i l
d 17 i MR. ZEBROSKI:
Well, let me comment on that in a very 5
l
{
18 '
simple way.
When you look at the accidents which have what I
19 l people have called unanticipated design problems, you come back to a
i 5
20 an almost -- the unanticipated things are rarely mystical.
Some-21 body somewhere thought about them, identified them, but they did 22 not enter the perception or actions of the designer and operator 23 '
sufficiently.
In other words, it was a communication and j
24 l cultural problem.
Somebody could have designed it right.
25 Probably many people could have designed it right.
But under the i
I ALDERSON REPORTING COMPANY. INC.
j l
34 I
1 !
pressure of the particular circumstances, that designer and l
1 j
2 operator didn't have it communicated to him, and we see many, i
l 3
many instances of that now, in analyzing the problems that plants l
l 4
are having currently.
Nearly all of them have the characteristic 5
that some guy is having trouble with this thing, and five or ten
]
E 3
6 sister plants don't have that trouble.
He didn't get the message R
M 7
of how to stay out of that trouble.
A j
8 So the solution is not very much new science or new 1
0 d
9 engineering science.
The solution is better communication, 4
Y
]
10 better adaptation of standards for the things which you know 3
I l
11 how to do right, and bet *c enforcement that people have j
8 g
12 implemented when they put them in, l
5
{
j 13 So I see it as a cultural -- the aircraft industry has
=
l 14 precisely the same problem, and so does transportation generally, E
J 2
15 l for example, or heavy technology in general.
I S
4 l
16 MR. LAVE:
Let me get back to this in a second.
One j
j d
l 3
6 17l of the things that the Food and Drug Administration has done i
w 1
=
'j 18 faced with the Delaney clause is to contradict the direct 5
j
{
19,
wording of it, which says that no substance found to be a n
]
20 carcinogen may be added to food, and the NRC in trying to flnd 1
i j
21 ways to -- the FDA, in trying to find ways of getting around that,
-6 22 decided that a probability of 10 was " negligible".
And the 1
i 23 l statute doesn't say no negligible probability.
They said that l
24 l if it's negligible, then the Delaney clause doesn't apply, and 25 j have gone through some calculations, and have actually been --
i 1
i ALDERSON REPORTING COMPANY. INC.
4 m.
35 1
and have litigated that out and at least one court has upheld i
I f
2 that 10-6 is " negligible" and therefore could be taken as not L
3 contradicting the Delaney clause.
4 That is an instance where a quantitative risk assessment 4
g which I think everybody would agree is at least 1,000 times more 5
a 3
6 conservative than reality, given the techniques for making it, R
R 7
led to an outcome which I think was clearly socially beneficial.
M I
q j
8 And so I guess I have to back off at least my concurrence with d
9 Panel B 's report a bit, that is, where you are faced with an z
l c
h 10 absolute statute, a non-balancing qualitative goal, then z
c j
11 quantitative risk assessment can be very helpful in a small number j
3 I
y 12 of instances, and you shouldn't generalize that, since you have 13 [somethingwhoserisklevelwassurelylessthan10-9 in order to 3=
1 3
14 '
pass this test.
m I
I w
i u
2 15 The other one I auld like to say about it is that if a
i j
16 l one doesn't take the quantitative risk assessment too seriously, d
l I
d 17 i then I think it has been helpful in many other kinds of decisions w
i l
E 18 by EPA and FDA and so on, by using various highly conservative U
{
19 i techniques, and may I add that inevitably once you start the
\\
n 20 l process, whenever there is uncertain _ty the agency resolves it, i
I t
t 21 I that uncertainty by choosing the most conservative assumption that 22 l they can reasonably make.
I 23 '
So that what happens is that by calculating some i
24 i quantitative risk estimates, there is some discipline that is i
I 25 ' brought into the process, and I think in general it is salutary, ALDERSON REPORTING COMPANY,INC.
l i
, ~ _
.y4..
-. v
.c.y-
36
{
i 1 !
but what those agencies have not tried to do with this exception I
-6
)
2l of defining 10 as negligible, what they have not tried to do is i
l 3
to define any numerical goals, and I think that it has been e
4 helpful, without having to have numerical goals, by putting in e
5 the quantitative risk estimates, and so I guess after that sort
?
j 6;
of meandering discourse for a while, I would say that since the i
R i
)
7 NRC statute is not an absolute one of striving for zero risk, and s
j 8
since none of the qualitatitve goals that we proposed were d
i d
9 absolute ones, then that first instance has not been a -- I mean i
4 c
10 quantitative safety goals are not going to help very much in 3
l j
11 getting around that.
1 mj 12 The second one is, think that there are good reasons i
2 13 for continuing to go ahead _and using PRA to calculate quantitative l
=
mg 14 safety levels, and to use those in settirg goals and in various 2
15 anlyses, but I guess I'am unconvinced that there is anything to w
=
j 16 be gained at the moment given the levels of uncertainty in A
d 17 trying to set some numbers, like the probability of core melt, a
=
h 18 l and then using PRA to see if you comply with them.
I P
i 3
19 '
MR. JOKSIMOVIC: Why did you say the last sentence?
I a
l 20 l MR. LAVE:
I beg your pardon?
21 MR. JOKSIMOVIC:
Would you expand on why did you make 22 '
the last statement?
Why is it not useful to assess the 23 l
probability of core melt?
24 MR. LAVE:
I think it is useful to do it.-
I guess that i
i 25 i what I don't find useful is, again, judging from the techn ques J
t ALDERSON REPORTING COMPANY. INC.
t
~
~
~
37 l
1 that are used for calculating the risks of exposure to 2,
carcinogens, at very low levels, the technique used by the
]
3 agencies are caywhere from 1,000 times to 100,000 times --
1 4
MR. JOKSIMOVIC:
Yes, but that reflects uncertainties I
i e
5 in our knowledge.
Our knowledge about carcinogens is very j
E I
.l 3
6 poor.
o I
{
7 MR. LAVE:
Yes, and I was getting the impression that 1
,~
N l
j 8
PRA was going to give you a fairish range depending on the I
d l
d 9
various techniques that you used to calculate it.
j 5
I E
10 MR. JOKSIMOVIC:
The PRA simply reflects the state of E
E 11 knowledge.
If the state of knowledge is poor,.then certainly d
12,
they are large.
If the state of knowledge is pretty well known Z
l 1
5 j
'd 13 then they are certainly small.
Suppose that.
You don't fault 5
S 14 the technique.
You fault the lack of knowledge.
U j
15 MR. COCHRAN: The fault of the application of the i
5 j
16,
technique.
Let me see if I can rephrase it.
You described the 3
i 4
l
)
l 17 ;
way the FDA managed to j ump through a hoop because they couldn ' t 5
i 5
18 meet the law.
In fact, that's exactly what the NRC is doing now.
i
=H
[
19 They don't have a quantitative goal in the Atomic Energy Act, I
i 20 l but they have all sorts of regulations further down the line.
I f'
21l Then you come to a plant that you show that doesn't meet the I
l 22 i regulations.
And so they have tried to figure out because they I
l 23,
don't shut plants down and don't deny licenses a new way to i
24 :
jump through the hoop.
l I
\\
25 And so, right now, there's some techniques of 1
I e
ALDERSON REPORTING COMPANY, INC.
1
[
38 1
jumping through the hoop and giving them the license anyway, even 1
l 2'
though they don't meet the regulations, and those techniques i
3 are, well, it doesn't matter for a while, we can still operate 4
this plant without meeting the regulations because nothing is s
5 going to happen until we study this, while we study the problem.
9 3
6 PRA is, a newer and better way to jump through the hoop 4
R R
7 when you don't meet the re.gulations.
You don't meet the s
j 8
regulations on. cantainment, so you say, well, we have done a e
d 9
probabilistic risk assessment that says you didn't really need i
I 10 the containment anyway, so we're through the hoop.
Then you come
_E g
11 to emergency planning.
You haven't got the emergency planning 3
g 12 program in line.
Well, we won't -- you know, if you don't have s
i y
13 !
it for a few weeks, it doesn't matter, because we've got a i
l 14 containment.
We've just done a prcbabilisticarisk assessment.
ue 2
15 And so, really what we are arguing over is which way j
16 do you want the Commission to jump through the hoop?
Do you want
~d i
I d
17 l to give'them some new techniques, or do you want them to jump E
18 through the way they have been jumping.
=
+
{
19,
MR. LAVE:
I guess in answering that, to go back to the a
i 20l FDA decision, the FDA is peopled with a lot of advocates who would i
21 l like to protect the public health as stringently as possible, yet 22 l you can push those people and get them to admit that there are i
23 certain things that they think are silly, where the law wants them 24j to make a decision that they don't want to make, and so without t
e 25 j generalizing to this, I think that there are times when you get 1
i il ALDERtON REPORTING COMPANY. INC.
l
l 39 1.
hung up on your own statute or goals in a way that you think is not 1
2l in the public interest, and in the case of the NRC, it could well 3
be that the NRC has set up some goals that are not in fact in the i
1 4
public interest, and at that point anybody should te searching o
5 around for --
1 M
d" l
6, MR. COCHRAN:
I really don't think that's the correct a
E 7
analysis.
I mean, it's --
8 MR. ZEBROSKI:
Tom, I think you are maligning the n
d d
9 Commissioners, in a really unjustifiable way.
They have been i
h 10 presented a lot of evidence that they have caused something like 3
ij
$30 billion of delay costs and increasing that at a billion a l
d 12 year, and they said, so be it, that is not our job, our job is E
i
^
c l
d 13 I to keep it safe whatever the consequences.
So I think your
}
=
i y
14 maligning statements are really very uncalled for, and maybe Us-E 15 Peter Bradford will have an opinion on it tomorrow.
16 MR. COCHRAN:
Well, I will stick with them.
3 A
MR. RATHBUN: Ed?
g 37 l
j MR. O 'DONNELL:
Lester, I had a little.prbblem following 18 l
(
19 your logic here.
You seem to be saying that FDA has a zero risk x
e n
l 20 l statute, but they have in fact not done that, and they have 1
l 21 adopted a 10-6 quantitative goal, and that was -- seemed to be an l
4 22 l acceptable way of doing it, but since the NRC statute was non-23 zero, that is, it was some finite number, or finite risk above 24 i zero, that therefore they would not be in a position to adopt 25 a quantitative goal.
ALDERSON REPORTING COMPANY. INC.
40 I !
It would seem to me that if FDA has demonstrated or has l
2 adopted a number like 10-6 as a means of satisfying a zero risk 1
3 standard, why would not NRC be able to adopt a similar number as i
1 4
showing that they made their risk standare?
e 5
MR. LAVE:
It is not able to.
It had to do with h
j 6
usefulness.
That is --
R j
{
7 MR. O'DONNELL:
Well, all right, we are talking aboat N
l 8
usefulness here, though.
What is useful to NRC.
O c
9 MR. LAVE:
Yes, but I think the point I was trying to i
l o
j g
10 make is that the experience of other agencies is that where 3l 11 there is uncertainty, that.1the regulators inevitably because of 3
y 12 challenges in court and so on, will gravitate towards the most
=
f f
13 conservative assumptions, and it seemed to me in this case, if a
l 14 you have had some reasonable qualitative goals, you might get a x
{
2 15 better outcome than if you set quantitative goals and then had w
8 j
j 16 the regulators striving for the most conservative assumption, 1
w d
17.
then seeing whether it all met up.
w l
{
18 l 2
s So it was a judgment about usefulnest, certainly not A
l
{
19 a judgment about statutory capability, which I r:an' t make.
20 MR. COCHRAN:
I would like to read from a letter from
?
I j
21 John Kearney of Edison Electric Institute, Chairman of the NRC, 22 dated June 9th of this year.
This is concerning egulations for l
l l
23 ;
radioactive waste disposal that were put out for --prior to these 1
l 24 l regulations being put out for comment.
Let me read it.
j l
)
25q "We recommend that the draft regulation, 10 CFR Part 60,l 1
1 J
ALDERSON REPORTING COMPANY, INC.
.n
'l j
l 1! not be issued oven for comment in its present form.
Our concern i
)
2' over this regulation is.that there are potentially serious 3
licensing pitfalls in setting quantitative numerica l objectives 4
which cannot be proven, such as the 1,000-year package ar.J the
-Sth e
5 10 release criteria."
Ae 3
6!
This is a case where the industry doesn't want R
7 quantitative criteria that cannot be proven,. whereas in the A
i j
8, reactor safety case they want quantitative criteria where they d=
9 can't be proven because they think they have a new trick to 4
Y 10 ttilize them.
Thay will use the mean value, and won't require j
j 11 proof.
Yet in the waste case, it is sort of fear of the 3
j j
12 j unknown, I think, fear that it is a different group of people I
5 j
j y
13 l that haven't been in this business of handing out licenses whether I
14 l you meet the regulations or not, and so they are a little less E
i 2
15. reluctant to have the same sort of philosophy used over there.
I E
l j
j 16 l Exactly the.same group of people.
I I
d 17 l MR. KOUTS:
I don't think that's really quite that w
i l
5 18 l parallel, Tom.
I would hate to get involved in the disposal of 5
i E
I l
19 l waste here, which is an entirely differ at subject, but in that g
5 l
20 l particular case 7 think the indue'
.jsction was to the
- 21l, physical impossibility of demonstrating that you can preserve t
I 4
22 waste without its entering the Siosphere for that period of time.
l 23 They were not opposed to applying models to do this, but they 24 were opposed to the idea --
i 25 MR. COCHRAN:
They were opposed to setting criteria i
ALDERSON REPORTING COMPANY,INC.
l
j 42 j
1 ! where you cannot demonstrate that you are meeting them.
i l
2 MR. KOUTS: But the demonstration they talk about was i
d 3
the physical demonstration, not an analytical demonstration.
4 MR. COCHRAN:
I disagree.
l g
5 MR. KOUTS:
Are we getting anywhere with this?
N
{
6!
MR. RATHBUN:
I am not sure.
Herb has some follow up R
7 questions that he wanted to pose to the plenary session group i
j j
8 here, and I would just like to turn it over to him and let him d
4 d
9 go through those.
I N
I 10 MR. KOUTS:
In nur panel we never did really get to the E
j j
11 subject of method of implementation, and there is a section of the i
S y
12 draft document which deals with implementation, deals with the 1
i E
I j
g 13 l soft method of implementation, I think, as it was termed this
=
j j
j 14 l morning.
And I wonder if we can 't have some comments on this E
i 2
15 '
general subject of the proposed method by which the criteria are 4
w=
g 16 to be applied.
I d
17 !
I assume you -- well, let me start it off L.
.aying a=
18 that if I have no comments, I assume everyone agreer that the 1
=
H 3
19 soft method is the one to be appl 4, yes?
b 5
1 20 MR. PERROW:
The overall comment is I don't see what 21 is going to be implied in here because it is so general and j
i 22 so vague, but a more pertinent comment is that apparently it is 23 '
not going to be applied even if there were specific -- because I
24l for example if the calculated risk estimates when compared with 25 the quantitative safety goals indicate an unacceptable risk, l
i 4
ALDERSON REPORTING COMPANY, INC.
i
43 J
1 it should be necessary to show that other factors override this i
2 finding.
Now, the implication there is clear that if you have 3
an unacceptable risk, you can make it acceptable with other
)
4 factors, soithat I really find it difficult to see how anything l
=
5 could be implemented very sincerely if all you do is look for I
E l
3 6
other factors which will override the findings, and in case I o
e7 i
A 7
missed that sentence, there is the other sentence, well, there I
i j
j 8
are several sentences in this section.
i d
d 9
MR. ZEBROSKI:
What page are you one?
~2 i
e
]
10 MR. PERROW:
That was on Page 27, near the top.
The i
3 j
E 11 bottom of -- or the middle of 26.
NRC should not require j
3 12 demonstration and compliance with the goals in individual 1
=j 13 I licensing cases as a basis for licensing.
There are other factors.
z l
E 14 Nor indeed should demonstration of necessity and adequacy of the w
1 2
15 l safety rules or generic safety standards -- quantitative safety 4
a 1
x l
l y
16 goals be required.
l l
17 '
It seems to me it is a basically weakened kind of 18ll i
5 l
E implementation, but the other thing, to go back to the main P
lo l thing, I don't see any steps spelled out here on how this is j
5 j
20 ;
going to be implemented.
I don't know what is going to be done.
i i
j 21 How are you going to go into plants, are you going into-plants, i
22 l and saying, you do not meet this goal?
How are you going to 1
i, 23 arrive at it?
24l It goes back to the question of whether you can l
25 j really sensibly use the probabilistic risk assessment -- that is
/
i 1
i ALDERSON REPORTING C OMPANY, INC.
l
?
~
44 I
1 !
going to be a problem in itself, so that may be why you have a l
2 !
problem in being spe :ific, but implementation and verification 3,
certainly don't satisfy me.
And this came up, I would point out, j
4l at the last meeting in April, and both in -- I think in two 1
l g
5 sessions, ours and one other, it was raised, in theusummary of S
j 6
the plenary session, in the summaries made in the plenary l
R 7
sessions, and it came up again in the plenary session, too, that 4
s 8l verification and implementation was not spelled out in the I
d 9
draft documents we had before us.
f ieg 10 So that was -- I think there was a fairly strong 3
h 11 indication that some people in this group were quite concerned a
j 12 about this, and so when I got this version, I was very 5
y 13 l disappointed.
t
.=
i m
i j
g 14 !
MR. MAZUR:
I. think there is a more general issue b
E 15 even than concern with this specific statement, and that is, I l
M i
g 16 feel to the extent that a goal statement is published, whether
'A i
b^
17 I or not it has implementation procedures in it, and whatever they W
i I
18 I might look like, it does tend to convey t he inference that i
P l
l 19 l nuclear power reactors are indeed meeting these goals, and by i
M l
l 20 l virtue of what we have been discussing here, I think it is l
21 very important to point out, quite apart from the specifics of 22 the implementation, a kind of a broad preamble to it which says i
I j
23 ' right out in front, for public display, that the goals are one i
24 I thinc, but that at present there is no way to ensure that 4
25 ] nuclear reactors are in fact meeting them.
I I
i l
ALDERSON REPORTING COMPANY. INC.
I
45 l
I I think this is a crucial feature of public under-I j
2 standing with regard to the implementation notion.
1 J
3 MR. ZEBROSKI:
Isn't that the thrust of the Lewis 4l report, though?
I mean, it made th3t point extremely i
5 explicitly and that was entered into the Commission's record, g
a j
6l and the executive summary was criticized for being over-g 7
simplified.
I think we haye been all through that exercise.
A j
8 MR. MAZUR:
I would surely want it as part of any 0
i d
9 goal statement, though, to make it quite clear to those who i
c 10 encounter it that there is a severe dysj unction between the fact 3
j 11 that goals have been stated and the inf erence that they are n
[
j 12 being met with any assurance.
,=
4 13 MR. KOUTS:
I think you are overstating it, though,
=
14 Alan, aren't you?
3 i
i 2
15 MR. MAZUR:
I don't think so, but I will accept 4
j y
16 MR. KOUTS:
Ed, do you want to respond to:that?
i I
l 17 MR. ZEBROSKI:
I think he is right.
I think if you --
2 18 ll' 1
=
5 you know, the Commission, both the Commit;1oners and the staf f,
=
4 s
j 19,
as near as I can tell, are people who are dedicated to show that i
n 20l they are regulating in a responsible fashion to keep things i
i 21 safe.
I think they have -- the: int 0itive safety goal is that 4
22 l you should be at least as good as Surry appeared to be in I
l 23 '
WASH 1400.
Ahearn stated that informally a couple of times when 4
24 i be was chairman, and that was the rationale of accepting the I
j 25 preliminary reports on Zion, Indian Point, and -- They appeared i
0; 1
i
- i ALDERSON REPORTING COMPANY. INC.
\\
46.
l l
I 1 !
to be somewhat better than Surry.
So I think inevitably the i
I t
2 statement of the goal, if you don't immediately shut down a lot i
1 3
of plants, implies you have accepted it.
]
4 MR. MAZUR:
Which is incorrect.
Which is a false 5
inference, and we should take pains.
I think it is a responsi-e E9 i
6 bility here.
We should take pains to ensure that that false E
7 inference is not promoted.
1 I
j j
8, MR. ZEBROSKI:
Let's work like hell on the verification i
d j
=
9 process.
i i0
+
10 MR. MAZUR: Well, it seems to me almost in principle we a
z j
j 11 are not going to come up with a verification process short of 3
y 12 l actuarial experience that shows that the goals are not being i
5 I
{
j 13 ! met, in which case the whole thing is irrelevant anyway.
=
mg 14 MR. COCHRAN:
You can't do that.
You cannot have 9
15 actuarial experience, because every piece of 6.ata you get, there s
{
y 16 is a " lesson learned" which sets aside all historical data and I
j d
17 j makes you have to start frcm scratch again.
u a
5 18 MR. MAZUR:
Without even being as cynical as you, I 1
=
H j
19 l can still draw the same conclusion, and that is obviously if n
20. you violate the goals you've got a real problem, and the whole i
21 thing will go by the boards anyway.
h 4
j 22 i MR. COCHRAN:
You don't have to be cynical.
You can l
23 '
look at the Maguire licenses.
1 24 MR. MAZUR:
All right.
25 l' MR. RATHBUN:
Ed?
Let's let Ed --
i i
I r
1 i
I i
)
ALDERSON REPORTING COMPANY. INC.
l
I 47 i
1 I
1 MR. O ' DONNELL:
NRC has a statutory requirement to 4
1 2l make a finding of no undue risk every time it issues a license, 2
i 3
and every time it does that, it is making that judgment.
You 4
know, the current regulations are set up deterministically. We i
e 5
have defense in depth as the basis or the underpinnings of 8
l i
3 6 i.
regulation, and I think it is important, and I guess Panel A e
1 g
g 7
brought attention to that, we have to retain that, that the J
8 quantitative goals and PRA are not a substitute for defense in i
d j
=
9 depth.
They are an attempt to make the decision making process 5
)
10 more rational in terms of determining need for changing those i
?
I 5
11 regulations to make improvements, and I think it would be 3
a d
12,
totally inconsistent for the NRC to make the kind of statements 3
a
=
l E
13 you have suggested, that, well, we have these goals but we E
i j
j 14l don't know whether -- that these anals define acceptable levels l
j 15 I of risk but we don't know whether the plants that we have
=
j y
16,
licensed and already have made that determination on meet those
^
j p
17 '
goals.
l l
18 l It seems to me it has to be implicit that if the NRC
=
r w
l
}
19 is saying, this is our quantitatiw definition of what we have a
i 20 been saying all along with respect to undue risk, are these I
21 numbers, and they would have no basis for allowing plants to i
22 l continue operating if they said, well, we don't think that the i
j 23 plants in operat'on meet these goals.
Of course they have to i
l 24 i believe that.
I 25j MR. MAZUR:
That is precisely the kind of inference I I
Al.DERSON REPORTING COMPANY. INC.
j 7, - -,, -, -,, -
r--.
,c,
,-,,-a
ow I
l r
But the inference, you were inferring fear.
i MR. O 'DONNELL:
of the acceptable 2
l I that along with the statement of the goa s, h
statement that 3
levels of risk quantitatively, would have to be t e 4i oals.
we don't know whether they are meeting these g I
I think if I followed you you visualize a 5
e An MR. MAZUR:
like the risk d
6 l!
l situation where we have these quantitative goa s, f
o I
E 7
ly is one in a to an individual near the plant of dying ear
~
E 8
d j million per year --
d 9
10 f MR. O'DONNELL:
Right.
i 2
b dy
-- and the plants are operating, and any o E
MR. MAZUR:
t 5
11 '}
ht if they are would make the inference you just made, t a B
i but in fact we don't d
12 !
13 l4operating, they must have met those goals, z
3 d
know that that goal has been met.
E s
14 l Well, the NRC shouldn't.have issued I
d MR. O'DONNELL:
i u
E 15 a license if they didn't believe that.
16 That is a statutory requirement.
3 i
^
l MR. MAZUR:
17]!
E MS. WEISS:
Come on.
make the 18 Yo'u.are saying it shouldn't E
I MR. COCHRAN:-
C 19 t statement because it will expose a Ee.
x l
l and thei 20 l I think the use cf the goals l
MR. O'DONNELL:
No, hings, but it does not-,
r 21 :
PRA should be to attempt to verify those t ill licensing plants based 22 j Whether or not we have goals, we are st j
t which have been found 23 on the existing deterministic requirements, 1
- \\ \\
to be acceptable for licensing.
l l
ALDERSON REPORTING COMP ANY. INC.
f 49 1
j MR. MAZUR:
But in the end it is a guess as to 2
whether those goals are met or not.
3 MS. WEISS :
All you have done when you have signed off l on a plant is said, as of this date, this plant meets all the 4
i e
5 regulations.
That's all the Commission ever did when they made N
3 6
the finding, that there wasn't undue risk.
They never said, e
,I
[
7 certainly never said quantitatively what the combination of N
j 8
meeting the regulations got you to, and so to reach the d
inference that Mr. Mazur suggested would really be an incorrect
=
9 j
10 inference about what the NRC did when it licensed a plant.
.z.
g 11 a
i d
12 !
z t
3
- i 13 I o
S 14 l
=
i E
2 15 l M
i j
16 l l
p 17 x
ww 18 i
=
I w
i 19,!
g n
20 l 21!
i i
22,
23 24 i l
l l
4 25 j
.i i
\\
\\
lI ALDERSON REPORTING COMPANY, INC.
NBC 1
Sutt-n/rct NO faps 7 flws Jeani.
MR. MAZUR:
You understand that I r_;a not saying that the I
2' plants don't meet those goals; maybe they do, and I am just 3
saying that there is no way of-knowing that they meet those I
l 4
goals.
j e
5 MR. KOUTS:
Can you enlarge on what you mean by saying 3
6l that there is no way of knowing if they meet these goals?
U t
E 7
MR. MAZUR:
Yes, I have a nuclear power plant here in the 3
j 8
middle of the room and we have stated a goal, and that is that no d
C[
9 nuclear power plant should expose a member of the public living zcg 10 near it to more than one in a million chance of death per year, Z
_.)
11 and here we all are living near it.
There is that thing
?
j 12 operating, and we know that we have this Nuclear Regulatory d'
13 i Commission which presumably would not have licensed the think if m
i g
14 l the goal was not met.
E l
2 15 !
It carries to me the inference that this thing must 16 j
therefore be offering us no more than one in a million chance d
i, d
17 that we will die, that one of us will die in a given year, when I
5 18 l in fact there is no way as far as I can see that the NRC can be b
19 l l
sure that that is the case.
5 I
20 I MR. KOUTS:
What do you mean by being sure?
21,
MR. MAZUR:
Well, I mean even being reasor1bly confident.
t 22 l MR. KOUTS: What do you mean by being reasonably i
23 '
confident?
24 :
MR. MAZUR:
If there were a probability risk assessment 25 l technique which would allow the generation of an absolute i
1 f
(
ALDERSON REPORTING COMPANY, INC.
I 1
.2 l
1 l probability of that that we could have confidence in.
I 2
MR. KOUTS: See, that is exactly what I said.
l l
3 MR. MAZUR:
And that it told us that the risk was
)
4 actually ten to the minus nine, and I would say that it had been I
c 5
met.
Ea 3
6 MR. KOUTS:
Are you questioning the margin -- are you o
R 7
making statements which are related to the margins of uncertainty n
E 8
attached to probabilistic analysis?
n d
I 3
d 9
MR. MAZUR:
Yes, certainly.
i ob 10 MR. KOUTS:
Okay, that is what I thought that you were s
5 5
11 doing. I think that really some very pessimistic views are being
<m d
12 drawn from what Bob Bernero said this morning that are much more E
4
=
N 13 pessimistic than what he would believe.
E E
14 MR. MAZUR:
I had a.nuch more pessismistic view than z
2 15 Bob gave this morning before I walked in.
i w
z J
16 MR. BERNERO:
Herb, I would like to ask Allan a e
W l
g i/
question.
I E
i 5
18 Wouldn't the same situation prevail on a qualitative 1
G i
19 goal if the Commission, for instance, said I am going to throw 3n 20 away PRA, it is too uncertain, but I do intend to state a new 1
21 way to assess reactor safety; that it should constitute no 22 significant increase in background risk, and I will devise some t
23 deterministic way to demonstrate that.
If the Commission is 24 clearly stating a new definition and not a mere explanation of I
25 what we meant all along with those safety evaluation reports, 4
- 1 ALDERSON REPORTING COMPANY. INC.
j.
3 l
r 52 1
j l
isn't the situtation the same?
2 MR. MAZUR:
No, it is not, because you have another 3
layer of problem there, and that is you have the word games going; i
4 what do you mean by significant, and the goals says that there j
5 g
should be no rignificant risk.
You never know whether you have 1
9 6
got it or not, not just because you don't have a method of i
R j
7 checking it, but because you do not know what significant is.
s
]
j 8
You don't have any criterion for saying when it is significant and l
d
]
q 9
when it is not significant.
i zoy 10 MR. BERNERO:
Which is not a whole lot different from:
3
=
1 Q
11 what we have right now where the statute says no undue risk and Y
12 there is a revealed standard of what that means in the enormous 5a 5
13 i pile of paper,and codes and standards that we judge and do not i
m l
14 judge as necessary.
)
[
15 i
MR. MAZUR:
You are drawing a similarity, and I would say l
=
16 j
j that was an important difference, though, when you actually put A
l U
17 i a number there instead of the wishy washy platitudinous kind of i
a=
{
18 statements, and you are really carrying a much stronger influence.
A
{
19 MR. PERROW:
May I ask for a show of hands by those of n
i 20 you in the implementation and verification section who feel that 21 j
it requires considerable revision or rewriting, is that in order; I
l i
22 l because you asked, Herb, whether anybody had any objectons and 23 have one and a couple of others did.
Judging from the panel that l
24 i I was in, I think that there are several others who have not i
j 25 spoke enough, and I think that we should get a signal this time ALDEi4 SON REPORTING COMPANY. INC.
I 3
t -
--w
..---,.y-,
4
(
i 53 i
i
- 1) again, if I might.
Whoever writes these things should get a i
2i signal again that there is considerable concern about the 3i implementation and verification sections.
i 4(
If it is in order, I would like to ask for a show of 5{ hands from those who think that this section is insufficient and e
R 3
6 needs a lot more consideration.
9 I
/
MR. RATHBUN:I think that is fine, let's have a show of l
j 8i hands, but as kind of a followup on that, I would like to get for d
I q
9 our benefit here a little more information as to what the show of E
10 hands means from the individuals who so indicate.
E) 11 l MR. ZEBROSKI:
I object in principle to a show of 3
i j
12 { hands.
It seems to me that if you want a poll taking that you 5
l 13 l could do it by mail. I thought that the reason to meet. was to 2.
i 3
14 l hear reasoned positions which could be comprehended.
z 1:
I 2
15 i MR. RATHBUN:
That is right, Ed, but that is precisely 5
t y
16 } why associated with that same -- and it does not have to be a i
d 17 poll.
18 MR. PERROW:
All right, I will withdraw it.
i 5
}
19 MR. DERBY: Ed asked"fdria. reasoned argument ~and I think n
20 '
that I will leap to the fore and state it.
To me, the 21f implementation section is vague for I think a very fundamental 22 '
reason.
I cannot distinguish whether or not these things that 23 ' are called goals are in fact goals or standards.
The disticction 4
24 that I make is that a standard or a cri'
.-ia is a barrier or a i
l 25 ; guide to reaction siting design, operatien of nuclear plar.ts,
I
-l I
1 ALDERSON REPORT!NG COMPANY. INC.
I
5 54 1.
signing off on various legal things.
A goal is a way of I
2 evaluating those actions historically and a guide for setting 3
the standards.
So I would look at the imp 1& mentation section 4
and I would say, well, I will reserve this distinction until s
5 I see how these things are going to be used.
They do not say; D
]
6I they do not say whether this is an evaluation of the consequences i
R j
7 of licensing activity that has gone on before and it going to I
f8 direct us solely into evaluating new licensing and setting
~
e d
9 standards.
i 1
o j
g 10 MR. RATHBUN:
Well, I think that there is some confusion z
2
=
j j
11 that continues on that very subject.
The thing that I would guess 3
y 12 that I would call attention to is the first couple of sentences 3
y 13 in the lead-in for implementation, that the goals are not
=
4 l
l 14 l intended to substitute for regulations; that individual licensing 5
{
15 decisions would still be made primarily on ch basis of x
l E
16 compliance with the NRC's regulations.
It is a goal, an W
d 17 i objective, and not a requirement per se.
4 w
f b
18 I have heard it said, and I think that it is basically E
l9 ;
true that it would be essentially an adjunct to another factor 5
l 20 l or to another cut at that problem, another perspective.
4 4
21 MR. COCHRAN:
Could we take some case studies; I would i
1 22 like to run through some case studies and just see what is going i
23{
to happen with this.
We gave some examples in our paper of 24 i some reactor safety issues.
I would just like to find out the
(
i 25 way this thing is going to go with this, with these case studies.
J j
ALDERSON REPORTING COMPANY. INC.
1
53 4
6 i
MR. MALSCH:
I would wanted to make a comment.
The I
2 comment has been made before that we should not use quantitative 3
safety goals as a substitute for the present defense in depth i
4 approach or the present system of regulations that are used.
i It seems to me that the difficulty is that the present system of a
{'
regulations lends itself to the use already of qualitative 6
n i
8 7
safety goals.
In applying i.he present regulations, you are i
Na g
1 M
required to make decisions as to which accidents to postulate d
6 9
j for design purposes.
And decisions are made every day that 1
ag 10 certain kinds of accidents are sufficiently unlikely not to be 2
=
used for design purposes and that certain accidents are of i
d 12 2
sufficient likelihood to be used for design purposes.
I E
13 g
So while the quantitative safety goals will lend E
14
~
themselves naturally to use within the regulations, the more 1
w j
s F
15 1
s accurate statement perhaps is that probabilistic risk assessment J
=
I 16 l
g should not be used as a substitute for engineering judgment.
That gives rise in my mind to the question whether there is a way i
x l
5 18 'I J
=
to use or make useful qualitative safety goals that do not 1
19 j
l depend on probabilistic risk assessment.
20 I
- l 1
Can you use goals in the regulatory process or to make l
a 1
21 I j
l regulatory decisions without necessarily having in place an 22 !
j agreed upon model or methodology of probabilistic risk assessment 5
23 '
1 given the fact that you have *o make these kinds of decisions 24 anyway is a way to use the goal in some useful way even if you 25 cannot rely upon probabilistic risk assessment.
i
)
i j
ALDERSON REPORTING COMPANY. INC.
c,,
~-.
m,__
,_.-.m--
7 l
i 56 1
)
I MR. RATHBUN:
Is it meaningful to an attempt to i
I l
2' establish a quantitative safety goal divorced from scme means of i
3 testing?
j 4
MR. MALSCH: No, the problem is that it is obvious that I
y 5j you should not promulgate a goal that is not going to be uceful, e
1 j
3; 0!
but the assumption is that the only way to make a quantitative i
E 7
]
y goal useful is to use probabilistic risk assessment, and there-8 fore, the usefulness of one or the accuracies and validity of d
i 9
j
}.
one is tied necessarily to the usefulness of the other.
I am c
i h
10 just wondering if that tie in is necessary or inevitable, 3
I MR. RATHBUN:
Why-don' t we pose that very question for l'
s" 12 discussion here; are there any comments on that?
4
=
1
]
5 13 MR. ZEBROSKI:
Another role of a goal is the implication 1
m l
5 I4 that it is an expected result of a set of procedures and i
l b
15 practices.
In the case of pressure vessels, you have a hundred I
E I6 year accumulation of experience of which about the last forty i
W j
h I7 years are almost a coherent set of standards and specifications 18 which actually have gradually improved over time.
The experience w
s I9 8
i is just about the ten to the minus six level you can _say that my n
20 intuitive shooting for one in a million which led to this body of 21 I j
j practices and standards, which is the implementation technology, i
I 22 l' is now more or less verified at a fif ty percent confidence level 4
23 '
that we have not had a catastrophic vessel failure of a high 24 I i
quality vessel.
i i
25 i
But applying the same reasoning to an overall risk i
l ALDERSON REPORTING COMPANY. INC.
l 1
57 1
1 assessment, there would be no way to get from here to there.
t l
2 The experience path is too long and painful, so you have to do it i
I 3
by the pieces some of which you do not get the ten to the minus i
4 six kind of experience to build it on.
3 i
~
5 MR. KOUTS:
Ed.
e h
6, MR. ZEBROSKI:
Chris just pointed out something to me R
M 7
that maybe you are bypassing.
On page 25 and on the next page, n
i j
8I there seems to me that there is a common ground between the 2
d j
9 pro and anti technique users here where you use it where z
c g
10 applicable and you use other things where it is clearly uncertain.
z 1
j 11 g Certainly, if you have a PRA which with whatever level of review a
a
(
12 and audit you can provide tells you that you have a wide margin 5
j j
13 of safety, like some of your SDA's, clearly you can say that that 1
m l
l 14 is a reasonable basis for decision.
i j
j 15 Conversely, if it gives you a clearly bad signal, you z
16 have something that you really do not want to live with, you j
A I
d 17 i-have got to respond to that signal.
I could interpret all this a
I
=
l l
18 l co say that there is a recognition that there vill be a large
)
5 I
(
19 grey area for which this cannot be the primary technique of
}
A
]
20 ;
reliance.
That means a very tempered statement of the basic 1
21 goal recognizing an area of limited applicability.
Nuvertheless, 22 it is a guide to actions and priorities.
]
I I
i 23 MR. COCHRAN:
Then would you concar that putting terms i
24 in the quantitative goal like estimated statistical mean is i
25 i unnecessary?
6 1
i ALDERSON REPORTING COMPANY. INC.
i
1 l
9 i
i 58 l
MR. ZEBROSKI:
Well, that is where you are going to get I
I i
l 2
the big disagreement.
I would not be surprised to see ten to the i
j 3
two or ten to the three differences in the numerical results i
i 4l when you say what is my safe side number versus a most i
I 5
g probable value, and that is unresolvable by present techniques.
i.
n.
I g
6; I think that that was what some of the panel were talking about i
R 1
0 E
7' when they say that there is a deficiency in verification.
When b
3
{
j 8l you have a widely safe or a widely dangerous result, then those 1
4 E.
9ll l
things wash out and you do not need to argue those things.
j
@ 10 ll i
Dr. Lave's summary said and the panel discussion said z
=
i j
j Il that the use of a statistically expected something or another was-i j
1 2 li 4
a very weak reed for the gray cases.
5 j
j 13 l MR. COCHRAN:
Am I reading you correctly; I mean do you 2
i j
h 14 l concur that it does not make sense to force to choose the best b
l
=
g 15 l estimate, the mean case, by putting that in your goal rather than
=
E I0 leaving that for an issue in terms of implementation.
t A
N I7 MR. ZEBROSKI:
I am not quite sure what statement of t
r IO l goal that you apply it to.
h
[
E l
"g 19 i
i MR. COCHRAN:
I mean that all three of the quantitative n
20 l goals have a phrase in there " estimated mean" or " estimated 2I statistical mean," and I am trying to get those terms out of the 1
22 ! goals and then when you are asking yourself whether you are meeting i
i 23 '
the goal, you would say, well, my estimated mean is this and my l
24, 95 percent confidence limit is that, and you can make some i
25 j judgment about how far off you are,and how uncertain you are, and
.j i
l ALDERSON REPCRTING COMPANY. INC.
l i
}
- ~ - -
Lupton 1
59
+
1!
so forth.
l 2
MR. ZEBROSKI:
I think I agree with you on that.
1 3
As a guide for where you put your attention, it is good. As a 4
proof of meet, don't meet, it is very -- that is what Bob was g
5 saying this morning, too.
N 0
3 6
MR. WHIPPLE:
The implication of what you are saying 1
g l
5 7!
ties into Allan's comment.
If you are only willing to use the l
j 8
goal for part of the problem.you can't carry the implication i
d i
d 9
that a reactor is meeting the goals.
I i
10 MR. KOUTS:
I think probably the idea of talking about E
j 11 reactors meeting goals is foreign to the section on implementa-3 y
12 l tion in any case for most of the discussion, because the method 5
I
)
i 13 i of implementation tries to avoid individual determinations of that i
E i
1 m
j g
14 kind except when they are absolutely necessary for other reasons, u
i e
E 15 such as for backfitting purposes.
You do an analysis for an E
4 j
j 16 individual plant on backfitting questions to determine whether or i
a l
d 17 i not backfitting is required and to what extent it is required, w
I
=
l
}
18 and there is a table which determines how rapidly you take A
4 E
19 certain steps.
5 i
n j
20 But in general, the implementation section is a means 4
21 by which implementation is application to deterministic j
j 22,
requirements.
Deterministic requirements which are themselves i
j 23!
the basis for licensing actions.
And there is this one step I
j 24 '
removed attitude in implementation which has to be borne in mind.
i d
1 i
25,
Ed O'Donnell has been trying to talk for some time.
i 1
i ALDERSON REPORTING COMPANY. INC.
1 i
-e-#e.
.--.-w,
-, =,.
r,.
r
,,,,m.
y,
11
)
$Q l
MR. O 'DONNELL:
Well, I ought to let these fellows go j
i 2l first.
i MR. PAGE:
I think my feeling is also along with some 3
F f the others that to have a -.by itself doesn'tLhave much 4,
e 5
meaning unless there is some way of knowing whether you are 5
8 6
close to it or far away from it, and that means that you are e
7 going to have to rely on some sort of probabilistic risk h
assessment to put some meaning into the numerical goal, and 8
M N
the probabilistic risk assessment itself doesn' t have much 9
i i
l 10 meaning unless you have some way of seeing whether you are Cz l
l jj doing well or badly with that.
g 12 MR. KOUTS: Of what?
z l
=
5 3
13 l MR. PAGE:
Unless you have some way of validating:the i
E I
l E
14 probabilistic risk assessment.
It seems to me that our approach i
s!
15 to validating probabilistic risk assessment may be leaving.out j
i 16 a great deal of information that could be used skillfully,
~
n l
'A I
d 17 l and that is that the're seems to be a feeling unless these i
5 f
E 18 Probability numbers are pretty high so that you can get 1
i
(
j9 frequency information on them by repeated accidents, you.can't I
A l
20 say anything at all about the assessment as a whole, but I think I
]
qi that that is getting up far too early.
There are techniques 22 of validating probability risk assessments on unique events i
23 which happen only once, so it is clear that you don't have to 24 have frequency information to validate probability assessments, 2
23 which is a very important observation, considering that 1
f ALDERSON REPORTING COMPANY,INC.
i
l 12 61 l
1 dealing with these low probability numbers you are not going to 1
i l
2l get much frequency information.
I 3
So what you do have is --
.o give you a data base is 4
that you have lots and lots of predictions, and you ban'begin to g
5 see how these predictions are coming out in light of each year's l
8 j
6 experience, and in doing this you can get around or dealvdth R
l s
7 the kind of problems that Tom Cochran thought of, which is j
8 suppose that people make probabilistic assessments of events, and i
d d
9 they say probability is real low that they are going to get melt i
o I
g 10 down in a breeder program, and you have two melt downs out of E
1 l
11 four, and on the one hand people will'say,well, it doesn't matter a
d 12 because we have learned from it and now the rules are different, z
1 5
i d
13 we will design them differently so the probabilities are less.
E 4
j j
14 l Okay.
You know, that -- there may be some truth in E
2 15 that.
On the other hand, it may also be that to have two i
w a
=
I l
j 16 l accidents out of four where the underlying assessments of i
j t'
17 i accidents were far lower'than"that, it says something about your l
risk assessment techniquer and you are throwing that information 5
18 l
-1
=
i j
{
19 ;t away by saying we are going to go on, revise our learning curve M
j 20 l and so on and so on.
I 21l Well, that is fine, but this learning curve tells you l
22 '
something about how your assessment techniques worked in the l
I l
23 '
past and that is --
24 ;
MR. ZEBROSKI:
That didn't exist at the time of EPR 1 i
25 and Firmy 1, so your statement is way out of --
i i,
i ALDERSON REPORTING COMPANY,INC.
I
13 j
l 62 l
l j
j !
MR. PAGE:
I'm sorry, I couldn ' t hear you.
1 l
j
}
2 MR. ZEBROSKI:
The technique didn't exist at the time 3
that Firmy and EPR 1 were designed.
That was pure engineering 4
judgment with very little background experience with the system.
5l MS. WEISS:
But surely -- in the licensing was a e
i I
h l
6:
judgment that they would perform better than two melt downs in 3
I m
k7 four, so I mean the observation that experience showed that 8
8 something was wrong --
n l
d o
9 MR. ZEBROSKI:
There was no licensing for EBR 1, and i
i h
10 barely a primitive licensing process, a siting process for
.i 3
j E
11 Firmy 1, so to say that the technique was tested and found
<k d
12 wanting is just seriously inaccurate.
z 5
d 13 !
MR. PAGE: Well, I think you are missing the point.
l r
1 i
l E
14 l The point is that when a risk assessment is done today, there I
-=
j 2
15 j are literally thousands of implicit probability predictions w
i,
=
l 16!
buried in that risk assessment.
Now, if we did the risk n
i M
]
f 17 '
assessment in a way which could be scored that means two things.
1 5
E 18 l One, that you force people to define the risk assessment and i
5 l
l C
19 !
document it carefully enough so that after the fact you can go 1
5 n
3 20 l back and say, what was the prediction for this event, you know, i
l 21 i which apparently, according to Lewis, is very hard to do on the I
l 22 h existing assessments, j
]
23 The second thing is that you have to be able to, for a i
24 '
given event after the fact, you have to be able to say, was4 f
i 1
i 25 this on the fault tree at all.
And so part of your predictions t
l
}
i ALDERSON REPORTING COMPANY, INC.
l
63 l
l l
ji are the decision nodes in your fault tree.
Another part of your l
2l prediction is to say what fraction of accidents are happening in 3
the course of a year were in the list of things that you were 4
predicting, and it seems to me that there has been very little 1
e 5,
discussion of that sort of thing, and that leads to some sort of M
i a
I d
6l handle on the problem of predicting design error or human error e
l a.
7 which has beea neglected.
s 8
8 So, all I am trying to say is that we have an enormous n
d d
9 amount of information which we ace not using and which we could be Y
E 10 using to increase our confidence in probabilistic risk assessment, z
11 but it seems to me this is fundamental if we are going to rely
<3 d
12 !
upon it in order to give some meaning to numerical goal setting.
z
=
i 5
13 l MR. JOKSIMOVIC:
We are doing a great deal of that.
~m E
14 i MR. PERROW:
I think it is an excellent point.
I would w
b!
15 like to come back to the previous discussion for a minute and 5
16 :
ask, I guess, Ed and Herb about this.
On Page 26 it says 3i 17 safety goals will be applied in situations where regulations do i
=
18 not dictate -the regulatory decision, and it gives an example E
i
[
19 of considering the request for exemption from safety related 5
n l
20 l requirements.
Now, I just heard Ed say that we now know that
-6 21 pressure vessels are safe to 10 o.
one in a million.
Would this 22 mean -- now, we have that.
That is a probabilistic risk l
23 '
assessemnt.
We have that now.
Could a plant come in and say, 24,
well, we have this risk assessment on pressure vessels now, will 25 you please remove the requirements on emergency core cooling?
Or s
i i
l ALDERSON REPORTING COMPANY. INC.
l 1
1
i l
64 1
something else?
Because we know that the pressure vessel from i
2 our many years' experience now, and even as I heard during 3
one of the -- we even know how they will perform under nine 4
feet of river -- cold river water with some salt in it, because e
5 they have been used in boats.:
I am speaking of the Indian Point N
6l
-- on the basis of that we can now release the plants from other R
7 kinds of safety related requirements that bear upon exposure s
j 8
or failure of the -- couldn't they come in and say that?
d d
9 MR. KOUTS:
I am not really the person to say what Y
10 the regulatory staff might be faced with or might do under these j
11 circumstances.
I would be surprised at any kind of relationship 3
y 12 l of the kind you talked about.
I think pressure vessel failure
=
y 13 has very little to do with requirement for ECCS, for instance.
i l
l 14 I can't see that that would be connected in any way.
I don't i
E i
l 2
15 :
know -- I don't know really what a good example to apply to this, l
r a
l
=
y 16,
Maybe there might be --
A i
d 17 f MR. COCHRAN:
How about banning the containment at
[
5 l
5 18 24 hours2.777778e-4 days <br />0.00667 hours <br />3.968254e-5 weeks <br />9.132e-6 months <br /> --
=H
}
19 ;
MR. KOUTS:
There is no regulatory requirement for that.
n 20 MR. COCHRAN:
The Clinch River reactor, in the 21 !
licensing, the staff proposed that the containment on a design i
i 22 l basis accident explosion, that the containment need only last i
23 '
24 hours2.777778e-4 days <br />0.00667 hours <br />3.968254e-5 weeks <br />9.132e-6 months <br />, and it could be vented.
Now, that is an example.
I 24 would like to know how this is going to be used to determine i
25j whether that is a good rule or a bad rule, whether the
?
i ALDERSON REPORTING COMPANY, INC.
I
i l
65 l
4 i
i containment ought to be -- last for the longevity of the 2
accident or whether they should be allowed to vent in 24 hours2.777778e-4 days <br />0.00667 hours <br />3.968254e-5 weeks <br />9.132e-6 months <br />, i
3 I would like to see some of this stuff -- see us go through 4
some of these things like this and see what is going to happen.
5 MR. KOUTS:
That would be an interesting exercise, i
9 i
3 6
Maybe Bernero can she some light.
R I
7 MR. BERNERO:
The question of examples came up in l
5]
8 Panel C, and I offered one example.
I believe in.the talk d
d 9
earlier this morning to the whole session I spoke of an assessment I
i O
i 0
10 of a plant, Crystal River, where there was a calculated l
3 l
l 11 probability of core melt of sout ewo times 10 per year, which
-4 4
]
p 12 is above the postulated goal level in here of one times 10 per
}
3 E
13 year.
2 E
l j
14 '
And in that particular plant, the vulnerability of the l
2 15 core melt is tied up in a rather narrow circle of equipment I
E I
y 16 associated with the auxiliary feedwater system, and one example s
i 6
17 i of how such a goal, if it existed, if one had a goal as drafted 1
m i
=
M 18 here in this talking paper, you know, this Xerox paper we are f
=
i H
l
}
19 all speaking to, the NRC would aay, your plant -- although it n
20 satisfies the regulations for auxiliary feedwater systems and 1
j 21 redundant AC power systems and DC power systems and so forth, i
22 that estimated reliability of decay heat removal is not quite i
i i
23 ;
enough.
The probability we calculate by this recognized
-4 24 li procedure is two times 10
, go fix it, and fix it in a way to
+
l'
-4 25,
make it less than one times 10 and therefore meet a goal.
a h
I' 1
ALDERSON REPORTING COMPANY,INC.
1 i
l 66 l
l l
i l
1 j On the other hand, if you read the proposed goal that l
2 the AIF has submitted, there is a fairly significant difference.
3 It would follow those same steps up to that point, and say, 4!
aha, you have something above the core melt probability secondary I
g 5,
goal, and based on that, you should now do a full scale risk a
6l assessment of the plant, and really look at the public health R
l 8
7' risk, because in that particular instance that plant is in a j
8 very remote site.
It is on the Gulf coast of Florida.
It has d
d 9
virtually no one living near it.
And although its probability 10 l l
of core melt might be a little high, its risk to the public E
I j
11 health and safety might still be acceptable.
That would be a
's j
12 l difference between the AIF proposed goal and this proposed goal.
5 l
g 13 :
So there would be an example of --
=
l 14 l MR. COCHRAN:
How it is used to undermine defense in 5
2 15 ;
depth, right?
How it is used to --
16 MR. BERNERO:
Which one are you speaking of?
j^
\\
d 17 i MR. COCHRAN:
The second case.
The first case, you 5
18 i don't need the goal.
The staff could simply say, we have l
E
-4th
}
19f' calculated the probability core melt is two times 10 and 5
20 we think it is too high, and they could go to the licensing board 21 and say that, and they don't need all these goals to say that.
l 22 l MR. BERNE RO:
Well, the staff can also on the.next
~
23 plant calculate two times 10 and do the same thing.
Now, in l
24 '
that particular instance, the two times 10~4, there existed i
j 25 l no goal, and I' wrote to the licensing people and said, I think l
i ALDERSON REPORTING COMPANY,INC.
~..--..--.
1 67 i
1 you really ought to do something about that decay heat removal 2
system, and the owner, I think, feels, at least from what we 3
have gotten from them, they feel the same way about it, and the 4
only question is how soon something is done, but lacking a goal
-5 5
there is just as much r 2.on to say go fix a two times 10 core e
Au 6
melt as there is for a two times 10-4 That is --
R 5
7 MR. MAZUR:
By the way,the factor away from the goal s
j 8
if it were there'is only a factor of two, and I find it very d=
9 interesting that this -- the result is that accurate, that one ie g
10 can unambiguously say that you are a factor of--
3 5
11 MR. BERNERO:
I would be the last to say that_the
<B g
12 result is that accurate.
=
5 13 MR. MAZUR:
So if in fact one made the calculation, E
y 14 and if in f act the goal were there, wouldn 't in fact the people 5
a E
15 it was going to cost something say let's review the calculation j
16 and --
G p
17 MR. BERNERO: They might, but other factors come in, as a
=
j 18 it did in that particular instance. The type of failure, the P
I
[
19 l type of core melt estimated is among.the more severe core melts.
M 20 l It is not one of the less challengingLbut one of the more 21 l challening core melt sequences.
Increasing the gravity of the --
22 increasing the concern.
And other -- you know, there is more 23 '
to it than the mere probability of damaging equipment, you know.
1 24 MR. ERNST:
I think there are other aspects, too, 25 and that is, that is only one contributor.
If you have one i
i f
ALDERSON REPORTING COMPANY,INC.
68 l
I
-4 1 i contributor that is two times 10
, you will probably find that l
2' it is a far dominant sequence and we are not just talking abcut 3
a factor of two for that sequence.
It would probably be a very I
4 dominant accident sequence.
It may be a decade more than the 1
l e
5 next dominant.
3 N
h 6
MR. BERNERO:
I wonder if I could just -- on the R
)
R 7
exemptions, I can 't remember - ; perhaps Marty Lofts can ' remember i
Xj 8
-- the wording of 50.12, but when you seek relief from d
=
9 regulations, a risk assessment, at least I don't foresee its I
io i
g 10 possible use as addressing one issue and showing sufficiently 3
l l
5 11 low risk to waive requirements in a totally different arena.
N i
d 12 An example that I can think I, and I am just synthesizing this, 1
h 13 there are requirements within the existing regulations and E
l l
14 guides and codes now that show that the reinforcing bar in the 2
15 concrete of tra containment building should be spliced-sith s
i
.j 16 i Cadwell splicing, certain procedures should be followed to l
17 i splice the reinforcing bar, to be assured of a good high s
18 integrity joint in the reinforcing bar, and if in retrospect f
5
}
19 l someone discovered that they did not meet that due to control l
20. system -- or quality control errors, or somebody goofed, or
?
l l
21 l somebody falsified records and was discovered after the fact, l
l that the containment was built, and there was a section of 22 I
23 splices that weren't done with Cadwell, they were done some other l
24 way, a risk assessment might be used to seek a waiver on the i
25,
grounds that the contribution of that anomaly or lack of I
ALDERSON REPORTING COMPANY,INC.
._.._._,_.y
-_..m,._.__m..
_y
- - _ = =.
l, 69 j
l I
I I
meeting the requirements is not significaat because of, and l
2 some legitimate, analysis fer it, but not to go from the Cadwell I
3 or the containment risk over to ECOS or anything like that.
4 MR. COCHRAN:
That's an example of the hoop that I was e
5 referring to in response to Lester's comment.
This. probability i
Mn 6,
risk asseusment approach would be used precisely that way.
i A
7 Another way to jump out of the regulations where you don't need j
8 them.
d d
9 MR. 22RNERO:
Or to decide whether you really do need i
h 10 to meet them on that limited population or location in the E
5 11 pl ant.
MR. KOUTS:
O'Donnell I think has had his hands up.
d 12,
z l
l 13 l MR. O'DONNELL:
Yes.
I guess I get back to the a
E 14 question Marty raisednof whether or not you can have safety goals u
2 15 without PRA, and again, I guess it might be conceivable that s
j 16 you could have some conservative deterministic way of just A
p 17 j demonstrating you have met the goal, but again, the reverse --
18 I cannot see how we can have the -- using PRA in a quantitative E
g 19,
sense fc. a decision such as the one Bob Bernero mentioned n
20 l without having - me criteria, goals for acceptability, f
21 l That is, when it a system design acceptable or when l
22 isn't it.
You do a quantitative analysis.
You have got to have i
23 some sort of quantitative number.
I 24 With respect to the exemptions, I see the use in this I
25 way.
Right now, I mean, a utility can come in and request an ALDERSON REPORTING COMPANY. INC.
70 1
exemption, and the NRC has to deal with that.
They have to make 2
a decision. Right now they would do that on I don 't know what 3
basis, but PRA 1 could see being used in that centext where the 4
applicant would bear the burden of proof for demonstrating that a
5 that exemption is acceptable, and the standards for that, he may 6
3 6l have to demonstrate that even with the limits of uncertainty, R
R 7
that he still meets.whatever goals have been set up.
A j
8 But again, you know, I agree with the beginning d
=
9 statements of this. implementation section.
The regulations..are i
O g
10 still going to be the basis for licensing, and the question 7
3 11 really is not whether a plant meets the goal but whether the w
j 12 l regulations meet a goal.
That is, if we are going to try and E
I y
13 l determine whether the body of regulations we have are rational,
=
i l
14 l we may want to set up a quantitative safety goal for judging --
2 15 ' for setting a quantitative expression of acceptable risk and N
j 16 then using PRA on a generic basis for seeing whether those l
17 ' regulations in fact deliver a design that would in fact meet the
}
18 j
goals.
P 19 The question asked of the individual plant still is, n
20 does the plant meet the regulations.
21 l MR. COCHRAN:
With the techniques in the hands of the t
22 industry, like you and others, what would you guess would be the 23 ' ratio of instances where you would use this technique to seek 24 l exemptions from existing requirements to the ratio of cases where 25 you would deny a license on the basis of not meeting the goal?
i J
4 l
ALDERSON REPORTING COMPANY, INC.
l
.. -.. _ _ ~..., - -
i 71
)
MR. KOUTS:
Well, this is really speculatien, and it y j i
i i
2' is probably not going to take us very far.
1 3
MR. COCHRAN:
I think it's about a sillion to one.
4 MR. O'DONNELL:
I would be willing to, if we could set e
5 rational goals, and accept the methodology, let the chips fall 5
8 6i where they may.
t I
I I
E 7
MR. COCHRAN: Would you be willing -- what position would 1
E 8
you take when Congressman Markey introduces a bill that says the i-0 l
d 9
Commission must certify that it has -- that these goals are met i
4 j
h 10 before it can deliver a license?
It must demonstrate these 1
3 5
11l goals are met.
Are you going to testify in favor of such a bill 3
1, 0
12 or against?
r z
=
l 3
13 MR. O'DONNELL:
I would have to read the bill first.
1 o
4 m
j 2
14 MR. COCHRAN:
I just read it to you.
w l
2 15 (General laughter.)
{
16 MR. KOUTS:
Mel, you have a comment on this point?
k 1
w 17 l MR. ERNST:
Yes, just a couple of observations, I w
18l x
guess.
One is, and I don't think we really feel that way, but 3
4 b
8 19 ;
sometimes I get the feeling that PRA is a dif ferent kind of a
J j
20 j science other than engineering, and I think really PRA is an 1
I i
21l attempt to try and quantify or better quantify what we commonly I
i 22,
call good engineering judgment to try and help the decision 23 process regarding good engineering judgment to some extent.
l
)
24 Secondly, I guess there have been a lot of hoops-
!l l
25j around here, and I would like to say that at least the draft ll l
ALDERSON REPORTING COMPANY. INC.
1 i
.=._ __ --
73
,I 1
such rules are being formulated.
I think that focus gets us out i
1 2
of this interminable mess of what you are going to do when some-3 body has a question on the rules that are made on applying the 4
standards.
I would like to get away from it.
I am tired of g
5 hearing it.
You guys can fight that battle in another place.
Q s
6 What I would like to do --
a R
2 7
MR. BRADBURN:
Before we leave implementation entirely, A
j 8
I would like to just remind all of us of what our panel said do 9
about implementation of qualitative standards, because I think 5
1 5
10 we can talk about implementation of a quantitative standard, but 3
]
5 11 just again to get the notion that something should be in there
- J 12 l and the focus about the process by which the thing goes on, and E
I j
13 i particularly what essentially we were urging was that there be
=
E 14 something in the report and implementation that calls attention w
2 15 to some sort of monitoring feedback of information to make the j
16 process, the learning curve and so forth that we were talking i
s i
6 17 l about before better, and that is an important job.
It is not i
l M
18 perhaps as controversial in a way, but it seems to always get I
5 C
19 left out, and maybe because everybody ' thinks it is self-evident t
3 n
20 or it is in there and so forth but it doesn't get in, and I think l
21 we should -- we urge that it be spelled out in some sort of 22 language like we were suggesting or something analogous.
23 '
MR. DERBY:
I want to finish my point, that the 24 ;
principles that we are talking about are very important to be that[
1.
l 25 ' only principles, to be guides, and noc standards, and if someone l
ALDERSON REPORTING COMPANY. INC.
4 i
74 a
i l
I l
can raise a flag when we get into rulemaking and standard setting 1
I 2i and it is the -- of regulatory activity, we ought to pull away I
1 from it, because if you want to set a number for that, this is 3
i 4
the wrong place.
i e
5 MR. KOUTS:
Gordon Corey.
A N
d 6-MR. COREY:
Herb, I just wanted to make one simple i
e E
I a
7 response to your question which was asked a long while ago, i
A i
8 which I think was, how would you go about implementing the n
dd 9
results of PRA, especially if we didn't have quantitative Y
l 10 standards, or if we established only tentative quantitative l
E 11 i standards, and I think I would subscribe heartily to the last l
I d
12 paragraph on Page 25 and what follows, t
z J
5
{
d 13 MR. KOUTS:
I would like to say a few words more on o
4-
=
)
E 14 the subject of implementation, hoping that it might clarify the i
Uz 2
15 l situation a little bit, and simply repeat the process as it has a
x 16 been described to me as one analogous to what happens under the M
l g
17 Environmental Protection'Act, where similar problems attend the I
s 5
18 analysis and attend the conclusions.
5 j
{
19 l The Environmental Protection Act, according to my 1
5 i
20 understanding of the process, requires that a certain i
21 determination be explored, a determination as to whether a 22 certain balancing -- a certain balancing leads you to a conclusion j
23 '
that you can proceed because the costs are exceeded by the i
3, benefits under the circumstances that prevail.
The universe for 24 f
25 l the determination which is to be made is very large, and there 4
ALDERSON REPORTING COMPANY. INC.
I 1
d
1 75 t
1 1,
are no good guidelines as to how big that universe is.
What are i
I h
2 you to include among the costs, what are you to include among the 4
i 3
benefits?
And as a result, the carrying through of the 3
4 j
4 analysis is beset by a number of great difficulties and the i
=
5 results are attended by a substantial uncertainty, and the way l
N l
3 6
out of this is that you do the work, you do the best you can, and i
4 e
Rg 7
you lay it all out for public exposure in an environmental l
a j
j 8
impact statement.
I d
]
d 9
And this is precisely the kind of implementation which i
t h
10 is considered here, where uncertainties are large for other z
1 g
11 reasons, and where some sort of balancing hcs to take place, 1
2
]
f 12 and in the last analysis you may not be sure whether your I
5 i
s 13 conclusion is precisely on this side of the line or precisely _on I
E j
14 that side of the line, but at the end of the process you come to j
j 2
15 a conclusion that, okay, it looks as if it is in the right
}
q j
16 ball park, let's go ahead, or if it looks as if it 13 not in the 1
A
{
p 17 right ball park, let's not go ahead, and here is the analysis all j
18 laid out for public view so that other people can second guess I
E j
19 you, do the Monday morning quarterbacking as to the way you did g
i n
20 the job.
l
}
21 And this is my understanding of the implementation i
22 I Have I put it right?
l process.
j 23 !
MR. RATHBUN:
I think so, and I think that is l
24 l certainly a central theme that we have had in mind, and I guess j
l 25 I do think that the implementation that was envisioned here is I
i I
ALDERSON REPORTING COMPANY. INC.
l i
)
76 l
1 1
-- the spirit of it, anyway, was best characterized by this 1
2' Paragraph at the bottom of page 25.
A 3
Is there any other comment on --
/,
MR. LAVE:
Yes.
Can I just respond to that for a l
g 5
second?
I think that there is a logical inconsistency in there.
N 6l And that is, the reason why you set quantitative safety goals, R
-5 R
7 like 10 is so that you have a technique for meeting them, you j
j 8
meet them, there is no comment, everybody goes home.
That is, the 9l 4
d whole reason for setting the quantitative safety goals is to get 1
d i
Y 10 around some long process considering qualitative features and z
4
=
g 11 all sorts of other intangibles.
~s j
12 You know, if you've got a building design that says E
3 j
13 the - stresses on this thing cannot be more than blank, you
=
j 14 I calculate it, a couple of engineers talk.
You never hold a w=
9 15 public hearing.
There is no reason to hold a public hearing on 4
=
{
y 16 that.
Everybody goes home.
You only hold public hearings and 1
A i
f 17 l talk about intangibles when you've got some problems.
Well, if i
s i
M 18 !
your technique for estimating the quantitative risks at any 4
=
I 19 particular reactor or in any case give you a wide range of n
20l variation, and even there you are not sure you are within that 21 wide range of variation, then by definition you are dealing with 22 l some intangibles, you had better hold some public hearings on 23 them, and at that point I don't see what stating quantitative 24{
safety goals has added to that.
25 q And in particular, I will tell you that it has added
)
i i
ALDERSON REPORTING COMPANY, INC.
l
l 77 i.
l l
a big mess to it, that is, it may be self-evident to some 1 ll 4
-5 1
2l People in this room that a probability of 10 is negligible or J
3 trivial.
It is not obvious to me that that is so.
And in order 4
for me to Cign onto that, not that it is terribly important that 4
e 5
I do, but if I put on my citizen hat and say, in order for me to 3
i j
i 6
sign onto it, it would take a vast amount of thought and i
4 I
R j
l j
g 7
calculation and so on, and being a lazy type, I don't want to do i
t 1
j 8
that, and I am sure other people don 't want to do that, and you
-4 9
start taking a look at 10 for the probability of core melt.
i i
E i
10 Again, it is not self-evident to me that that is a E
I j
5 11 probability that we can all live with and we can go home on that.
i 1
d 12 Somebbdy is going to have to spend a long time convincing me, and d
d 13 if what we wind up with is something where we've got lots of E
j 14 intangibles and we are going to have to consider on some merits, i
5 2
15 I don't understand why we want to go through setting those 4
s i
=
y 16 quantitative safety goals.
Why do we want to have all that i
51 l
d 17 l hassle if we are still going to talk about intangibles after the l
l J
E 18 j fact?
3 19 !
MR. SEGE:
May I offer some. comment on what Le~ ster 8
3 i
a 20 l said?
OneJ9ay of looking at quantitative safety goals is not 21 that it makes the process of regulation easier, but that it makes 22 I it better.
We have not said that it is the purpose of i
23 quantitative safety goals to make regulation process easier.
24 i What we do hope'is that the rationality and coherence of the N,
25j system of regulation may undergo or should undergo some i
ALDERSON REPORTING COMPANY, INC.
I
78 1
improvement as a result, and when it isn't possible to prove 2
with a degree of certainty or precision what degree of safety 1
3 a particular plant has reached, if one can nevertheless i
4 expect that there is some rough validity at least to PRA j.
p 5
results, so that in some number of cases of major safety issues E
6 they are of some value in guiding the development of new rules, i
n j
R 7
modification of old ones, or in judging some exceptional cases,
}
E l
l 8
that deriving some value from them in those cases where the d
j q
9 precision is sufficient for the regulatory action involved may 4
z 10 be better than not deriving any value at all.
}
3j 11 But certainly I for one did not expect the l
5 j
j 12 regulatory process to become a process whereby one does things E
i d
13 by rote, and much more easily than before, but I think that it W
=
)
l 14 is probably a widely shared view that there would not be a l
{
15 great easing of regulation.
=
l j
16,
MR. COCHRAN:
Can you give us three or four examples?
I i
p 17 {
MR. KOUTS:
Let's not have any examples.
We have some l,
a
=
l 18 questions, some comments.
The examples are examples of 1
j
{
19 ;
application, and I think we have all concluded that the people.
1 5
20 who do the application are not even in this room.
1 1
j 21 MR. COCHRAN:
Some of them are.
1 l
22,
MR. KOUTS:
We do have some people who have some i
l 2
23 '
comments.
Yes?
24 l MR. MAC LEAN:
I think it would help -- it would help i
i 25 ) me anyway, to at some point see how it would be applied in 1
i I
ALDERSON REPORTING COMPANY. INC.
l 1
1
l 79 i
1, certain cases.
I know that several people here have suggested I
2 that we shouldn't be talking about standards and rulemaking.
3 We should be talking about goals.
On the other hand, this 4
paragraph that has now several times been endorsed on the bottom i
5 of Page 25 says that the quantitative safety goals will be used 9
6!
to assist decision making with respect to major safety issues 5
R 7
in the development of new safety rules and standards and the j
8 assessment of existing rules and standards, and so on and so j
d d
9 forth, and I think it was the point of Lester Lave *s comment that i
t O
10 you can't really separate these, that if you do establish a E
5 11 quantitative safety goal, once that is shown that that goal is 3
y 12 met, there is no more question that can be asked about rules 5
d 13 or standards, so it is very important, I think, to consider these E
l l
14 together.
e
{
15 l Furthermore, if I understand the upshot of a lot of 4
=
-j y
16 l the comments, and where there is agreement here, if anywhere, x
]
d 17 l that about concerning the vast uncertainty of probabilistic i
M 18 l risk assessment, the limits on what we can -- on the refinements 5
l 1
19 j that are available to us now, and even the usefulness of it, M
a 20 l that is, to identify outliers, and maybe nothing more, perhaps i
i 21 I we should be talking about a quantitative -- a probabilistic 22 risk assessment as a standard and not as a goal at all.
And 23 specifically as a standard that will help us determine when a 24 license shouldn't be issued, or when a correction should be 25 ) ordered, and for nothing more.
That is perhaps the state of the i
ALDERSON REPORTING COMPANY,INC.
80 I
art is such that it can only be used now as a, say as a necessary
}
2 but not a sufficient requirement on the part of the licensee
]
3l and as sufficient but not as a necessary condition for refusing 1
4 a license, and that is as a standard, not as a goal.
i e
5 MR. KOUTS:
Mr. Bradourn?
j N
i 3
6, MR. BRADBURN:
I would like to raise a question about n
g
]
g 7
George Sege's statement that quantitative standards might make i
s 8
rulemaking or regulations better.
Just in the limited amount of i
n d
6 9
observation I have had about bureaucratic processes and Y
10 regulation in general, is that it is sort of quantitative i
i 2
11 Gresham's law, and the quantitative things tend to crowd;out 4
i a
d 12 '
non quantitative standards, and I would expect thac if one gets z
i
=
h 13 l goes ahead and has some kind of quantitative goals, or what
)
]
]
14 l I think are standards here, even in spite of all the things that 0
{
}
15 one says on Page 25 or beyond about the flexibility of their use, z
16 that they,will in fact end up getting mandated, and ge' ting i
3 A
l g'
17 incorporated into a law, and we do not have a technique --
E I
E 18 everybody says that we will be able to -- people can agree on as l
l 19 :
to how you met them, and you will be in a much -- it will not 1
5 i
20 lead to better regulation.
In fact, it will lead to worse l
l 21 l regulation. That possibility has to be given very good 1
l I
22 l consideration.
I 4
j 23 MR. LAVE:
I have another sort of feeling about that, I
24 and it follows up, I think, on some of Doug's comments.
That is 25 that as a practical matter, if you were talking about how you l
J I
j
[
ALDERSON REPORTING COMPANY. INC.
l d
~_.. -.. _ _ _ _ _ _ _... _ _ _ - -,, _ _ _ _, _ _ -
81 1
1!
would establish quantitative safety goals, there is only one way i
2 I know of of doing that, and that is within a context, that is, 1
l 3
this is one of those cases where I think you would have a par-i 4
ticular reacter, you would look at it, you would take a look at i
e 5
what the PRA said, you'd take a look at the design, and you'd 1
3 n
I 6
say, okay, here we have our best estimate of a probability of a o
-5 a
7 core melt is 10 it's unlikely it would be less than --
-4 8
8 greater than 10 and so on, and you would look at that and say, I
d i
d 9
is that acceptable, and everybody would gather around und talk 3
I 10 about the reactor, and the answer would be aye or.nay, and I E
j j
11 think you have a kind of common law development that would gc on B
d 12 from that.
If it turns out that for that.particular reactor that i
E=
l 13 set of numbers was acceptable, then you would go on to the next E
14 one, you would take a look at it, and you would start off and w
2 15 say, okay, these numbers were acceptable in the last reactor, E
i
?
16 they dominate here, is that acceptable, and that would be the a*
1 d
17 I basis of~ how it is you do it.
w=
18 I don't think anybody could sit down in a room like 5
-4 j
{
19,
this and say, I think 10 is the right number, let's all vote l
n l
20 l on it.
You have to do it in a context, and that again suggests i
i 21 that what you would want to do is, you would want to establish a
22 the qualitative safety goals as besu you can.
You would want to 4
i 23 do your PRA.
You would put that evidence into the record, in 24 ;- wnich it would be challenged, people would know about it, and in h
1 j
25 the end you would see how effective the evidence was in the actual 0
ll ALDERSON REPORTING COMPANY. INC.
l
-__ _=
82 1
1 Ij licensing procedure, and I think what would grow out of that i
2 might be some quantitative goals to -.or might not, depending 3
on the nature of the challenge, how consistent you got over time, 4
and so on.
I e
5 MR. JOKSIMOVIC:
So what are you saying,that it is N
N 6
premature to do this?
I C
l R
\\
^
R 7
MR. LAVE: No, I think that the establishing of those 5
1 i
j 8
numbers is not something that people can sit and do around i
0
=
9 the room.
I think it has got to come out of an actual sort of I
i 10 licensing procedure, regulatory procedure to do it.
And in the I
5 5
11 course of that one can also hear by the objections how good the 1
<3 d
12 PRA is in each case, that is, when they start using alternative E=
l d
13 assumptions, what numbers do you get out, what do people i
G
]
l 14 accept, and I think that this is sort of an organic development 5
]r 15 that I am talking about rather than an a priori development.
=
j j
16 MR. JOKSIMOVIC:
Well, it is my understanding that A
j p
17,
when they made a decision at the London Heathrow Airport to I
5 l
j 18 '
allow blind landing, they had a meeting like this, and they i
5 i
E 19 agreed what the number should be, and on the basis of demonstra-A j
20 tion by the vendor it was approved and the system has been
]
21 extremely successful and it was already exceeded.
j 22 MR. COCHRAN: Do you want a counter example?
I 23,
MR. LAVE:
I think you are supporting my point, not 24 ;
arguing against it.
There was a particular context where they had 25j some qualitative goals in mind.
Somebody presented a nunber.
.i I
i ALDERSON REPORTING COMPANY, INC.
y!
83 l
+
1 !
That number-- I think it wasn't the number that swayed them,it l
2' was the whole context of what they were doing.
I think the number 3
that they wound up there is probably a number that is.of.some i
I 4'
use to us.
l c
5 MR. COCHRAN:
Well, there was the case that Carl 1
E a
3 6
Morgan talks about, where they sat around a room like this and I
2 7
safety people came in and made the presentation that the chance s
8 j
8l of a snap reactor burning up in orbit was chance in 10 and so they l
a
=;
9 shot it up and ft fell into the Irdian Ocean.
3 6
10 MR. KOUTS:
That was not a true risk assessment, that z
=
i 9
11 was a very bad analysis.
s y
g 12
(. General laughter.)
E I
N 13 MR. RATHBUN:
Let's take one last comment here.
I have E
i 14 good news.
It is 5:30.
Marty.
w J
E
{
2 15,!
MR. MALSCH:
I just wanted to make a response to the S
l 1
]
g 16 l earlier comment that L' r:er made.
It seems to me that i
l 1
p 17 development of a quantitative safety goal by the case by case I
i 18 i method somehow assumes that decision making in the first case 4
=
1 i
j
?
19 l or the first couple of cases somehow is superior than
)
a 20 l decision making in a broad rulemaking context, and it seems to i
I 21f me that there is more danger of error in case by case decision 22 l making when results of the first case may be in a defined context i
1 23 but with very limited participation. There is more danger that i
24 that is going to be inaccurate than in a situation where it is a i
25 l broader proceeding and more people are involved.
t i
ALDERSON REPORTING COMPANY. INC.
1
I 84 i
l 1I MR. LAVE:
I think that is fair enough.
I mean, I
\\
l 2
am certainly always on the other side of the argument I was I
3 putting before, but I guess if you were going to do it in a i
4 generic rulemaking, what you would have to do wouldi be to bring i
1 e
5 in your numbers for a series of reactors, that is, make it -- put I
h 1
-4 j
]
6 it in a context -instead of just saying, gee, you think 10 is
~
i E
7 okay, and so far the only way I have heard the discussion about j
A
-4
]
j 8
these numbers has been of that sort.
Gee, do you think 10 is 1
a c;
9 okay.
And I honestly don't know how to react to any of that,and i
?
j 10 if we had some context, and if we said that PRA is going to get 1
5 l
l 11 you numbers within the range, depending on the assumptions that 3
j j
12 you make, then I would be much nore informed and be able to j
Ej 13 comment on them.
J l
14 MR. MALSCH:
I think you are maybe agreeing with 1
5 2
15 Tom Cochran.
w I
i y
16 MR. LAVE:
Yes.
1 A
1
-4
)
b-17 l MR. COCHRAN:
10 means one core melt or 60 percent E
j
{
18 chance of a core melt from the reactors that are existing and in c
{
19 the pipeline.
}
l 20 MR. LAVE:
No, I --
4 J
]
21 MR. COCHRAN: That is easy to figure.
l 22 MR. LAVE:
No, I know what the probability means.
.I 1
1 j
23 guess I don't know in fact how that orresponds to reality.
l
]
24 l MR. COCHRAN:
The reality is, you don 't know whether 25 l you are meeting it.
1; 4
1 1
l l
ALDERSON REPORTING COMPANY,INC.
l I
i
'o 85 i
1 j
j!
MR. LAVE:
That's right.
l I
2 MR. RATHBUN:
The hour is 5:30, and we are through with i
l 3
the program for today.
Mat Taylor will talk here at 6:00 4
4 o' clock and present some information that he has on core melt 1
e 5
Probabilities, 1
A n
i 8
6i Thank you very much.
1 e
R I
g 7
(Whereupon, the meeting was concluded, at 5:35 p.m.)
s i
8 8
a d
l 6
4 1
Y i
j g
10 t
z 1
E 1
4 II J
iis l
g 12 Ei j
g 13 i
=
E 14 i
E I
15 I
E j
16 l
d 17 l
- a 5m 18 19 1
n i
20 1
21 !
i 22 I
I l
23 l
24 l
!i 25 i
4 ALDERSON REPORTING COMPANY,INC.
I
!O NUCIZAR REGULATORY COMISSICN This is to certify that the attached proceedings before the th in the matter of:
NRC SAFETY GOAL WORFOtlOP Date of Proceeding: July 23, 1981 Docket !! umber:
Place of Proceeding:
Harpers Ferry, West Virginia were held as herein appears, and that this is the original transcript thereof for the file of the Commission.,
Rossie Sutton Official Reporter-(Typec)
,A
~
Official Reporter (Signature)
I f
I g
I 4
e j
- }
i
-.