ML19309C776

Transcript of 800331 Meeting in Bethesda,Md Re Briefing on Two Approaches to Treatment of Inventory Differences in Nuclear Matl Accounting.Pp 1-52.Viewgraphs Encl

ML19309C776
Person / Time
Issue date:	03/31/1980
From:	NRC COMMISSION (OCM)
To:
Shared Package
ML19309C777	List: ML19309C776 ML19309C782
References
REF-10CFR9.7 NUDOCS 8004090261
Download: ML19309C776 (55)

Text

,

t

1. "%c, o k_,

s %,*

UNITED STATES N UCLE AR R EG UL ATO RY COMMISSION in the m atte r of:

BRIEFING ON TWO APPROACHES TO THE TREATMENT OF INVENTORY DIFFERENCES IN NUCLEAR MATERIAL ACCOUNTING

%?

m m

_,fl 8 5

,- N or o

,,['

1 81 h'

ci

-=

s A r9

' r1 7

PIace:

Bethesda, Maryland EeG) l Date:

March 31, 1980 Pages:

1 - 52 l

INTERNATIONAL VERSATIM REPCRTERS. INC.

499 SCUTH CAPITCL STP.EET. S. W. SUITE 107 WASHINGTON, D. C. 20002 202 484-3550 8004000 %...

?

g NDPc NDP l

,,,, y g i

1/1 i

I i

UNITED STATES OF AMERICA l

NUCLEAR REGULATORY COMMISSION j

i

_______________________________x In the Matter of:

~

i 6

BRIEFING ON TWO APPROACHES TO 7

THE TREATMENT OF INVENTORY l

3 DIFFERENCES IN NUCLEAR 9

MATERIAL ACCOUNTING to

_______________________________x East-West Towers i:

Room 550 l'

4350 East-West Highway 13 Bethesda, Maryland I

14 i

1*

The Commission met, pursuant to notice, for i,

presentation of the above-entitled matter, at 10:00 a.m.,

I i

John F. Ahearne, Chairman of the Commission, presiding.

18 I

BEFORE 11 I

JOHN F. AHEARNE, Chairman of the Commission

o JOSEPH HENDRIE, Commissioner 21 I

RICHARD T. KENNEDY, Commissioner 5

PETER A.

BRADFORD, Commissioner 04 s

i

, e. o..s / c e Amm : =:.

me e 44e*N mfM -e,

r-t 1

i

=

=

1

,i i

I-Marjorie Nordlinger, j

4

!~

Office of General Counsel E. Hanrahan a

9 D. Lurie i

4

-s i

K. Cornell N. Haller 4

i l

R. Moore s

i t

[

l e

i f

I P

i 10 i

e f

ri g

12

~13

.i l

l

?

i i

']

'i 4

f

?

14 e

II i

i l

0 p

'19 l-i g

l l

i i

l0

• 1 M

4 em 4

e

4

==,

t 1

mm'seseems mipswges f u i

l ao san /m M N. t

• WrTT 'tf f

.Q.

+~ "+ * - - *

• T e

sacz sc I

f I

CHAIRMAN AHEARNE:

The Commission meeting will come to order.

I am delighted to have this l

opportunity to have a Commission meeting out in I

l 4

this location, or at least the sense of geographic l

t location, and any more members of NRC than where f

1 6

we.

mmallp hold our meetings.

I l

COMMISSIONER KENNEDY:

Mr. Chairman, before we l

3 l

begin, may I ask if we could have a show of hands l

1 i

of the number of members of public who are here, since i

10 this is a public meeting.

If CHAIRMAN AHEARNE:

Would you like to compare t

that to the number of members of public who we usually i ~.

i have at our meetings.

l' 14 COMMISSIONER KENNEDY:

If I am correct, this I

13 may be the ultimate in our move toward full implementation!

14 of the Sunshine Act we are making certain that too many l

17 of the public will not be exposed to the extensive j

18 i

damage which might result from Overe= the sun's 19 rays because this 10 the first time since the Sunshine

o Act in which I believe the bxal number of attendees l

21 i

of the public has been zero and I think the record ought to note that.

+

CHAIRMAN AHEARNE:

The record will certainly 4

note that Mr. Kennedy believes that this is the first i

l tamsn rcen.6 vese n= sum =9 set x G ENN &WMb F4R. E s kmt'W 6

L 2

i 4

0 paGE NC.

I.

i time the attendance of the public is zero.

I do not I

recall Mr. Kennedy ever having taken that similar show of hands at other meetings.

COMMISSIONER KENNEDY:

No, but he has looked f

3 around the room.

I knew what the answer was going to i

l be before I took the poll.

1 7

CHAIRMAN AHEARNE:

Delightful.

Any other 3

opening comments?

Peter?

l 1'

9 COMMISSIONER BRADFORD:

I suppose I could l

10 work one out if you would like.

COMMISSIONER HENDRIE:

Yes, I would comment i:

I that the value of the Commission meeting in Bethesda i

13 i

is to provide an easier and less time consuming j

14 opportunity for members of the staff interested in a particular subject the Commission has before it to f4 comment and sit in.

It always seemed to me that one 17 8

of the penalties we pay for the separation between I4 up county and downtown offices is that difficulty in time consumption if the staff is to attend meetings.

,0 I suppose you will look around and see what i

other available rooms there are.

7 3

The subjects this morning have to do with 3

safeguard matters and the safeguard staff -- let's see a

semeplace a bit north and somewhat east of here lonc l

1 ie% sc n am c.

(

J sam. ca -ei,e=ar. s. e sur v ss

e 3

1 o

nas we I

I I

enough so that it is not a very handy walk on a rainy j

morning.

The scattered location -- even coming this I

far still makes it very cumbersome for cognizant staff l'

who might like to be here and hear this discussion to attend.

l 4

i I do not know what you can do about it, but 1

7 t

I recommend some examination of what facilities there j

g l

might be in Phillips, Silver Spring, et cetera.

l 7

i If the Commission is to be on occasion, mobile to l

for this purpose, I do not see any reason we need anchor l

11 6'

here in East West Towers for all time.

For instance, 10 particularly for subjects that are of interest to staff's !

13 elsewhere.

!s COMMISSIONER KENNEDY:

With Spring coming, may I suggest a tent to be put up in one of the local parks which would give access to the public.

Let

,.S also add that I also share the concern is for the division of the agency into its many parts 19 i

and five years ago, many of us thought we were going 20 to be able to correct that problem.

l 21 i

I suggest that those of you who five years t

from now who will be meeting in this room from time to time I hope will be able to address that problem in a

,A more forthright ways than we have been able to approach I'

eeva..$ 'depenew L-,

n j

ass 33hifts C C.Se M 97

4 naz sc l

t' I

it up until now.

Furthermore, let me say, that I agree I

with Mr. Hendrie that I would like someday, and I will, i

P 2

inquire of the safeguard staff of the great benefits l

realized coming over to East West Towers this morning i

4 i

3 from Silver Spring, it is a different venue, at least.

t 1

e 0

CHAIRMAN AHEARNE:

If the Secretary will attempt' I

to find similar accomodations or accomodations in other 3

~

buildings and we can have a rotating show.

Peter, did 9

t you have a comment?

I

!0

^

COMMISSIONER BRADFORD:

I was going to say I

If at the rate we are going, I do not think anybody is l

1*

going to care if we do not get down to business.

13 CHAIRMAN AHEARNE:

Fine, thank you.

Kevin?

14 MR. CORNELL:

We are here to brief you about i

13 statistic approaches to material accounting, and I would 14 like to turn it over to Norman Haller to make a few 17 i

remarks.

la i

MR. HALLER:

Thank you, Eevin.

Mr. Chairman, 19 we were asked by then Chairman Hendrie in November to prepare a briefing for the Commission on the statistical approaches to material control and account-l ing.

Specifically, we were asked to prepare an 3

overview of the current approach for evaluating o

• wfw ~o.wacrearwa m %

<D.

i o

5 l

noz sc

~

l

}

1 I

inventory difference data and of the alternati"e game I

theory approach and to compare the two approaches.

Now, the tentative work has been done by 4

our applied statistics branch, Dr. Roger Moore, the head of this branch is here with us today at the table and also Dr. Dan Lurie will be giving the briefing.

i 1

e Dan is a member of this branch and has done s

the bulk of the statistical work here and the presenta-9 tion.

Now, because of the constraints on the time, f

to 6

Dan is prepared to begin now if you wish with the 11 briefing and the carry the view graphs through.

1:

He also has a couple of extra ones he would 13 l

like to put in if the time will permit.

l ta l'

CHAIRMAN AHEARNE:

Okay, well I think you best 13 5

move right into it.

MR. HALLER:

Dan?

I ty DR. LURIE:

Thank you very much, i

la The purpose of this presentation is to give 9

an overview of two approaches to the treatment of 3

inventory differences in nuclear material accounting.

The first one is a classical approach, also i

n known as a statistical hypothesis testing approach.

i

)

4 The second one is called the game theory l

l 2

approach and also is known as a strategic analysis e

i j

omp

.c %verw =a - x l

em e Q fNG"*._ t e Erft W

O 6

l saar we.

I I

approach.

We are going to cover these two approaches I

and compare the capabilities and limitations of the two 2

approaches.

This information is useful in providing a f

1 3

perspective on statement concerning loss or diversion.

6 Before we go on, I would like to hasten and make a state-i 7

ment that the two approaches are not competitive, they i

3 are complimentary.

They address different aspects of j

9 treatment of inventory differences.

10 I would like to make a short analogy as the 11 two approaches may be compared to a smoke alarm and I

burglar alarm, if you can afford, you buy them, if you l

I:

cannot afford them of course you have to assign your l

priorities accordingly.

The only time that the two i

15 i

approaches or two burglar alarms and the smoke alarms 14 will do the same task is if you have a burglar who 17 smokes a cigar, i

la

~

i t

The briefing is structured according to this 19 outline.

Following a brief background, the two concepts j

1 i

that will be introduced are concepts of ID, inventory

=

I difference and J1D limit of error of inventory l

4 difference.

u The classical approach then will be considered f

i 1.mpi.eenn supe r atmnw w l

es amm. cwmm, trous?. t. m.rm et

0 0

7 a.cz sc i

l t

I the game theory will follow and the conclusion of this presentation done over a comparison of the two approaches.,

3 I

A brief background is as follows.

l 4

Between July of '76 and January of '78, NRC r

studied an application of game theory approach to treatment of inventory difference. culminating in NUREG 7

0290 and NUREG 0490.

g s

6 In April of '78 MC&A Task Force rec 3 mended to 1

i the Commission the role of game theory be reviewed by j

to a peer review group.

This was done through NUREG 0450.

11 Between November of '78 and August of '79 i

the peer review composed of experts in statistics and 13 game theory evaluated the game theory and the classical 14 approach and NUREG 0950 was the result of their evaluation.

I In September of '79, Mr. Moglewer talked to g,

3 Commissioner Hendrie and asked that he expressed

9 concern with the material accounting and. consequently

.g in November of 1979, Chairman Hendrie requested a 21 briefing for :he Commission.

I We start with the two basic concepts.

They must be properly defined because wrong

4 implementation of either one of these concepts and 13

' poor quantification are going to hurt both approaches W

0 8

sacz we f

I I

so it is a must that we understand what we mean by ID I

and LEID.

After this definition of ID a model for I

I f

ID will be derived, we will define the limit of error for inventory difference and finally we will relate e

these two definitions to field experience.

We start i

I with the definition of the first term, ID.

l I

i i

What is ID?

First, we have to refer to a 9

concept called MUF which was introduced in CFR and g

then the definition boils down to the last line, where i

11 MUF is equal to the beginning inventory to which the t-addition to materials was added from which the ending i

13 inventory was subtracted and the removal from the j

1s inventory was taken away too.

15 We are all familiar with this definition.

14 Since 1977, ID was replacing the word MUF.

Tne statistical treatment of ID begins with the model.

The model is that ID is equal to true ID, plus an error.

I would like to bring an analogy to the field of

,0 communications.

We have a signal and we have a noise.

l g

The true ID is indeed what we would like to capture, t

visit our target, the error is the noise part.

The l

4 true ID, a a rula, is not kn341 as a matter of fact, 2

it is never known.

It is known, it is not pointed to e

tanwin. % vine.m.m de SIWDe WW"4A N. L e. SurTT '87

O 3

9 mage NC.

l i

t I

any statistics.

I Similarly, the error is not kn341 because if 3

you know the error, you can subtract it from ID and f

get true ID.

?

I The true ID includes, but not excludes loss a

^t and diversion.

The error term does include instrument i

errors, it includes measurement errors, sampling errors 3

and bookkeeping errors.

9 r

Now, I would like you to note here that if 10 any component of ID is a random variable, then ID itself is a random variable.

In nuclear material, 1:

i we do treat ID as a random variable which leads to 1

13 i

the next view graph.

l 14 What is LEID?

Again, you remember that ID, 15 inventory difference is a random variable.

Being a 14 random variable, it has some measurement of uncertainty.

17 The measurement of uncertainty can be given kinds of i

la i

deviation, limit of errors.and range and other measures are still available.

,0 The limit of error is the one that we are j

p.

concerned about.

Two definitions are given.

There

=

=

are many definitions that float around.

Only two are 4

going to be given here.

.J The ANSI definition says that the limit of a

e v o.a <e=na m e m %

age sl2,Pte WITT 9F

4 10 I

sacz sc l

I error of an estimator, in this case, the estimator will be-ID, twice the extent of deviation of that l

3 3

estimator.

l The CFR definition is a little bit long and l

l r

4 confusing and can be interpreted in many ways, depends i

what the other one is and different licensees can t

i be reading into these definitions whatever they want I

l 1

3 i

to read into it.

9 I do not know who wrote the definition for i

10 CFR, I sure hope it was not a statistician.

CEAIRMAN AHEARNE:

I gather you do not 1

t-i believe that the two definitions are completely i

l' equivalent.

j 14 i

DR. LURIE:

They are definitely not.

CHAIRMAN AHEARNE:

Could you say a few T4 words about -- I agree with you -- could you say a 17 few words about why not?

l 18 t

DR. LURIE:

All right.

I will tell you what, instead, I will ask my chief to comment on this.

,0 i

Roger?

I

,1 i

DR. MOORE:

My temptation is to say more than a few words.

l 4

CHAIRMAN AHEARNE:

Go ahead.

2 DR. MOORE:

But we were' told that we only had I

lww=.ano 6 s upemens h '.<.

O M'M W M WD

nacz sc. _,11 o

s I

I an hour for this.

The point I would like to make as

[

2 succintly as possible is that there is no professional h

2 statistician who can interpret the 10 CFR definition and come up with the intended interpretation -- what i

I we take to be the intended one.

Now, I am going to i

6 steal some of Dan's thunder.

He expresses an opinion j

I on the bottom of this line.

We know the jar 921 is, t

it is all tied up with the 95% confidence interval.

9 i

i The problem is that the definition -- a problem in 10 10 CFR is that there is no statement about whether j

11 l

it is a symmetric confidence interval; there is no 1:

I statement about the underlying assumption.

There is, 13 i

in particular, there is no definition of what uncertainty {

14 I

component means.

Now, that is part of what it means.

12 I would like to say that after a considerable struggle to to put out standard in 1516 this was agreed upon and

[7 is a statement that stands up.in all statistical t

is literature.

Thank you.

19 DR. LURIE:

All right, thank you, Roger.

,0 Despite the confusion with defintion, the spb-it of __

l

,1 4

COMMISSIONER KENNEDY:

Does 10 CFR antedate l

ANSI 15167 DR. LURIE:

Yes.

.4 l

3 CHAIRMAN AHEARNE:

But it still has not a I

rewmaro.46 vemeare. 81psseost '=r.

6m t

8 12

{

nos sc i

i great deal to do with it in a sense that 10 CFR even I

before the ANSI standard was written, the 10 CFR would 2

still be a very difficult thing to figure out wouldn't it?

I DR. MOORE:

From a purely gossip standpoint, j

i 6

I would like to also make the point that there was an l

i effort to get into the game early enough to fix 10 CFR I

70.51 before it was put into'hard print.

That effort 9

failed.

10 COMMISSIONER KENNEDY:

What was the date 11 of it?

l DR. MOORE:

I do not know exactly.

I believe j

13 5

it was in 1975, and 76.

I believe the first opportunity j

14 that I personally had to look at it was in the period 13 I

of 1973 to

'74, and I have not checked all those musty td old records to find out the details.

1, i

DR. LURIE:

All right.

Despite the confusion in definition the spbrit of LEID implies that when there l

19 is no loss of diversion, the probability of ID in

,0 absolute value exceeding LEID is about five percent of 1

i the time.

CHAIRMAN AHEARNE:

I gather what you mean 4

by that is that by the spirit of LEID what people in a

}

general around here talk about what does it mean that i

i lewm r o..

,e r a.:cs.s i e s

l m e M s?*Erf. t e surTT -s7

o 13 I

..ca sc i

I I

is what they think it means.

l 2

DR. LURIE:

Well, you remember that I said l

2 there are many ways to interpret the definition on j

the right and if you know what it really means, you l

i 4

I are looking for this particular definition you can i

I find the circumstances and situations in which this 8

t l

defhtLtdon will be in agreement, in general agreement.

I CHAIR N AHEARNE:

I was trying to capture t

9 what you mean by the spbd.t?

to i

l DR. LURIE:

All right.

If you are willing 11 to make all kinds of assumptions, like the normal 12 l

distribution, and symmetric distribution that is part I

ta l

of the normal distribution, than you said 1.96 is equal

{

u to 2, than the two of them are essentially the same.

13 All right.

One concept that I fcel that i

14 I would like to bring out in this point you see that the ID has a vertical bar on the right and a vertical Is i

i bar on the left.

This means that this is an absolute 19 value of ID.

It means, if it is negative, you drop the negative sign in the front of it.

I

,1 s

Now, again, we have a spirit of the definition

=

3 saying that five percent.of an-ID will exceed the 4

corresponding LEID.

Now, go to the field and find out u

if their experience verifies it.

f I

.emp ec swv

• w I%

1

14 f

sacz sc I

What does their experience show?

Taken from the White Book over experience of four and a half years, 17 l

facilities were looked at and 803 inventories were 4

reported in which LEID was also reported with it.

e Out of it 375 of them exceeded their oma LEID.

Nearly 6

I half.

You would have expected about five percent, 7

l roughly 40 but you get nearly half.

It is my opinion, 3

l that something is wrong.

9 i

Next slide please --

10 CHAIRMAN AHEARNE:

I gather that your opinion is not solely restricted to yourself.

i MR. HALLER: There are others who share that 13 opinion, yes.

DR. LURIE:

Here is the classical approach.

q In this briefing, a framework is presented for testing i

7 the hypothesis that true ID equals to zero, i

tg Relative to this framework a management 1

i dilemma is identified.

This dilemma leads to NRC's i

19

o current practices.

The consequences are examined by 21 what statisticians call operating characteristic curves, l

I and finally, the deficiencies in the current practices

2 are listed.

I' Again, we remind ourselves that we have a

'd model in which ID is composed of true ID and an error.

't

= vin.m 6 ',e.eams amx f

ase e M N. L. e udffT '87

O 15 I

s.oz se.

l i

i In the absence of error, the unreal world, the decision I

is obvious.

If ID is identical to zero, there is no f

loss and there is no diversion.

4 If ID is different from zero, you can t

I say that a loss or diversion cook place.

Even if you 6

l have one gram discrepancy because you don't have any i

7 l

error, this one gram is going to toll you that there f

3 was a loss or diversion.

i 9

But, again, we do not live in the unreal world, to we have to descend down to the real world and let us l

11 see what statistics does for us.

l All right.

What we do here is first we i

I:

in a large threshold in such a way that if ID l

set ta i

exceeds this threshold, we are going to make some kind i

15 of statement about the diversion or loss.

14 Let us consider this statement here in the 17 first column.

The best of both worlds is when true i

18 i

ID is equal to zero and the ID fell short of the threshold.

0 Our conclusion, therefore, essentially again if the true ID if I know is equal to zero,

' we made the correct decision.

If true ID be equal to 4

zero and ID was short if the decision is going to be 2

reached if the true ID is not equal to zero and our s

i m ecs.4% screava.

• m :=c me W CAF*Ul.

SWffT M

e e

16 i

nos se.

i I

we would get a false alarm.

If ID is not equal to I

zero and the threshold was larger than the ID or ID i

.4 smaller than the threshold, we have an undetected loss l

i or diversion.

Correct decision is going to be reached i

^-

if the ID is not equal to zero our statistical ID i

6 as being larger than the threshold and we made a 7

correct decision.

We know that something is wrong.

1 g

Of course, that we have some problems and the problem j

is that the probability of false alarm say five percent to i

it can be dictated by the management and we know that t

11 if che ID was larger than the threshold we could have i

1:

I I

made a mistake.

The mistake, the probability of 13 making a mistake is exactly five percent because we I#

i have decided on it in advance.

If, however, the 4

15 true ID is not equivalent to zero and we fell short of the threshold in our inventory difference, we do i

not know whether there was a diversion or not because i

la i

t 19 we cannot calculate the probability that a true ID g

was not equivalent to zero whether ID was smaller than the threshold.

In order to do so, we must know what p

=

exactly was diverted, how much was diverted.

If we n

know that five kilogram is diverted, we know what is 4

the probability of getting an ID smaller than the 2

threshold.

But if we know that five kilograms were 8

u %

i m e C,4P'tlA iT*E? 1.

• Rf A T

nas ns

$~

we do not have to use any statistics.

i Here is a management dilemma.

We have to set the threshold to compromise somehow between the probability i

4 of false alarm and the probability of undetected loss 3

or diversion.

t 6

i If the threshold is too small and let us look i

7 at the bottom line on the left then we are going to 3

have a high probability of fa'ise alarm.

If the threshold 9

is too large, and again on the right hand side below to i

we have low probability of sounding the alarm in case U

of loss or diversions.

We had to compromise between these two concepts.

t Next slide please.

l 14 What are the practices that we have right now?

13 I call it a departure from statistical principle.

j I4 First, by CFR, we have upper limits on LEID 17 i

and these limits are given in percentages of addition 13 or removal from material in process.

This is percentage l

l 19 of the through route, whichever is greater.

.g Now, you know the limit of LEID by LEID limit.

I 21 l

Remember what it says.

The name of LEID sometimes has l.

C l

to be spelled out.

It is a limit of error.

An LEID Limit is a limit of the limit of the ID.

What is a management action that we have to take according to

~

I e

mm., vm m. m,,,n,s ;% e k

gata,N QN STWEE*. $. e.

sufft 'et

18 Q

s asGE % C.

I I

i 1

l pages note of 1974?

If ID is larger than LEID, you 4

t notify NRC.

If ID is larger than 1 LEID limit, you notify NRC and you take a reinventory.

If it is larger 4

than two times LEID limits, you notify NRC, shutdown f

h and continue to search until a certain criterion is 4

met.

i l

If ID is smaller,than LEID, it is implied j

3 l

as a rule that no loss or diversion of this, this is 9

i a practice and if ID is smaller than LEID, we tend or to i

one tends to say that there is no loss or diversion, 11 when actually, we cannot make a statement like this.

j 10 t

Next slide please.

What are the consequences 13 that we have to worry about?

In this slide, I tried to depict the probabil-g g

ity of sounding the alarm or the probability of no g,

alarm when LEID or 1 LEID or 2 LEID are set at a i

i g

threshold and different amount of material is diverted.

Take for example that a bullet in the center p,

8 o

it says that if 1.5 LEID is stolen, the probability n

of detecting or sounding the alarm is

.5.

Let me back up and say one more thing.

I should mention at this point that let us assume for the sake of this example 04 that. sigma is equal to 1 kilogram.- Therefore, LEID

• d being two sigma is equal to 2 kilograms.

Now, let us L

s m % v - m. % %

[.

i

19 c

• • as we.

I assume take another example.

If the threshold is set I

at 2 LEID, maybe 4 kilogram, what is the probability B

that some 3 kilogram you go to the upper bullet and you A

read on the left the probability of no alarm is about

.84, probability of alarm is about.16, that is 16%

f I

6 chance you will sound your alarm is 3 kilogram is l'

I diverted when you set the threshold at 2 LEID.

3 e

Now, if you are going to go one step further i

i I

and y'ou plot the LEID as a rule it will be 10 on the curve to the right of the right of the three curve.

Now, there is a good chance that 3 kilograms i:

I will be diverted and you will not sound the alarm, or i

13 you will sound it with probability very close to zero.

l u

Next slide please.

Deficiencies in current 12 practices.

14 4

COMMISSIONER HENDRIE:

Can you just pause

.7 for a second and remind me again in setting the limit i

la i

of error on inventory difference, the licensee estimates 19 on some basis or other his errors in the various 20 quantities primarily is measurement errors I guess i

,1 in actual material quantities and then in effect assuming that those errors are random errors, assumes a gausian distribution and goes up 2 sigma to set his l

2 limit of error.

s

~ ~ ~,

- u aus sintf.e Cam N. E e. ud87T '97 I

20

.I n

nacz sc l

I DR. LURIE:

Well, the process of constructing I

LEID is very complex.

I am not quite sure that we l

really know how to compute it.

There are many f

3, components that have to jive together.

Many of them t

e 5

follow the normal distribution, some of them, I am i

I sure do not follow the normal distribution.

I am not 7

I sure that anybody really knows how to compute LEID, j

1 given all this --

l 9

l COMMISSIONER HENDRIE:

Yes, we have licensees 10 and their licenses say that when they exceed LEID 11 i

certain things have to happen.

So whether we know i

I2 i

how to compute it or not, there are a number of operators !

13 in the field who in fact do compute it and what I am j

ta saying is that I believe that their LEID is in effect 13 based on their ectimate of their measurement errors.

14 Now, there are some other errors that can 17 feed in there, but I think it is the measurement errors is isn't it that determines their estimate of the LEID?

l 19 COMMISSIONER KENNEDY: Is hold up considered

,0 I

a measurement error?

,1 i

COMMISSIONER HENDRIE:

You are measuring i

7 l

3 quantities coming in here and how much is in this tank

4 and how much went down that pipe and how much went out u

this stream and you are either estimating or measuring n

i m ece6'isse.c=r _

age @

adfTt 't,

21 f

nacs se I

i f'

and each of those estimates and measurements have some error in it and you crank these in fairly standard formula assuming that they are all random in gausian l

A i

distributed and so on and come up with a standard l

5 j

deviation, I think.

So that whether a given licensee 6

for a given campaign has a large LEID or a small one, 7

i j

depends on what his estimate, primarily his measurement 3

errors are I think.

9 P

That is, if he thinks his measurements are 10 really great and have only a few percent error, when i

in fact they are lousy and have ten percent error, I:

I why then indeed you expect him to run over his LEID 13 at a greater rate than the one in twenty that the i

system is nominally shooting for.

g DR. LURIE:

Well, I will try to avoid the g4 g7, answer for very obvious reasons.

I do not really know

g what is going into each one of these components of 19 the error and I do not know how the licensee is

o computing it but if he exaggerates in some way or the 21 other then of course he is going to inflate the LEID

=

and as a result of it the ID will be short of the M

LEID, if he is going to be --

l 24 COMMISSIONER HENDRIE:

There is a clear i

2 incentive to produce a somewhat larger one, nroduce it

?

Iweemm% ',spesnee #ursreups !<

l I

=act we. 22 l

o

=

l l

on the large side of a reasonable band than the small side and you are less likely to run across it at any i

gium point.

Okay, why don't you go ahead, I just wanted i

4 to remind myself in effect --

{

t CHAIRMAN AHEARNE:

Well, the thing is I under-l 6

stand it he is not for example allowed to put in any i

I human error, or estimate human error to that.

At least I

that was the explanation that' was given by NMSS.

9 COMMISSIONER HENDRIE:

If a guy stials stuff I0 is that a human error or is that 5

If i

CHAIRMAN AHEARNE:

That is a different kind t-I of human error.

i COM'ISSIONER HENDRIE:

Oh, that is a --

4 M

l 14 CHAIRMAN AHEARNE:

But, as you know many 15 measurement errors have a certain human compliment.

14 COMMISSIONER HENDRIE:

Yes.

Onward.

I am 17 sorry to divert you.

is DR. LURIE:

The deficiencies that are 19 outlined here, six of them, are listed here.

The 20 first one, ID is questionable criterion for guarding 21 against' loss and diversion.

The problem I have here is do you really want to treat negative ID and positive ID in the same way

,4 or would you split your effort a little bit differently i

t e mminvene.m. h te

(

,.. ~,,..,

23 nas ns and end up primarily against positive ID which usually I

reflect loss or diversion.

i The second part --

f 4

i COMMISSIONER HENDRIE:

Well, I would think j

l that if someone was sneaking strategic nuclear material l

5 into our supply lines, why we would want to know that, i

i I'

wouldn't we?

I 8

DR. LURIE:

Well, would you like to put 9

a probability statement on it?

I The second one is no uniformity in the 11 l

definition of LEID which was identified earlier for you, l

1 l

two definitions side-by-side, the ANSI and the CFR I

13 defnition.

The actual behavior of ID to LEID ratio is very different from most that is expected from the g

behavior of the ID based on the spirit of the definition.

g Why 375 out of 803 exceeded their LEID?

7, The next one, LEID-LIMITS are not based on statistical

[

jg

9 principles.

We cannot make statis tical statements

o any more.

21 Hypothesis testing procedure cannot make statements that no loss or diversion occurred.

There is no way for us to do it.

Just because we have a

4 small ID, short of an LEID or than the threshold does

~J not mean that there was no loss or diversion, e

lamannaw Yeseems h !<

aus e surTT e7

O 24 l

nez sc.

i I

And, finally, the LEID limit made it too high I

for as a rule they are larger than LEID and they will 2

result in high probability of undetected loss or l

l diversion.

l I

I Now, I am going to have an insert, since you 6

I were kind enough to listen to me without interruption, 7

or without too much interruption.

I think I will have I

time to introduce three more slides.

I think I have 9

i some extra copies of the insert.

This will divert us to away from nuclear material accounting and probably set it j

up a stage for some of the ocrests to be introduced.

l 12 i

This is the parking problem.

All right.

Here i

13 l

is the problem.

A commuter comes to work every day f

is he has to pay 15C per hour.

He comes at seven and leaves 1.!

at five.

His cost for last feeding his meter he is 14 going to pay five dollar fine and we make an assumption that the parking attendant reads the meter exactly i

la once a day.

If any different strategy, I am not i

addressing it.

,0 t

The distribution of time of reading is given

,3 s

u below.

Suppose it is the way that he is operating.

=

i g

He is going to come let's say, he never comes before

4 nine o' clock.

The probability that he will come between u

one and two o' clock is 25%.

25% of the time he visits i

leve vene.m. %-

w.

emWWTt.C.a N N,te sytTT 't?

_.-&L mm

o

=

nas8c. L i

l 6

l I

will be between one and two.

Somebody observed him and I

plotted this, now he is going to make a decision i

2 accordingly.

l A

I Okay, let us go to the next slide.

The I

4 commuter is going to find out some time of the time B

6 in which he is going to fill the meter.

Let us take i

7 l

any time here, take the one that is underlined, or the s

i one in white.

All right, at ten o' clock, the probability, 9

i and this is even from the first chart, it is not to terribly clear there.

The probability of coming to 1

11 the meter at ten o' clock and getting a ticket by this 10 time is only 1%.

If he is coming and there is no 13 ticket he is going to put $1.05 in the meter.

i 14 Now, therefore he is going to pay either

$1.05 or $5.00 the cost is one of the two, either to

$1.05 or $5.00.

However, you have an expected cost.

The expected cost of $1.09.

This is a game l

maybe if you played it continuously continued to pay at ten o' clock, on the average you only pay $1.09.

3 21 M st of the time, you will not be caught.

And,'if you j

look at the right column you will find out that

$1.09 is the lowest number that is going to pay on the average, l

4 This of course is according to expected costs.

The l

2 entire thing is going to be a strategy set and we are o

temsm.mmavwemme h f4 me umam surft 'n

nca sc 24 I

going to the next slide to show the concept relative l

t to this.

1 l

His objective is to minimize expected cost, j

4 I

he cannot minimize a particular time or cost at a given l

time, but on the average he knows that he can play this j

6 game and that there will be more time if he plans to i

I 7

t fill the next meter at this time.

Decision variable, j

l 1

i is time for filling the parking meter.

The strategy, i

9 I

l to set up all the options that are available to him.

to He lines up all the things that are available to him, i

11 i

all the lines and he is going to find the optimum 10 l

strategy which will tell him what he has to follow.

I 13 Now, I want to remind you this is a very is i

simple example and there are many things that I am not 12 taking into consideration like probability that he will be in a meeting and he won't be able to feed the t,o meter on time or that he will break his leg on the la i

1 way running to fill the meter or the probability that other things could happen at the same time.

3 1

g Let us go to the next slide and we will see I

=

that.

Again, the same concepts that we c=nsi&nedearlier l

l n

can be shown here.

So, the objective here in the game j

4 is again to maximize some kind of' expected cost or l

2 penalty to defender.

Defender is NRC at this point.

\\

w.

~=

x

. - u -

,-a t.

un o

27 o

e nez we What is the cost?

We are going to give you i

I an equatim, a mathematical equatim in which all i

I 3

cost and benefit consideration will be spelled out.

{

What are the decision variables.

Number one, I

what is going to be my alarm threshold before it was l

I LEID limit or 2 LEID, or 2 LEID limits, and the second 1

7 decision variable is going to tell you what you are going 3

to do once you found out that you exceeded this alarm 9

threshold.

Well, here it is called the estimator of the amount diverted.

11 The strategy in this case is going to be 1:

the set of all combinations of alarm thresholds because l

13 l

we can consider all of them and also that associated l

ta 3

i estimator amount diverted.

What would be my strategy?

15 i

My optimum one?

Either one that would give me a best to combination of alarm threshold and associated estimators 17 of amount diverted.

18 i

Now, since we have a game, we are going to i

19 play it.

Now, what are the rules of the game?

Suppose

,0 1

that we set the alarm at a given level, call it "A".

l 21

=

Three moves are going to be played here.

The diverter is going to remove X Kg of Special 4

Nuclear Material, u

The Defender takes inventory and l

t t

i i

bresp.rcs..

dese.c= w

=c as tuu'% C.Mu'*CE. f**tE* S

• Suf?T 'O

8

=

nas ns i

.I I

I what it finds out on the ID, that is what is given to j

2 him.

Now, he is going to take the knowledge of ID and I

is he going to act as if y kg of Special Nuclear Material A

has been diverted.

Notice that one ID is not necessarily r

I the same.

He is going the route for declaring that I

I a certain amount of nuclear material was diverted based l

I on this ID consideration.

This y variable is going to 3

give you an index of intensity of search, it just acts t

9 as if so much material was diverted.

i 10 i

This is in game theo~ry often called the 11 estimator or estimators of amount diverted.

I:

All right, the criterion for decision is 13 given in a cost function.

What is the penalty to 14 defender.

It will cost him for the clean out of the l

12 inventory, shutdown and clean out inventory, especially i

to if their strategy tells them they had to shutdcwn it 17 i

will cost him the cost of the material X kilograms that 18 he did lose, it does cost him that much.

It does cause l

19 I

him to search for Y kilogram of material.

You will remember that although X was l

,1 4

diverted we did not know what X is the Y ID was given f

he is going to look for Y kilograms of materials.

It is going to cost him so much to find Y 3

2 kilograms of material to search for it.

f

'eaNas desearse h x j

m e eam stuurt, s. n wr?T s?

U

=

29 i

nez sc s

i i

i i

Finally, there will be some penalties some 2

final failures and things like that.

This is going 2

i to be the cost due to wrong assessment of their i

4 their mild diverted, the underestimated or overestimated 3

is going to cost the company or going to cost NRC or f

I t

6 depends on where it is.

But he is going to gain something:

I if he plays his game because utility, or he may gain I

I i

something because utility diverted from the recovery e

of materials have some value.

i l'

10 CHAIRMAN AHEARNE:

Does that assume in your 11 l

game strategy that the material is to recover?

i:

DR. LURIE:

It may or may not be recovered, is no.

This may be zero, but if he recovers then he j

14 t

has to add to consider it in his costfacter.

12 In order to understand it a little bit better I4 I think we will have a once through in one of the 17 i

facilities and we will see how decision is made.

We l

18 have a generic facility, a small platonium facility, l

19 I

this is a generic one, where sigma is equal to 300 grams, 20 t

.3 kilograms, and if you play this game, we find out i

,1 i

that the diverter is a reasonable diverter, he can play optimumly and we are going to look right now on 4

the right hand side to find out what is going to be 3

the penalty to the defender.

Let me go through one of a

iwre maecs.es % r==. ra

==c.

de sDUT'e Ca*=*Gn. ST*G*. S e MITT '91 t

30 4

nez sc.

I these plans.

Supposed we are going to take the white I

line again.

Suppose the loss threshold, is set at l

5

.57 kilograms.

If I set it at this point, I am going to look at the ID, if the ID is going to be smaller f

I than 570 grams, I am going to search for 125 grams, l

6

.125 kilograms.

If the ID is larger than the threshold e

of.57 kilograms, I am going to search for.975 of the I

kilograms.

This comes out as a mathematical, analytical l

9 solution to the game.

If I am playing the game, my 10 maximum penalty is going to be to the defender is going i

l 11 to be 7.4 kilograms this is now actually if you compare 12 it to 7.4 kilograms to' dollars, this would be your 13 penalty.

Multiply 7.4 by the cost of one affected 14 kilogram.

13 COMMISSIONER KENNEDY:

What is the basis for id estimating Y?

17 DR. LURIE:

Everything is analytical i

18 i

considerations and you go through a lot of matrix I9 i

inversionpomputer manipulations and many, many hours

,0 of calculations.

l This is based primarily on going by trial and error.

Suppose that you set the alarm threshold at.57 and.you are going to play with different kind 3

of numbers.

Now, you are going to say if there is a o

iarews.ario.as vu. ear w #N,%

G

O 31 l

nez sc.

l-1 l

i diverter, if there is one, and he is an intelligent

{

B 2

diverter, he has a way of playing the game in such a i

2 way that it will cause the defender the maximum amount of penalty.

l 4

3 If this is the case, we are going to go 1

0 around and find different Y values that are available i

I to use and find out which one of this Y will minimize l

3 depending on the right hand side and this is going 7

i to be the solution to the game.

3 Right now we have three numbers in this case.

11 i

.125,.975 and 7.4 this is a solution to the game.

The I':

solution to what would be the best for the defender 13 i

assuming that again a diverter maybe somewhere in the 1

is facility.

12 CHAIRMAN AHEARNE:

To what extent is it driven I

I4 by the assumed values of the terms such as the cost 37 i

due to wrong assessment?

j 13 DR. LURIE:

All right. This is part of the l

19 problem with the present study.

This will be identified later on.

An arbitrary decision was made by the,I l

,1 baLiese wrong assessment of ID is going to be fifty times that of the amount of the divergent or a hundred times.

Two statisticians were sitting down and were 3

l trying to find a reasonable amouct to assign to this l

8 I

wrum.v o Vipear = at-sem M l

m manw surft -er

o 32 I

sacz sc I

very cost function.

l t

CHAIRMAN AHEARNE:

Does it also assume a l

i significant relationship between the value of Y and 4

the cost of the search for Y?

i I

DR. LURIE:

Yes, it does.

l 6

I have another slide, if you want me to show l

l I

it to you later on that you have the value of Y and C x Y is given.

The more ydu have to search for it I

the more it is going to cost you.

Now, look at again at this chart.

Set the aLust ti threshold to a different level.

Let's say.90.

You l

i find out that the maximum expected penalty to l

W defender is g. ing to be 10.1, somewhat larger than 7.4.

o 14 l

Now, what this chart gives you is not only i

13 what the defender ought to do once he sets an alarm j

16 threshold but what possible threshold that he can select 17 I

for himself is one that is going to give you the minimum is t

of all the maximum that expects penalty to the defender.

19 I

It is a MINI-MAX property.

You are going to look for

0 the right count and select the best one that will give I

1 you minimum penalties.

Exactly what we did in the parking example.

We follow the white line and we

+

decided to fill the meter at ten o' clock.

This will tell me for this kind of facility, assuming that all n

wrunu v o m vgree. m. acmcvss '%

f ass e Ref?T 87

4 33 l

paca sc I

cost parameters are correct, my best bet is to set I

I my alarm threshold at 1.9 stigma or.57 kilograms.

l 2

I will take it one step further.

Up to this point, we assumed that the diverter does know what is e

alarm threshold, it puts him in a big advantage, because 6

if you know what is alarm threshold, you know exactly i

i what strategy to follow.

g If I am a thief and I am trying to steal 9

something from one store, if I know what time they close I

10 the store, if I know what kind of mechanism is going to 11 trigger the alarm, I am going to do a better job than 1*

if I just played at randcm.

13 i

The question can be denying this information, j

14 and how do we do it?

14 l

All right, one way that you can do it is just I

I4 have an invente:y at randcm, but that is not practical, 17 l

as a matter of fact, I would not be surprised if you i

is r

had to prepare for an inventory.

Another way we are 19 doing it is keeping a secret.

How long can you keep

,0 t

it a secret and the third one is play the alarm at rardcm.

g It doesn't take you any extra effort from what you have 3

done up to this point.

The cost function is the same.

l

4

~The estbuticr.s are the same but now you are going to a

improve your gama because you have denied the diverter i

l

.my ro.

va m. h x l

me anstw casma. rmust s. e. wrre u 1

m i s c,

34 r

O I

• *c8 'c

\\

I from one important bit of information.

He doesn't know I

about threshold, now he is going to steal and after the i

ids collected the defender is going to decide whether 4

to alarm or not.

He is going to use a random mechanism 2

i once again he is not just tossing a coin or throwing 4

a dye.

There is a method to performing it and if you 7

go this route, you are going to rabce the penalty to 3

i f

i 2.5 kilogram.

You remember the optimum was 7.4, I have 9

i reduced it by a factor of three just by denying this important information to the diverter.

L 11 i

I Next slide please.

Deficiencies.

The 1:

i present formulation of the cost function does not appear l

to account for all eventualities.

I do not know if

?

i anywhere it would be spelled out what would be the g

penalty to us if enough material diverted made into g

7 a bomb and dropped over a populated area.

There are ig many other things that have to be considered.

I Finally, input into the cost function from 19 I

go decision makers and experts in relevant discipline is J

1 lacking.

I l

I have all the respect for the two mathemati-i cians who sat down and performed this function, they 24 did a tremendous study, however, the input from the 2

experts or people who are really familiar with all the e

=vessu% vapeew 0 i

m.

35

=

psas nc.

f i'

consequences is lacking.

Decision makers have to be I

invdved.

Indeed, the study is incomplete.

Deficiencies 2

l could be alleviated by further study.

4 i

The concluding part is going to compare the l

e two methods giving the advantage of the game theory l

6 over the present practices, the disadvantages of the 7

I game theory approach over the present practices and j

3 instead of giving my recommendation I would like to i

9 of fer the peer review recommendation of i

10 NUREG 0950.

i 11 l

Next slide please.

The advantages.

They ought to explicitly consider the possibility of a 13 diverter.

It is written down in the model.

We take it 1s into consideration.

Th2 classical approach does not mention anything about diverter.

Here there is a model, this is one of the so possibilities, j

g, r

9 Advantage number two, cost benefit consideration

.g The strategy of the defender is based on utility

1 considerations.

Number three the alarm thresholds are derived from optimality criteria rather than by arbitrary l

4 decision.

This is not two times LEID limit, but you "J

go to a chart, you go through a mathematical technique e

imm vasear w ac-smpe, x f

de sudT% CAF*th FPWWET. L e h tft 87

pea 36 t

o

=

nor we.

}

i i

and you find out what is best for you, thhre is more arbitrainess in this case.

l Finally, for the appropriate model, maximum I

a f

assurance is achieved.

This is what we have shown in the two charts ago, where you are going to the right f

8 column and you pick out the one that will give you more l

7 t

i benefit on the average.

s t

l The disadvantages have to be said in here too.

9 There is dif ficulty in developing a cost function.

It i

10 l

cannot be devloped without input from decision making.

I 11 l

It is a tremendous task to write a good cost i

function and to accomodate for as many possibilities 13 or eventualities that you want to consider.

3, Decision makers must come into the game.

g Continued subdating of the cost function and its g7 parameters are necessary for maximum effectiveness, is even if you got a model, tomorrow the model will change, i

1 because tomorrow the cost of plutonium is going to 19

o change, tomorrow the cost of search will change and 21 other things will-change to.

l The Game Theory approach requires more informa-i

3 tion in its application than the classical approach 6

24 and therefore requires more resources, no doubt about

~J it.

The present approach is the very minimum.

o twwn % vese e e posrreos saic.

s g

ass M M N. E e RdfPE '97

• =

noa ec.37 i

i I

I The generic requirement for regulation l

I concerning game theory approach are more difficult to l

i 2

provide than for classical approach and this is a i

problem.

Again, as I told you, instead of offering my recommendation, here are recommendations given by l

the peer review group.

I think that number one is the most important one and NRC takes steps to improve gametheorydevelopmentapplicabletomaterialaccounting.l i

9 Whatever has been done up to now is not 10 sufficient.

Number two, after the presumed successful i

11 l

l development, NRC take steps to implement the game theory 10 i

as a decision making tool.

This again depends on i

13 number one.

Is c

Number three is not really relevant directly to the game theory.

It has some recommendations relative to to the rule of three and five kilogram coming from 17 NUREG 0450 and as I say this should not be implemented i

is t

at this point until more information about cost benefit f

analysis are available.

t Number four, research to be initiated on use of game theory to multiperiod data and this again is very complicated.

If you do not go for number one,

4 we cannot go for number four.

a Number five, research and development to be n

y

'T.4%~,3M'.7

38 9

naar sc.

i I

conducted for discipline outside of Nuclear Material I

Accounting.

I guess I got carried away.

We got very 2

excited with OGC findings, and that is all.

COMMISSIONER KENNEDY:

Very good.

CHAIRMAN AHEARNE:

Joe, this was at your l

request.

i 7

I COMMISSIONER HENDRIE:

Yes, what is the current 3

status of the game theory in development effort?

l DR. LURIE:

All right, I have learned --

f 10 COMMISSIONER HENDRIE:

Sid told me when I talked to him back the end of last year and I have now 1:

I gotten and neglected to bring the appropriate notes.

13

{

So, if somebody can bring us all up-to-date on it and i

11 is it under research or is it under technical assessment 13 divsion of safeguards?

f DR. LURIE:

The information that I have is for fiscal year of '81, 125 KR allotted for each.

For 18

'82, 75K and '83, 25K, I do not know how they are working this.

,0 4

\\

COMMISSIONER HENDRIE:

What is the current --

3

=

CHAIRMAN AHEARNE:

Kevin --

i

(

MR. CORNELL:

Those numbers are correct.

l 3

There is nothing that I am aware of in

'80.

l l

3 MR. DAVIS:

It is zero in

'80.

We went to i

o i

1

)

te% <ese r w #wm %

I C

3 sacz m 39 research and asked for help in this.

We asked that l

I I

if funds were available for 80, but other things i

d had priority and did not see those funds for '80 and 4

will be allocated in '81.

I a

CHAIRMAN AHEARNE:

It is funded in research?

I

~

6 MR. DAVIS:

Yes, sir.

f i

7 COMMISSIONER HENDRIE:

Does that mean --

3 j

where was it, Sandia that the work was going one or 9

I is going on?

l l

10 DR. LURIE:

Berkeley I believe is the one.

It l

COMMISSIONER HENDRIE:

Does zero fuhding i:

1 in '80 mean that we lost that group?

l 13 i

DR. LURIE:

Would you like to answer this?

Is i

MR. CORNELL:

About Berkeley, I believe they I3 have not reached their decision as to where this 125k i4 would be spent and it is possible that it will be --

COMMISSIONER KENNEDY:

But if it is zero 14 i

in '80 --

9 MR. DAVIS

Can I come up for this, I have

.g a little information for this.

Besides the money that we have had for game theory, for '76, 145,000,

'77, 130,000,

'78, 56,000,

4

'79, 10,000 and '78 and '79 were peer group monies.

2 That #or '80 zero, because of priorities, and we have o

e.co. 6 so.w = c m.-

l

~ ~ ~ ~ - -'

nez sc.

20 i

i a plan for '81 in research for 125, and '82 for research j

i 2

175 and '83 for research 25.

i I

t 2

MR. MOGLEWER:

I would like to make a correctionj i

4 for the the 76 and 77 for the numbers, were the contracts for Berkeley Laboratory but half that money went for 3

the development of the so-called MaClans model and 7

application as MUF simulation model and it was not game i

I a

theory work, I believe only half the funds for those l

9 I

first two years --

10 i

CHAIRMAN AHEARNE:

Could you answer also the question that Dr. Hendrie asked.

In absence of ta l

SI' funding is the Berkeley group dissipated?

I 13 MR.' MOGLEWER:

Essentially it is, strong 14 efforts would have to be made to reconstitute.

I 13 CHAIRMAN AHEARNE:

Joe, any questions?

l 16 COMMISSIONER HENDRIE:

Yes, I guess I would 17 comment without knowing quite where it leads me in terms is i

i of you know specific dollars and priorities anc the l

research budget and so on.

I would comment along the

,0 t

following line.

The regulations, the present practice

,1 in inventory difference matters when he inventories and when he notifies, when we blow the whistle on the top of the factory and all run counterclockwise and so 3

u on over an inventory difference appear to me to be 6

-o ne 6,ev=a - x ep 33WTie WM 97

o ucz sc JG f

i i

a plan for '81 in research for 125, and '82 for research I

175 and '83 for research 25.

l Iwouldliketomakeacorrectionf 3

MR. MOGLEWER:

forthenumbers,werethecontracts!

4 for the the 76 and 77 for Berkeley Laboratory but half that money went for 3

O the development of the so-called MaClans model and I

application as MUF simulation model and it was not game i

I theory work, I believe only h'alf the funds for those l

3 i

9 I

first two years --

l io CHAIRMAN AHEARNE:

Could you answer also it the question that Dr. Hendrie asked.

In absence of i:

i 80 funding is the Berkeley group dissipitad?

I 13 MR. MOGLEWER:

Essentially it is, strong 14 efforts would have to be made to reconstitute.

i3 CHAIRMAN AHEARNE:

Joe, any questions?

14 COMMISSIONER HENDRIE:

Yes, I guess I would 17 comment without knowing quite where it leads me in terms t

is i

t of you know specific dollars and priorities anc the 19 research budget and so on.

I would comment along the 1

following line.

The regulations, the present practice i

,1 in inventory difference matters when he inventories and when he notifies, when we blow'the whistle on the

~,

top of the factory and all run counterclockwise and so 3

a on over an inventory difference appear to me to be t

' m e m 't w ru am ? r.

de 33WT% ON N. L e 34t73 97 4

L z :. -

41 l

C 2

nos we I

i i

not very useful.

That is, we say well, we would like I

to know that all the materials there within some fairly f

precise estimate and we get that all set up for a i

2 i

i licensee and then we discover he is alarming left, right t

and sidewise everytime he does a campaign and he comes f

3 I

around and says I can't run the plant if I am on inventory!

7 92% of the operating hours, so we checked the limits a

and say right, I guess we had it set too low and that 9

I is not practical in view of your measurement of to uncertainties andone thing or another and that we 11 attempt to provide compensating measures in physical 12 l

security and in material control in the physical sense.

I 13 l

We are not quite ready, on the other hand, 14 to give up and say, never mind these inventories, just 14 use those for your bookkeeping purposes and we will 14 depend for material protection and control on the physical I7 I

security and material control efforts, we continue to l'

is i

I give some, at least, considerably more than lipservice i

I must say to the inventory difference in measurements

,c i

and what they mean.

But what they mean is very hard to say.

The licensee who handles his stuff goes through l

a campaign and comes out under his LEIDs and that does l

t 3

not raise a ripple then, but I think we are not prepared l

to stand uo and say that is a proof positive that no s

!=,w re. 6 eine,.c= h %

i, n h.edf\\nr%1Q

42 r

0 7 642 N C.

I l

material disappeared someplace.

On the other hand, he f

i goes over the limit and has to shut down and do l

hnencarias and we all run around and look hard, but when 4

we get all through with that exercise, once again we 8

are now unable to say two things.

We are unable to l

say that that material was diverted and we are unable l

7 to say the material was not diverted.

In the absence r

a of being unable to say one or the other why I sometimes wonder is the measure useful, is the exercise useful?

9 10 Now, the group theory approach appears to f

i 11 me to still have some difficulties in it.

The game l

i I:

theory approach.

i i

13 CHAIRMAN AHEARNE:

You are going down to la group theory approach.

I COMMISSIONER HENDRIE:

That may be next i

to years approach.

The game theory approach does have the dif ficulty how you arrange the game and how you come f

I3 out in it depends very much on how you write the cost function and it is not so clear how you ought to assess o

the values in all those elements of the cost function i

21 l

or indeed whether the current models have all the 1

terms in it that one really ought to have.

On the other hand, the game theory approach

4 does have the prospect that it doesn't set some t

t e co.%'s w m. h %

l

~~

a. n.,

..u-

4 _.

43 I

n.or ye i

i arbitrary level and say over five kilograms, apparently I

1 unacccounted for and we all panic, and under that j

everything is honky dory.

It sets a strategy for 4

trying to track a potential diverter which is clearly 1

3 better than having this highly mechanical sort of l

fixed threshold and certain things you do and limits 6

l i

7 adjustable o'n through put and so on.

That is, if there l

3 is a diverter, it appears to me that the game theory I

9 approach considerably improves the prospect that 10 inventory techniques and statistical techniques will 11 catch him or at least put people on the track.

12 I

Since that is the case, it appears to me i

13 l

that we have' sort of two things here in the inventory 14 field.

One of them is the system that we use now 12 to regulate under in part which appears to me to be id of not zero but diminishing usefulness in a practical 17 sense and another one which we are not implementing i

la i

i with the licensees which at least seems to me to improve i

19 whatever usefulness there may intrinsically be in inventory techniques.

f

,1 4

It remains of course a question whether one l

i l

should not simply throw this who.te body of inventory l

I I

l

4 matters, just set it aside and say good do your 2

bookkeeping and charge one another for material lost on o

.my cm 6 veneem

%-s== x l

essou m e.aarau.,cuur.s e marrt st

.- u -

~

44 O

O

==

paGg NC l

I that basis, but with regard to protection of special l

4 nuclear material we will just go back and depend on l

e I

physical security and physical control of the material l

l and people watching people and so an.

I Since I suspect we are not ready to do that, l

and since the difference between the present ID regime I

7 and the game theory regime is as I believe I

s i

i characterized it generally.

It seems to me that we 9

are really putting minimal resource into a technical to area which has at least the capability of getting II l

whatever there is to be gotten out of inventory matters II i

with regard to protecting special nuclear material.

l 13 1

l I am not sure them that we have all the priorities ta entirely straight in the safeguards area.

Now, the budget' is not all that large in the whole area, did somebody l

1,,

s guess what it is in the safeguards research?

{

MR. DAVIS:

We are talking about 12 million, g

I think in safeguards research and TNA.

.g gj COMMISSIONER HENDRIE:

So, out of 4.7 I am not sure the ratio ought to be 4.70 rather than I do not know, 4.55 and 150 or something like that.

4 It appears to me there is the prospect for 2

yield on research and game theory which is relatively i

1amss er cm.., vo wr w

< c.

et WWDe 4LAN fMEE*. 5. e Wafft 97 eammCDs. & & mm j

4 45 l

sact sc.

I i

i high compared to the other side of the business.

End of comment.

I COMMISSIONER KENNEDY:

Where does this stand 4

in the action sense right now?

Has there been a request 2

implementing peer reviews recommendation?

g 6

l I

DR. LURIE:

I don't --

i 7-COMMISSIONER HENDRIE:

I don't think we are s

l implementing the recommendations of 0950.

9 COMMISSIONER KENNEDY:

The theory -- there j

10 has not been any staff proposal that they should be 11 implemented.

MR. DAVIS:

What the staff proposes.

The 13 staff does not disagree --

CHAIRMAN AHEARNE:

When you speak of the

4 l

staff?

MR. DAVIS:

The elements of the stay j

g, l

g safeguard staff in their look at game theory migh' i

19 be measured in degrees of enthusiasm.

I wouldn't say

c that we have two groups.

One deadset against it i

21

' and one who Ehh* it' the answer to mankind's problems.

l I

There are degrees of enthusiasm.

U I believe the consensus of the staff would Id be that game theory does have potential as Dr. Hendrie stated, does have potential and should be looked at e

intumenoma6'tmenme h <

aED S2JThe CAN ffB4E*. E e SWITT '97 L

mn A o

46 4

sacz we.

I t

I think though that many of the staff would agree that i

the feasibility of the game theory approach in a I

2 practical actual safeguard situation has not yet been A

I demonstrated.

So, consequently what our research j

l noney is intended to do is to aim toward that demonstration i

6 move it from potential to actually some demonstration.

j 7

I l

s

' We have not approached on the basis of looking at all the peer groups recommendations and 9

l making a decision to implement them all or not Lmplement them all.

We think more information is needed to lead I

If us to some type of decision.

I s

So, I do think, as I say, that we are not 13 i

satisfied with the current material control and I

accounting.

34 We do have a major effort underway to upgrade that.

One of ou'r efforts at the present time since i

g7 ja a limit of error is critical to both game theory and t

79 to the classical approach is a significant outlook 20 and limit of error.

I would like to interject here, i

I 21 however, that as I understand it, the licensees at the present time are not totally free to determine this M

limit of error.

There are guidance out and of course l

3 what they have done is reviewed by NRC staff.

l 2

Hopefully, this makes it more rational.

i.,=

T

. v-,.

% c i

en gasTse CAM

• tin, ff*Er*, E e. Suf?T 'S7

.- 4. :. -.

4 47 l

maca *c.

CHAIRMAN AHEARNE:

Well, I think we saw the guidance, or at least part of it in 10 CFR 7051.

MR. DAVIS:

I think that there is as is qpite 4

common with our regulation, there is a body of guit e

J associated.

6 CHAIRMAN AHEARNE:

Yes, trpreting that I

I theo ry.

i i

l l

MR. DAVIS:

One other remck-I would 9

like to make, since we have been discussing this the to last couple of weeks.

There was a slide shown earlier It I

on, which says does inventory experience.

What does 12 i

inventory experience, show, which would indicate a l

13 i

somewhat gross situation with regard to ID versus LEID.

i 14 i

We looked at those numbers and think it may be not quite 12 i

that high, it is nothing to brag about --

CHAIRMAN AHEARNE:

What percentage would you say?

f la MR. DAVIS:

Maybe half of what that would 19 show.

Twenty to Twenty-five.

Well, like I said it

g 21 is nothing to brag about.

i CHAIRMAN AHEARNE:

Well, it is not only nothing 2:

to brag about, it is basically fundamentally

4 inconsistent with the definition of a limit of error.

'J MR. DAVIS:

Well, we won't get into argument l

n.o

. %,..= - x me uh,7%

htf1 't?

,,e 48 C

P AGE MC.

about limit of error.

Well, you know how we cut limit of error very finely without the human error.

t f

f CHAIRMAN AHEARNE:

Yes, which is the fundamental!

4

.i inconsistency I think would --

MR. DAVIS:

Well, I would like to emphasize that the staff for sometime has been workinc 7

towards i

t l

a relock at the classical approach and we are somewhat j

s i

l l

optimistic -- very optimistic that by changing some r

9 method of the classical approach we will come to a l

better control of the situation.

Il COMMISSIONER HENDRIE:

But e,ven if the 12 i

I improvements are likely to be in the direction of 13 recognizing some --

i MR. DAVIS:

More timely basis.

gy g3 CRAIRMAN AHEARNE:

Putting a larger --

g7 COMMISSIONER HENDRIE:

You are going to i

18 rationalize some human error and you are going to l

I

9 rationalize bookkeeping --

o CHAIRMAN AHEARNE:

Well, there is another 21 word, I was going to say you look at data and ycu look C

at the way systems really operate and you reach a C

conclusion here is what the error is.

I' MR. DAVIS:

Well, let me make a final comment.

-e We agree with the comment of the briefer that we don't I

e

-ve r e r

m- !<

- n un,r nr s. urn n m ur

I o

nas ss so I

consider these to be competitive cysts s but complimentary :

I systems and the degree of disagreement really is I

competition for the resources that go into the two 4

differences.

3 I

COMMISSIONER BRADFORD:

Except that you could 6

i not enforce against both the systems simultaneously.

t 7

l MR. DAVIS:

One of the difficulties I 3

personally have with what I see as game theory is 9

i fitting it into my somewhat traditional mind of j

10 regulation, particularly the aspect of shifting limits i

i It i

and unrevealed 1imits and this type of thing.

t:

i I would have a very difficult time and thus I

13 suspect of some members sitting at the table here and have a really difficult time of explaining why we react one way to a situation and a different way to the :same situation of another licensee, which is l'.

possible as I understand it with this approach.

l t

l g9 COMMISSIONER HENDRIE:

Peter, I think if

g you are leeping for a minute or two a long way beyond l

21 where we are and maybe beyond where we will ever get.

I If one attempted to regulate on a game theory basis, you wouldn't try to implement ir to write in the 4

regulations all of these mathematical details.

What

• 3 you would say is material is to be protected by a

% % wn.. - %

as e Cap *3. JT*W'*. i e 1WrTT '97

I nas sc 50 i

combination of inventory techniques, physical security f

techniques and physical material control techniques and i

that the inventory techniques would involve for each l

l 4

licensee an approved statistical approved by the 1

Commission staff's statistical i

6 each analysis and then for licensee why you would 7

set up a game theory approach.

That is, you would not a

I have then for a given licensee a fixed so many kilograns 9

on an inventory or apparent mismatch and you shut down to or whatever you institute for that licensee this kind of regime without spelling all of that in the regulation. ;

You would not have ids in the same sense.

13 l

g j

COMMISSIONER BRADFORD:

It is going to be legally pretty touch but I don't think we want to

3 g

pursue that any further.

7 CHAIRMAN AHEARNE

My only comment would be 13 that I would appreciate Norm if you can get together 19 with NMSS and sometime in the near fuhE2 at least 20 give me a paper talking about the statistical treatment

• 1 whether one can make it at least a valid statistical i

treatment, the concept that we have because the way U

I end up from listening to your briefing ic that the 4

current system is terrible and that a revised system needs p lot of work before it can be useful, we have 1

.wuss.aro.6,o=eanu *troo m.%

& W LM*'*'3%

%tT1 -91

W

- o e

uca 3c. 51

~

I got to do something in between.

I COMMISSIONER HENDRIE:

And it is not so clear a

3 to me that the work for some improvements so that the l

game theory improvements, so that it is necessarily e

the best distribution of research funds in view of 4

the relatively small amount we are talking about on 7

the game theory to let it continue on through the balance of this year to zero.

Now, we are already 9

l well into fiscal '80 but you might begin to pull j

10 t

people together at some appropriate center of excellence on the basis of a relatively few dollars 10 in hand for the balance of '80 and knowing that in '81 13 i

you were going to put what a hundred and a quarter i

ts or something like that --

13 CHAIRMAN AHEARNE:

Perhaps EDO could look 14 at that.

And would you see if we can at least get on a sound statistical basis.

I 13 t

l MR. HALLER:

Yes, we would.

19 CHAIRMAN AHEARNE:

Thank you very much.

DR. MOORE:

I think we can get it on a sound l

21

=

statistical basis by getting the fundamental medels

=

and the fundamental concepts right the first time.

4 Now, if it turns out that in practice we J

miss it will be because the practice missed, not because t

>=rne

% 'sm e

• g n -m. -,..,

O O

7&435

.*tC.

s i

I the model missed.

That is a very important point and l

I it must be made.

i r

t CHAIRMAN AHEARNE:

Right, thank you.

There

}

1 i

i ought tio be a shifting of personnel I believe.

I 2

(Whereupon the meeting 6

on the above matter adjourned ;

7 i

at 11:22.)

I 3

7 i

10 i

11 l

i I:

I i

l 13 I

i' 14 13 i

?

t 17 i

i 18

},

19

o i

21 i

i l

4

's i

.1 e

e.m,...',- ~

i

-[

j e m rom. n. m.m o

.- 2 :.

,