ML19207C153

Data & Licensing of Nuclear Power Reactors:Overview,Some Philosophy & Example

ML19207C153
Person / Time
Issue date:	08/14/1979
From:	Abramson L, Moore R NRC OFFICE OF MANAGEMENT AND PROGRAM ANALYSIS (MPA)
To:
References
NUDOCS 7909060501
Download: ML19207C153 (28)

Text

.

INTERIM REPOR T Accetuon No.

c,ntuctrror m or rroject

Title:

" Data and the Licensing of fluclear Power React. ors:

An Overview, Some Philosophy, and an Exarple"

',. ct of th; Document:

Saoe as Title, ry;

,f Docun <nt:

Confercnce Paper 139th ?nnual is.eting of the Ir.orican Statistical Association August 13 thru 16,19/9 a ther(s):

u Rcger 11. I' core & Lee R. Abramson 0 2te of Ceca <nt:

3/14/79

,ocsn > mc !ca.; 1s. t. > :. mc of fi< e or oit, iun:

Roger 11. l'oore, Chief ASB, MPA This document.ms pr ep.2rei primarily far preliminot y or internal ute. It has not

..sj full review and approval. Since there nuy ba substantive chang.3, this reo dei ument shoulJ not be consideral final.

Prepare <f for U.S. Nuclear Regulat..ry Commission

// < hington, D.C, 2C555 19gqdg iN r s R e., R eeo R T

/

.B 10 lesearci and Tecinica N Assistance Report

.k er r

DATA AND THE LICENSING OF NUCLEAR POWER REACTORS :

AN OVERVIEW, SOME PHILOSOPHY, AND AN EXAMPLE by Roger H. Moore and Lee R. Abramson Applied Statistics Branch Division of Technical Support Office of Management and Program Analysis U. S. Nuclear Regulatory Commission Washington, DC 20555

[ Prepared for presentation at an Invited Paper Session, " Energy Data:

Availability, Utilization, and Analysis with Application to the Nuclear Power Industry,"

arranged by the Section on Physical and Engineering Science at the 139th Annual Meeting of the American Statistical Asso-ciation, Washington, DC, August 13-16, 1979.]

Abstract This paper:

-- Outlines the process used by the U. S. Nuclear Regulatory Commission to license the operation of nuclear power reactors.

Considers certain aspects of the role of data in the licensing process.

Illustrates these considerations with a description of a problem involving the estimation of the frequency of airplane crashes into a specific plant site.

I.

PRELIMINARIES The Nuclear Regulatory Commission (NRC) has many responsibilities, codified in the Code of Federal Regulations (particularly 10 CFR), regarding the civilian use of nuclear materials in the United States.

Perhaps chief among these responsibilities is the licensing of nuclear reactors for the generation of electrical power.

Before the NRC will license a nuclear power plant, it requires assurances with respect to four major concerns:

Public Health and Safety

-- Environmental Protection kkh kgggg{gj gg{ Tggigjca

-- Materials and Plant Protection (Safeguards)

ASSISI 8DC0 Repor[

Conformity with Antitrust Laws.

NOTICE This paper represents the views and opinions of the authors' it is not officially expositive of the Commission's position and not binding on the Commission.

o s v

,o s.# A U_.d o

. Vast quantities of information -- written and verbal, objective and subjective, beguilingly acceptable and recognizably enigmatic -- are generated in the process of assuring the Commission and its staff that a particular license should be granted. As used in this paper, the word " data" refers to those elenents of information that are expressed quantitatively by their very nature or are capable of being so expressed by a suitable process.

Thus it is inevitable that statis-tics -- the science / art of the treatment of data -- is drawn into the arena of nuclear reactor licensing.

Wherever statistical concepts are invoked, it is incumbent upon statisticians to participate.

To show that nuclear power plant licensing is of considerably more than academic interest, consider Figure 1.

Superimposed on the U.S. map are the sites of almost 200 nuclear reactors, either now in operation or expected to be in operation before the end of the century. The partitions of the 48 states designate the geographic areas that are the responsibility of the NRC's five regional offices.

II.

NUCLEAR POWER REACTOR LICENSING -- A OUICK LOOK Any licensing process can be made as simple or as complex as the concerned parties' interests, needs, and involvement require.

Nuclear power reactor licensing is no exception, but it does have certain terminology and features whose identification serve as clarifying guideposts:

-- A nuclear power plant is a group of one or more reactors located at a single site.

[Among other terms used for plant or site are facility and station.]

Each reactor is a unit.

[Nonnuclear units also exist on some sites.]

-- Plants often are named to reflect geographical [e.g., Calvert Cliffs (MD),

Haddam Neck (CN)] or eponymous [e.g., Duane Arnold (IA), Cook (MI)]

significance.

-- Units at a plant with several existing or planned-for units usually are designated by a sequence number [e.g., Point Beach 1 and Point Beach 2 (WI)].

Initial licensing of a nuclear power plant has two phases:

Construction Pemit and Operating License.

Construction Permits and Operating Licenses are subject to amendment throughout the construction phase and the operating life of the plant.

-- Each phase has its peculiar elements and compositions of technical reviews and public hearings, all of them aimed at providing the Commission with the information needed to satisfy its responsibilities.

To fix some of these ideas, consider Figure 2 which displays the " parallel tracks" that are typically followed before a construction permit (CP) is issued.

Here, an application is the request to construct one or more units.

~p @,'f

[;

/ S 'i hll M.nf.39 r e' < > o[ i a n l

],1 L,

i_

>l!j

- ' :I

.t~

.i bi. 1 N tJ t "

)

e 4

u<

c c.

f 4

ukjb50.N D i

a j

3, C

g 5

c 5

en 8

4 M} 4 f

Y w

4

$o s

ig t-c 4

w 7

g; S 45 RO O

'4 El w

A i--

i

)

4

~

4

• e 4

4

• g

~

4 2

m 4

g s

j j

5 z

2 4

=-

H3 r

.s"14 b

M 4

mi 1

4 LA.1 3

4 O

e i

a w

LY.

s m

W Q

e

<s Q

G

'1 r

c O

s 2

i c:C C u:C

[

/gv I

4 a

4 4

EE i

• /a l

s o *:: c::

r/

IM 44

)

c::

C.D ul LA.1 3

es o

O A

a :

M o

o i

C g

C w

I s

i O

i E

D l

1

[2 i

a E

L8

.I '

1 cn E

an=I l

4 tCC li c

1 M

k m

W 3

N s

g N

4 i $

o

?

5 i

5 Q

b b$ k b

kkkl 4

\\ *e A

2

+

fb $ $

ss s 4 e$

0 t b' 'Ija N

~

n

Figure 2 - PARALLEL TRACKS IN CONSTRUCTION PERMIT REVIEW PROCESS Public APPLICATION ACRS Hearing Tendered Docl<eted Letter

{

{

SAFETY 0

0 0

0 0

SER Suppl.

SER LWA Decision APPLICATION ENVIRONMENT Tendered Docketed DES FES CP 1

AND SITE O

O O

Or a

Decision SUITABILITY I

Public Hearing ANTITRUST e

Early Dept. of f

A Filing Justice Public Findin9 Hearing f,g f.2

'{'

Abbreviations:

C SER - Safety Evaluation Report ACRS Advisory Committee on Reactor Safeguards DES - Draf t Environmental Statement LWA Limited Work Authorization FES Final Environmental Statement i n <,s

Another way of looking at the construction permit review process is displayed in Figure 3 which delineates the interrelationships among the various segments of the t'RC organization.

For example, the Advisory Committee on Reactor Safe-guards must be satisfied by both the applicant and the NRC staff analyses be-fore the NRC Commissioners become directly involved in the technical aspects of an application. Moreover, the figure emphasizes the stages at which state and local officials and the general public are notified of findings and parti-cipate in hearings.

Once construction of a unit is well under way and as its completion date becomes more certain, the applicant seeks an operating license (OL).

Without an OL, the unit cannot legally be loaded with nuclear fuel and cannot generate electricity.

An important difference between the construction permit process and the operating license process is that there is no requirement for a mandatory public hearing in the OL activity -- but intervention by interested parties can and does occur at the OL stage and now is considered a routine part of reactor licensing.

Many issues involving data and their analyses arise at almost every juncture of the licensing process. Such topics as environmental ef fects, earthquake frequen-cies, meteorological phenomena, engineering safety margins, equipment reliability,'

and human error concerns continually occupy those involved in the granting of a license.

It is, therefore, important to recognize and to cope with some of the fundamental data-related issues in reactor licensing, and that is the subject of the next section.

III.

DATA AND NUCLEAR POWER REACTOR LICENSING If there is a god / goddess of fickleness in nuclear power reactor licensing, his/her name is " DATA".

He/She promises so much -- but she/he often delivers so little. He/She can illuminate deep problems -- but she/he can confound experts on both sides of a question. He/She can permit a flash of under-standing, offer a show of resolution -- but she/he can cloud with inconsis-tencies, darken with anomolies, shatter with irregularities, drown with a bnornal i ties.

He/She can gracefully support the framework of carefully con-structed licensing decisions -- but she/he can suddenly and effectively destroy the entire structure with the flick of a single event.

How, then, are data to be treated in dealing with reactor licensing issues?

The not-too-surprising answer:

Very carefully.

The nature of statistical problems in nuclear regulation has been described by Moore and Easterling [1978] as follows.

Highly visible Federal and State governments require open discussions... issues stimulate controversy.

-- Technically / emotionally intertwined Proponents, opponents, and regulators all tend to " harden" previous positions..

difficulty in changing regulatory decisions -- no matter what subject or who prcposes a change.

P001 OREL

==

- f-

@w, s.-

e%n ex

//

%N'~ 9 \\

• ci: :

C'C '?

C1ff ^ 4 h

\\p" k 4 5= [l d % d "N.

Yl!Il1

' -a..

5.p,

-

1 2 'r a "

[r$"^-&W- -*

= i.1.!- < ~ i !.

7:

FFe A _

I E I E 3' d. r t W --

'4*i>E23 E

• • • {g, _5

~E72>

d 7 M(,.e.W O %

2 3

ge g

V :; i)

_ tg3 7,

-==2 i 13 E.I I $

Mi#-

9,4 Wi f5E'5 7" M

%=lai' r

"3 im!

L f re -! ; : r 5-A 2 epi G( k a~3 4C e*s:n e

r t s : - = ;.: g Ug l p'O.u-5 e;74*

15 U 2 ;' !! j ;,w C @eM!

C 5 A i~.

4ii3!!!

3 A'. !

j hi[

h.

<i ris ih; m=a.

1.

+ $

:., $ ~! ?

i t Jead '

EWC "p

W9 S

Y i

w"~'g 5 b= =?!i s!!! 4 Lli

1. 3 ;[M2Ss'M<e e[

N M

• • E99 1

~

s M

M= m_ l re m s

< s t-

\\ '

a&

c

_e f

we Ema t

i a

m v

% y*=m' w m @p.p 0:1 we m2m ier m

&gg

^m gj= s'3 2,5 2 j

% m-g..

c C, b

n

[

E!iE !!ll'!!,9vEI __

[. 4f

=:

r_

rd =

3 MN WWIl p$x$CD

_i

-u! w !!, i d, i G #gH r,=--

g 1!!!

fii ne e8=

E F$ i 5 2 E i Y i f 5. 5 $f:'E'! mM-M'*'. s.: p

. ws nm a 'E{

L[.hM 4

a:t :-

==st n-g a.

i. ~ f $ h b.$ h$ e h yI 5 S

Nhhh L

G

Hil$lifill 4!di C@ PM fM W s O

= ; ;- :i s i s ?, ! i l l s G i l t i 6

%e' bf.'s 257-I b s s. i :} -j [ e ::i ; csi NM

{CG p=a,i

-+ "I g

7s Eg,

t=

e 3

9 Q

5;g y

==.

g O

O 115 NA dNNI DdMM &.- m,h. fe k d t

'l C. s %'h"e(#_

.# g

. J,.- c af a

4

.._2 5EE=

4

^Qh E' a Q

, g%~tB C/)

E;3 ' ' i ~ E

s. &

?

h],E 3,.

7 D,K D.-,

eA g::g =. _ _ :

g l- $ h. 5 I 9f

<.D

-/N y

4 e

p bf b

(~b, d

l-a a

m, _c i a=se2..-

WM[ema 5

s w-w A'M a$

q tjl,II g $1:

D $ ~f5}id 3

'4c j

? Ji EM 5g ri:

Ir!ii!

e[^ w.j$ h

)

Mk ID N d

f *t W

ce t--- o g=r m2n a# ;:3' we

@a

m. li.1 m:~ i~,

~ A-w i-2 T-a

~= t i 5 ~

r-

. 6

-- c s

n E3 iS3:

E E~~

l 3W 1

= - 4, s :ciar-e-1 CE utsr esh a

W E-kb_th D. $ 5 ? v s 'I b.NM..m m 5

h.rg-N llAl.

5y O

5 g%" *-

G ~UW

-c

-

s u ub.

e N N's N

) I ! *, k, _

m e

- =

=

4

!lf Is Y

• n, w+D,.y.r O I gnTe-d? T e g -f ;; ]

CL.

e t

f N

mI

=,s su s

.m<m

. _o b i ** E t=_

w

,=<-

b y J [* h 4C bs

,-,;$% sf-*

• T 3,I~

}

f g

5 c e

_.....t hs 4,'i 4. m.

2.. a. _.,n u.

e Z

~.fr,,- - M

• v 2

p' b_ g l. W

_e.

" j.+4citc 4

i2 U t J : 5 a, E'.* C*tg-=g.

2 -

r p's-e*

e a c y

6 -

U 3 s s

h,}

.igis i_: : -

7-I tg w

a%

=

Lqt m: tac-I u --<==

s s

~

D

%% case

=._= ~m w..o.

e 9.Js. s.,*,, g Q 'y

Influenced by data under-and over-runs Concern about what to do if no events of a type have been recorded... what choices are to be made when data gathering is especially fecund.

Often complex, scannino the spectrum of statistical methods Examination of an issue may require experimental design, regression, time series analysis, reliability estimation, nonparametric methods, econometric modeling, index creation, et al.

-- Non-textbook and difficult to define Many disciplines.. unclear and/or unstated assumptions... criteria are non-quantitative and/or nebulous

" quick and dirty" answer required...

combine subjective judgments with data.

-- Affected by lack of statistical tradition (such as that developed in regu-lation of drugs)

Special concepts (" common mode failures," "probabilistic licensing") still being argued data-based analyses and arguments often ad hoc and have been naive, misleading, even wrong.

Although most of these characteristics apply to any regulatory activity, two especially nagging aspects arise from nuclear power reactor licensing:

Paucity of data Core melts and releases of radioactivity are, by any measure, " rare events."

So-called classical statistical methods result in "unusably" large un-certainties on the rates of occurrence of these rare events.

So-called Bayesian statistical methods introduce "what is known" in the form of a prior distribution (See, for example, Apostolakis and Mosleh [1979].)

• Will the arena of nuclear power reactor licensing provide a vehicle for resolving the fundamental classical / Bayesian conflict?

-- Applicability of data Nuclear power reactors differ from one another; they have different designs, different operating conditions, different managements, different operating crews.

By various criteria, different operat'ng histories may be " pooled" to assuage the prcblems arising from paucity of data.

• What should these criteria be and how can they be encoded into regulations governing the licensing of nuclear power reactors?

The licensing of nuclear power reactors does not necessarily wait for the data-based philosopher's pronouncements.

Construction permits and operating licenses are denied or issued in terms of present knowledge and in terms of human capacity for understanding and agreement. Monitoring of operating reactors continues, and licenses are continued, modified. suspended, or revoked according to the pertinent technical and legal issues.

The role of data and their analyses in reactor safety issues continue to be debated; WASH-1400 (NUREG-75/014) and NUREG/CR-0400 provide noteworthy points and counterpoints.

The wide-ranging debate engulfs numerous segments of science and society in a way seldom encountered; Kuhn [1970] provides insights into the phenomenon of such uphrtvals.

Still, despite the difficulties and concerns, data are being collected and are being used by the NRC in the licensing of nuclear power reactors.

As Dr. Bernard Fisher noted on CBS-Tl News (June 2,1979), in discussing the raging debate over forms of breast cance treatment, "In God we trust -- all others must have data."

It is imperative that data relating to major issues be exploited for all their content.

To do less !s to surrender to pure opinion.

IV.

AN EXAMPLE A.

Cetting the Stage Harrisburg, Pennsylvaria, according to Rand McNally, is 102 highway miles west of Philadelphia and 10' highway miles north of Pashington, D.C.

Middletown, Pensylvania is some 10 miles southeast of Harrisburg, on the banks of the Sus-quehanna River.

Figure 4 shows the relationships among Harrisburg, Middletown, and their neighboring cities and towns.

Three Mile Island (TMIL is situated approximately 3 miles south of Middletown and is the site for two nuclear power units (TMI-l and TMI-2).

Figure 5 pro-vides geographic detai's.

TMI-l was issued an Operating License on April 19, 1974, and went into uonmercial operation on September 2,1974 TMI-2 was issued an Operating License on February 8,1978, and went into commercial operation on December 30, 1978.

Harrisburg International Airport (HIA) is on the north bank of the Susquehanna, adjacent to Middletown. Tne nearest HI A runway's end is 2.7 miles northwest of the center of the THI nuclear power reactor site.

The geographic relation-ship between TMI and the HIA main runway are shcwn schematically in Figure 6.

B.

Why There Was a Problem As indicated in Section I, the regulations by which the NRC licerises nuclear facilities are codified in Title 10 of the Code of Federal Regulations.

An applicant must prepare and submit a Preliminary Safety Analysis Report when applying for a Construction Permit and a Final Safety Analysis Report when applying for an Operating License.

Specifications for these reports are g5en in NUREG-75/094.

The reports are reviewed by NRC staff according to procedures given in NUREG-75/027, the Standard Review Plan.

g, a.i.J.e O

4

,1

,,o-i

' o, n

t m

y D-

,'t O

i o

,o g

t

% p7 i

E I

i p-f!

)

v

~

d,yF' C

d I

N

}w d

a l

ra A

Y

,6 0,

B

,7%

l[

l

, ]"

1

1. 8

/e

[

i

/-

t 0

~

t 1

(

s" iV g\\

/p[

( (eR

\\

S

\\

a R

U fl I

A

\\

9 i

i C_

J A

CJ-

\\

9 R

[ /e i

t t

I" iv t

M y

C n

A s

E

[

I o

a A

I E u B. H

~

a+o%,Y 1

i t

A 4

l I

E$

s c

D 8

R R

0 N

t A

f y

t

[

D 5

o 0

t I

,c i

0 T

N C

\\_

o,9

\\

~

n fO 0

N 1

OUfg g',

\\

~

s-1 U

n o

l l

N g[g

, x,i,

(

' k

' O g,

" s

+

A R

1 sa r

R]

S 1

g

. E a

x s

Nq.

>i m'

W n

,(

A I

m b

~

n a.

~

s~

N e

\\

t n

!I

_ N A -

0t i

t t

f t

o,

/

w 'S l

d'r J

D i

C k

LI i Nq%

ri c

\\

If

/ L n

S iR s

(

s l

[

M

- m

-i '%{ge t'

i T

U i

it

~

I J

I

c s

g,l E

i i

~

(o f,e i

X Ri

- /m W

n t

5!

.o 7',",M '

i s

s o,

1

,, 82 u

h A n

3 t

y_

c n

g D1 0

n n

'r P

g i

., G u

u. S gN

~

o B, -seU T

p R

0

,./

\\

it

\\

' L g

n Y

B u

Y i

~

r s

S A

\\

/

• - r I

M

~

\\

R A

R

\\

T B

R f

g' 5

s A

i.

E n

N IL e

ftl A; U.ia A,

u

?

0 l

M t

i r

A

[

o.

b4,

(u On st s

B 0

eI s

p n

R qt L

. n I

8 E

C n

A s

1 U

t wI 1"t y

Ca3 3

1 w

tu 2

s s

)i v

5

[kR 1

l

. o 0

y 3

Ns?

F 2

a

,k i

g r

n I

n t

I e

HnI

~

n c E ",,,o..,*Fj v

=

1 7

l O

U PQ ~

y Pe5 D

N s

A n

L n

f Y

R e

l A

N P

A M

U 0

1 O

g D

C r

4 uy bt e

si r

i n u

g n,ic I

A d

j i i

ai F

HV

,w@ sN& 4 yY

(

e.

e 7-r.

SpO.q i

-1C-d'Qy"S, N hQ [n[ w*k N.

,7 Y, M

'!>~. M " f ' M 7

&~Q 4

, n: cw

r.

x u v,,

. a#s

[_.

. c m J n\\ ;,t (77 y

. g..k.? m.

.p:n y

.5 s.

...:.,' t,1.' 2'l..d, v r.r r s-) )f K L

\\, ;..

1

.~ - s R

' y K\\_k,

.$he, C"*" ' 8

~

s', h r). 'h+..

?

a 7

~if - h

&gd-T ' W ' K W ]Q Ef, )'3. R S T F r

' G I.

f f

\\

i.-- -

,\\,}

r.k j i,,-6

. I l'- bp- '

y_ m -?--.

' s.

/s

.L lt

-1 p

sp.m-.r-

-~

,,/7 h

.?;(// / p 5%

v

\\

.k

'f

\\,, ' ; f 'N

.1 D

--a l,

f. %.A

~

' ( w

, s'I l

w 9..

~-

A'k 3

,'l j

j-

.x

a/,;, s 1

n Ih

.y a

.n

'O' s'k

/

w iv.w S -N

>L

~~; 5. Y l'

\\

h aty?-f-&..

y5,M' 2.

U

$(.1

.i.,

M(' - %

%'6 v

r"*

./

.. -h* ',I

'x pWL'/.Qs. Y(

%y

_'e M '.$

^

f 'N '. ? h,%

_,1.".L

)

L/

m y

- s

. D M'

.'Q.-%

M'

.NT[ j

/

)%.

• * * ~ '

fc.s

'A+.; %

\\

w

=hyt'&

p3 'a ).

-ii r

c af

'rf 9i

/,

^ ~

a;.

s v

--- j - +- 7p[m. ',>, s i.p2.u,>w.,a !0

,t h

V

[

i f... ':

%' \\Z ,_ / ':

f f;

=

I y.)

2 s,

,J 1 : s-

' ^

\\.\\.

n

>A s

..:?

• • '~-

1 3.

s.t & -

W'~ a - - / y.'!

' x $\\s o,

. n.

%. ~>

L '?

s s

og L

s, t

z s'.

-s a

v, w

s

,1.

.n.

a --

- 'if

['

tr-Figure 5 -

j I

/

j

.L-

.<-~ j/pc Geographic Relationship r .

- - a

, i

~

v

-x.

<l

.. _,.9 f t a 7,,/

's r

,(.

4

... -;. i

's, y/

u M..?.

TMI and HIA

/

T ' _-, _....,..'

a.

t.

f (q*. a su k *i 4 <.f

\\

s

..J i

re.

,5 I,

' b-s' j,

g, g g, g

{

i-j

. i. '. 9.'.

\\,N.'.f C'$

i

?'

' ~

.a.,

c,,

ie e, ' ~w.- q (g. [, cm, n,,

t k.'.. M. e

' {,

4

.x c

s.o.,

i

,f'. ' re*if,q,,.3PCRT!Cli$CF05r,$ CHARTS > f. -[._-. a f.

\\_

j

> e. e

'g s

s

, me g''*

.\\

,e

,.. j[

sv. c r, t.. -.: u,c m i

,.I p

_______m

• ~ l s' l

[y vi: D:.ETO.W N. P A.

STEE' TON. P A.

4 w:3,Q;. g

(

i 2

s

...P

./ e s

)

f

.v mm.

._.1 m

s

\\

t l s **'***

~j j

fs f, c/ -

~

e f

e n ess q

n o.s yl,, ;

g i, I

/e.

. /.,

6 6 e,-.e r..u e...,

ame.o,o.i -m,.,,

o.

L y g

...ss ema.3a.

m as.esaars van s

.e l'\\ '. f.

Y'.'

s$_'M?. _.S,?-$$Y.'~..

\\.

\\. \\.

'~

e

• ms

" 1

-. 'x g,.

's'

\\

I

==

=.ar g

s o

i x

's

,,7 i

%1 6

N

&{;..

(

N

's

\\

h

~.u.or,c r..s.o..,.., i.,.

j,/

co.io..

e s rm r

g.,

'~

• \\

e.1)

+

\\

~

,d '

/

....,.7Y*

=\\

'N

',/

.vi.

.,.,.u,oc. c.=.v.r.r.n..

,4,y@;4.,,,.g,e o.w

. v i mou. i., r.~.e s e

/

s-E 'I M

'd

^R~P

., a g.

A G"id X

El way

/

2. 2 V.i O

3L

/

2. 7 F.i 1

1.5 Mi TMI Figure 6 -

Geographic Relationship Between Three Mile Island Site and Harrisburo International Airport osc-U 'A.

• d.. ' (,.*9, Aircraft hazards are specifically cited for attention by both of these f!UREGs (Sections 2.2, 2.2.2.5, 2.2.3, and 3.5.1.6 of fiUREG-75/094 and Section 3.5.i.6 of11UREG-75/087).

In particular, Section 3.5.1.6 of fiUREG-75/094 states.

"An aircraft hazard analysis should be provided for... [all]

airports within 5 miles of the site...."

Furthermore, Section 3.5.1.6 of fiUREG-75/087 states :

"The plant is considered adequately designed against aircraft hazards if the probability of aircraft accidents resulting in radiological consequences greater than 10 CFR Part 100 exposure guidelines is less than about 10-per year...."

The guidance to the applicant given in flVREG-75/087 goes on to state:

"The probability per year of an aircraft crashing into the site... (P ) m y be calculated by using the following expression:

A L

M P

z

C. f!i j. A.

=

A J

j 5,3 3,3 where:

M

= number of different types of aircraft using the airport, L

= number of flight paths affecting the site, C.

= probability per square mile of a crash per aircraft move-3 ment, for the jth aircraft, fl. = number (per year) of movements by the jth aircraft along the

$J ith flight path, and A]. = effective plant area (in square miles) for the jth aircraft.

.. [The] choice of values for the parameters should be made judicicusly in order to arrive at a meaningful result.

The manner of interpreting the individual factors may vary on a case-by-case basis because of the specific conditions of each case or because of changes in aircraft accident statistics."

As shown by Figures 5 and 6, TMI-2 is well within the 5-mile criterion.

omo-.

u a.u'.. ' l 'ar-

. The issuance of TMI-2's Operating License on February 8,1978, was appealed to NRC's Atomic Safety and Licensing Appeal Board ( ASLAE) on the basis of, among other issues, the findings and conclusions reached relative to "the probability of a crash of a heavy aircraft into the plant".

In its decision of July 19, 1978, ALAB-486, the ASLAB wrote:

"The record does enable us to find reasonable assurance of safety given present levels of aircraft traffic in the vicinity of the plant.

But it contains sufficient incon-sistencies and ambiguities relative to air crash probabili-ties over the life of the plant that we must order a further hearing on that question. There is, however, no need to suspend the operating license pending the outccme of the hearing...." -- (ALAB-486, p.5)

The decision continued with an extended discussion of the " inconsistencies and ambiguities" in the record regarding aircraft crash probabilities at TMI-2.

Running through the discussion were the two threads singled out in Section III with respect to certain nagging aspects that arise in nuclear power reactor licensing:

paucity and applicability of data.

The ASLAB declared that it had " determined that the record must be reopened to receive additional evidence relative to the probability of crashes of over-200,000 pound aircraft at TMI-2." -- (ALAB-486, p.70).

C.

Stating the Problem This, then, was the state of affairs when we were asked to participate by attorneys in NRC's Office of the Executive Legal Director.

Not surprisingly, the issue had most of the features we've associated with nuclear regulatory matters:

high visibility, technical / emotional intertwining, presence of data under-and over-runs, reasonable complexity, difficulty in definition, and absence of statistical tradition.

But just what was it that we were expected to do? Certain matters gradually became clear:

Only " heavy aircraf t" (planes weighing more than 200,000 pounds) were to be considered because the unit was designed against any untoward consequences in the event a " light aircraft" crashed into it.

-- Various sets of data and interpretations had been performed and presented in earlier testimony, but the ASLAB was seeking more convincing analyses.

-- As a regulatory agency, the NRC is concerned that its positions be as defensible as possible and, if it errs, should err on the " conservative" side.

With respect to the problem at hand, these concerns led to a

" classical" approach with parsimonious and, sometimes, conservative modeling assumptions.

(This had the additional advantage of providing a contrasting analysis to the " Bayesian" approach used by the applicant.)

(> 3 o,

~. c h.

In developing its argument for another hearing on the subject, the ASLAB outlined the model as follows:

"Although the applicants and staff used several techniques to compute the probability that a heavy aircraft might crash into the TMI-2 plant, all of the analytical methods are variations on the following equation:$/

PA""^

For present purposes, the terms of the equation may be defined as follows -

P

= the probability per year that a plane weighing more A

than 200,000 pounds (' heavy aircraft') will crash into the plant (the single TMI-2 unit) crashes per year.

N

= the number of heavy aircraft operations (landings and takeoffs) per year at the airport that might affect the plant -- i.e., those occurring at the TMI end of the runway; operations per year.

A

= the effective area of the facility -- i.e., the area that the plant, as a target, presents to an oncoming aircraf t; souare miles.

C

= the areal crash probability -- i_.e., the probability that a heavy aircraft engaged in a landing or takeoff operation will crash at a designated position with respect to the runway; crash [es] per souare mile per operation...."

"M/ This equation for crash probability uses the same nomenclature and is exactly in the form given in NUREG-75/097 (at p. 3.5.1.6-3),

simplified to consider a single type of aircraft (those heavier than 200,000 pounds), and a single flight path (that associated with Harrisbura runway 310 /130 ),.

-- (ALAB-486, p.41)

Among the matters to be settled by further hearings were two items called out by the ASLAB.

... There shall be provided a complete set of those data on aircraft crashes in the vicinity of airports in the United States which would be pertinent to the calculation of the probability of a crash of a heavy aircraf t at the TMI-2 site...." -- (ALAS-486, p.72) can,.,

v.=../

/L:,

.... The data compilation shall be used to develop a model to compute the probability of a crash per operation and per unit area, at a site off the end of the run way.

The model should reasonably reflect the spatial distribution of crashes displayed by the data and incorporate conservatively any trends for the future which these data portend.

An attempt should be made to assess the precision that might he expected for probability values determined using the model."

(ALAB-486,p.74)

Another important element of the problem consisted of the ground rules under which we operated.

Other PRC staff members were to supply values for N and A, and we were to confine our efforts to the estimation of C.

Thus, the problem reduced to just two aspects:

Estimate C (the areal crash probability)

Assess the precision in the estimate of C.

Since C was declared to be "the areal crash probability" (more properly, the areal density), the problem required only a straightforward application of binomial sampling theory -- or did it?

D.

A Look at the Data The judgment of the NRC staff was that information relating to accidents occurring to both heavy and light aircraft flown by U. S. air carriers in the 48 contiguous states during the inclusive years of 1956-1977 would be pertinent to the TMI-2 problem.

Accordingly, records of the Federal Aviation Administration, the Civil Aeronautics Board, and the National Transportation Safety Board were examined. Attention was concentrated on accidents that were both "off-runway" and "within five miles of the end of the runway."

Of some 1500 accidents reported, a total of 51 met these criteria.

Two arrangements are pertinent.

First, in Table 1 we show the numbers of operations (an operation here is either a takeoff or a landiag), the numbers of " fatal or destruct" accidents, and the accident rates for both scheduled and non-scheduled flights.

Training and ferry flights are excluded from this tabulation because the corresponding numbers of operations are not known.

Next, we look at where the accidents occurred.

For the most part, locations of air crashes are recorded in terms of distance from end of runway (r) and angle of deflection from the " runway extended center line" (9).

But not all accident records yield a determination whether the crash occurred "to the left" or "to the right" of the runway.

Consequently, the air crash locations are recorded in a single quadrant as shown in Figure 7.

The dark squares be-tween the TMI-l and TMI-2 cooling towers indicate the locations of the reactors and their control rooms.

Note particularly that only three accidents involving

" heavy aircraft" are recorded here. Another arrangement of the data appears in Tables 2 and 3 which, respectively, show the takeoff and landing data in a rec-tangular r-9 arrangement.

The reactor location itself is indicated by TMI, Two training accidents on takeoff and two on landing are included in Tables 2 and 3, respectively, since there is no reason to suppose that their areal crash densities are any different from those for scheduled and non-scheduled flights by U. S. air carriers.

<t 1 o -

u x.. :.;a,

x

TABLE 1 - Off-runway Fatal / Destruct Accidents for U. S. Air Carriers (1956-77)

Within 5 Miles of End of Runway SCHEDULED NON-SCHE DULED Operations Rate Operations Rate 6

6 (x 10 )

Accidents (x 10-6)

(x 10 )

Accidents (x 10-6)

Takeoffs 86.3 11 0.127 2.36 2

0.847 Landings 86.3 25 0.290 2.36 13 5.508 Totals 172.6 36 0.209 4.72 15 3.178 (36 + 15) accidents "Overall Rate" = {l72.6 + 4.72) x 10b operations

-6 0.288 x 10 accidents / opera tion

=

ha O

I-a>

CC w

e[e w6* _

4 2

~

~

,' ~

~

. ~

g s ~~

h g~

4i

~

8

~ ~

5N_

\\

~-

9 i

~

iE

\\

)

3 p'

/g N\\

S

\\

()

g g

i r

i n

ij

_N

- )x ~N ~ ' /-

g 3i g

(

)

j g

h

~~

s N

/

r s_

3 U

/

~

J g\\

ir

)

- x

~ /

f N_/

' ~

(

f C

t

,~ d j

/

/

dx J

H b

ob t

7' J

l' t

(:

\\.

h{x'I/

7

/

x' Gl

)

(_s

~I g

M/.

\\'

i J

\\

t

\\-

/

N

"\\

i

().

er1 1

g

\\

~

i 9

\\

/

l 1

/'s s

J o

t 3e.

n I We

\\

gt p '.

s i 1 t

y s./

\\

i J

1 r

i

7. i

\\

l x

l

~

." W N

~

'l

.,p.

t A

it

,. h I'

')l f

l l

ii t

A 4

i t

i At

, Y n

A

~

l t

W l

sl -

a N

l nen n2 d

ei s

~

e r a s

ior ul w-U ra o1 h t

ev e e t n n

hF s t.

~

tt sh a tl t

i 1

I weo sT a ot w f

l f e

s ohi r

l n t

aroa0h oh t t egc peso rob r0t ps a

rn m c 0, y dt n c

f n ail ein rsyr e e or na r

r cee id si0b h

e si u

(, -

acd r an a0 t a1 L

r t nat aia ue 2d nn ot c s

c

. goh g e

eI i na nycsn inpt nnt y2 t o ioa anw iaa l

- h g ace w

oo v0 yvh c n1 t r c

h t neih l 0bl ti ot uu ott 1

s o

d i ob aust s o 0, d l n croa v

ven e

ss af l t t oo i0ti o nneri t ro.

o h ct n0enri cd nd n

a l n 2 a n e eah r ea s o hd s

cs r. r t ia d usi t t nee t git ga el l

ic l

a o

qet n

at r nnd nn f - o et ea eineisl e! tt cedt d p)l dhidhI. o rll a

ah rn i e0 pe ig ignb I e t ae cc-ml ciecium r vy l

gi 5

cx rob cerceoy ofi ner l

Ae( ca AwaAwps F ot Airo f

l 0

3 4.

N 1

2 T3_g Og=

c-

=

a g

y

( W[ ;-(J

). A ;

n

\\

l 3

m

- 1

=

cs

- ' e o

O h

l 4*

~

i i

i I

s l

l g

e I Ef i

i

\\

l ti k

l 6

e i

4 i

d l ~

~

~ l e-i

-~

i, i

om N

/

l

~

l me n

s b

Ie C*

1 om I

4 g

O

.C C l

s c

k

~c I

e i

s l

kN l

C s

O k

l I

es e

s g

a 80 e!

i er 28 l

i I

l 3 me l

v s

i 1

a1 -

m. ].

! e GA I

'[

l w.

Om I

1 C

l l

C u,

l l

I MA

@t=,E5l mi.

i "t

g

~

1 i

~

~

cc '

D j

r a

O O. hl Cm I

C

,2 e-i 6

s "e

.A } 4 S 's % L % % N \\ 5.# * %14 + % % %g %%%%%N % %L L % '

• +

! 5 l

l

)

l

.2 m.

l l

C I

$k O

4

( l.

I j

r.

RW s

I o'

J S-

' n hQ i

cl

's su 3

-1 C

g 4..

WC l

/

g Nh e=

Csis,persr>rr

.ee rasus i s a o r

' r r,. s

= <

C a-.

O N

i

~

m =

e

~~

w 3

.C ti >

co k

c <

- ev

/

a c, =

w w

• c O@

k

>=

D<

[

C

- w

.. a

^

V O g

-E

/

o c

2 s

4-e x

v 4

nQ

~

l x

et -

i m

5

\\

'2 C

s

. - + '

c c

=

,o c

c c

r

= l

?:

3 N

} _~ :-

_d

,N m

e w

x e

. I g l :4 e.

=.l

. s=.

. x

=%

.c e

- I -

j c.

- l l=.

c

\\

u l

ff. ' f I' V

/'

f - l 3 N

e, I cs 6

t._

l l

o $ ",,.hO o x..

i sv,%

l _ %)i i

as l

', e i

5

!e-l m

- i e

e m

s

~

j "I

E I

.i I

t i

E I

I h

l I

i;l i

l s

i l

V' 1

_ i i

i 9

l l

lN

• E I

- [-

i

, E-l I

i s

i l

i

,$2i e

Le i e

l l

N i

l d2 i

cm Cm l 4

l C

3A I

1 CC l l

I I

s h

I i

' ~c

)

i S

I l

I s

l l

C kN i

! $ k ll l

l

{

t i

C w

/

e lb3 3

• i i

m n!

i i

s l

I o

a

_ Esi i

s i

u 3_

e ni 4 -

x II Id2 f

l C

-, wj s

\\

s

=, el

'cv w

% % N \\ % s b \\ ]* E

- l '.a s s, % % } % % N % \\p s %it % w % %g %%

i l

E g

Ef E IN E i

1 o,

r

=

E S_i hd2 l i

l j

i g

-a

+

'",. _ i mm el CM w

I

,, e I-l l

a, i

m

==

a

. ~

rffr /* k / F f' r r d' r r sryF FAPf fr2r7i /f f 3'

Eo p

l 2=

~

_R I

R

.c h -

l

._-*2l-m sE l

p

.5 I

t$

l =T '

~

I i

~

5 = D I- ) r @g @i H

...=

i a

I

!d-i g

z n R

,g M

~

N N

jd2 p

g p,

g s

ES

!'l

?~E e

i -

v_

C C

cC O

0 j'

O

1. g

.E(

- I e;

~

i v

e o

. ;m.

o

=.

a m

_- m g

, i e.

=. !

f=

< i eg e

m

&y -

?

' _y e

w

,, - l ~

c.

, =

iig

/ J') r / V~J /

l

\\

g a

Gan..,

v.A.. /.m Q,,

E.

Formulating an Approach The problem is this:

Provide an estimate of the areal crash density at the location of the TPI-2 unit.

Synbolically, let the areal crash density at a general location (r, 9) be designated by C(r, 9) and assume that it is the product of two components.

Thus C(r, 9) = P D(r, 9),

in which P denotes the probability of an off-runway crash and D(r, 9) denotes the probability per square mile of c crast at the point (r, 9), given that a flight has crashed. At TMI-2, r = 2.7 miles and 9 = 34, so that the immediate problem is the estimation of C(2.7, 34).

And don't forget to assess the orecision in the estimate!

Table 1 indicates considerable differences in accident rates according to whether a flight is scheduled or non-scheduled or whether an operation is a takeoff or a landing.

With respect to the location of a crash, however, the judgment of the NRC staff was that whether a flight was scheduled or not would have no influence on where it hit.

These considerations result in the following refinement:

Provide estimates of

$j(r, 9) = P D (r, 9)

C 4

where i = T for takeoff

= L for landing and j

= S for scheduled

= N for non-scheduled.

And don't forget to assess the orecisions in the estimates:

The estimates of P9j, denoted P4

, can be obtained from Table 1.

The un-certainties are assessable in a number of ways, the most direct of which is application of Poisson distribution theory to estimation of a Poisson para-meter.

However, estimation of D (r, 9) and D (r, 9) require a bit more doing.

T g

Tables 2 and 3 show that no accidents have occurred in the r-G rectangle that contains TMI.

So the easy answer is that zero is the estimate of both probabilities D (2.7, 34) and D (2.7, 34).

But that would mean that all T

L four estimates of C;j(2.7, 34) would also be zero.

However, estimates of zero are unacceptable to almost all parties in this situation -- because, we believe, of the conceptual difficulties inherent in distinguishing be-tween a parameter (or a physical constant) and the corresponding data-and model-based estimate of that parameter.

There goes the easy answer'

$.&G

. he next take the following approach: Suppose the two measurements of hit location, r and 9, are independent of each other.

Then it would be possible to estimate the joint density D (r, 9) by the product of the two marginal densities.

The information in Tables 2 and 3 was arranged in two contingency tables -- with considerable pooling of adjacent cells -- and, based upon the chi-squared statistic, it was determined that r and 9 could be treated as though they were indeed independent for both takeoffs and landings.

So our model takes on an additional refinement, where the subscripts (T, L, S, N) are dropped for clarity:

By definition, the crash density D(r, 0) is the probability of an accident per unit area at (r, 9), given an off-runway accident.

Consider the differ-ential region with sides dr and rd9 and with area rdrd0, where r is measured in miles and 9 is measured in radians (the shaded region in Figure 8).

If dQ is the probability of an accident in that region, then D(r, 9) =

r de Now let g(r) and h(9) be the crash densities or r and 9, respectively, conditional on an off-runway crash.

By the assumed independence of r and 9, dQ = [g(r)dr] - [h(0)dG]

It follows that 0(r, 9) = g(r) h(9) r Although both the takeoff and the landing data bases show accidents occurring in the 2.5-3.0 mile interval, neither set of data shows accidents in the 30-35 degree interval.

So, once again, estimates of the areal crash densities are zero.

Still armed with independence of r and 9, we extend the rectangle in which TMI resides.

So, instead of looking at the half-mile by five-degree section, we expand to a one-and-a-half-mile by fifteen-degree section as outlined in Tables 2 and 3.

And this does the trick -- non-zero estimates of the arcal crash densities are obtained.

(In point of fact, this technique is not as arbitrary as it may seem.

Because of ambiguities in the accident records, exact locations of crash impact points are difficult to ascertain.

Thus, widening the rectangles of interest helps to " smooth" this uncertainty.)

The procedure becomes :

(1)

Let N be the total number of take-off accidents.

T (2)

Count the number of take-off accidents in the three half-mile intervals which include r (the distance to the site) in the center interval.

Let g

this number be X '

T n 3 o

• o O g) + rJ u 1

, s\\.. /

.y

<\\

/

\\

/.

/,/

/ rd6 e r

e

/ de\\

[^

i s-

/

/

/

/

\\

/

Shaded Area:

/

dA = rde 7 dr

/

/

/

/

\\/

Il/

l M

Eun.:ay l

~

Figure 8 -

DIFFERENTIAL ACCIDENT REGION

!? !l!G3 D " ' ! :l {

u ou

<D,,'T

;' [, l, If

.,c-I "

i ij l

I U' ),

i;\\,'<

j. ;

'~

g,, y r, - * * /1 U.K.,,.. U '

(3)

Count the number of take-off accidents in the three five-degree intervals wnich include O (the angular distance from the runway to the site) in g

the center interval.

Let this number be Y '

T (4)

Adjustment for the widths of the distance (r) and angle (0) intervals to yield square-mile units in the final result and make estimates as follows:

~

I (Thefactor1/1.5placesthe Estimate gT(r ) by gT( o) = 1.5 N

x f estimate on a,per-mile basis.

o T

^

180 T

The factor 180/15r places the stimate h (b ) by h (b )

• T5-~

• y

• estimate on a "per-radian" basis.*

t T o T o gT(r )h (6 )

T (5)

Estimate D (r, G ) =

T g g

ro The factor l!/r places the estimate by

=fx x h (0 g

T o

on a per-square-mile basis.

(6)

Repeat steps (1) through (5) for landing data, using th.e obvious notational v

p changes N, X ' and Y, to yield 'g

,~h (

, and 'D{ (r, O ) wi th the L

g g

additional multiplication by the factor of 1/2 in to correct for the

" folding" of the two sides of the runway in Figure 7.

• Because of the " folding" of the two sides of the runway in figure 7, it might seem that a factor of 1/2 is needed to make an unbias ed estimator of h (O ).

This would be correct if aircraf t always took off s.traight ahead and T o had no left-right crash preference.

However, at many airpor-ts aircraf t of ten take off either to the left or to the right.

Accordingly, i'1 they do crash on takeoff, they would tend to crash in the quadrant over which they took off.

(This is not true for landings, where aircraft typically make their final landing approaches along the extended runway center lir.e and have nc, lef t-right crash preference.) As a consequence, introducing a factor of 1/2 -in the formula for would make it a negatively biased estimator of h (O )

Since we do not T o know what factor to use in order to get an unbiased estimater of h (O )' **

T o make no correction and get a positively biased e'timate of h (O ).

We take this 7 g approach because of the regulatory desirability tr ue a "comservative" estimator wnen an unbiased one is not known.

_] l[ I c

U

'I Q [c G..i l

<3 3 f-r o-ll j;

' I! fI U A../_

sJtJ

.SJ i _ e L i -. ' ~

- 2: -

(7)

Estimate C )(r, O ) by 5

g g

,/N

/

^

C;j(r, G ) = P ) x D (r, O ),

g g

j j

g g

where i = T for takeoffs and L for landings and j = S for scheduled and N for non-scheduled.

But we still must address the assessment of precision in these estimates.

Our initial choice is to use confidence intervals and to bring Bonferroni inequalities to bear on the problem, in a manner discussed by Morrison

[1976, p.33], for example.

That is, we are dealing with the product of three estimators.

If we also form the product of the corresponding (1 - a) confidence limits on each of them, then we conclude that we have captured the true product with confidence at least as large as (1 - 32).

G.

What the Answers Are The data recorded in Tables 1, 2, and 3 are employed to estimate the four different values of C33(2.7,34).

The estimates, along with the intermediate results, appear in Table 4.

Confidence limits, based on Poisson theory and employing the Bonferroni method. appecr in Table 5.

H.

Envoi The matter of air crash probabilities at inree Mile Island -- and elsewhere --

remains open.

In particular, the final ASLAB decision on TMI-2-specific con-siderations has not been issued; thus the acceptability of this approach has not yet been detemined.

Among the remaining technical questions are:

-- How will aircraft accident data for 1978 and 1979 affect the estimates?

-- How can the independence of distance and angle be exploited? How can the confidence limits be sharpened?

(One idea is to use the method proposed by Buehler [1957] regarding confidence limits on the product of parameters for independent binomial parameters. )

-- What generic treatment of these data can be devised for application to o+.her sites?

-- How can uncertainties in future air service and uncertainties in the effective plant area be combined with the statistical measures of pre-cision to place bounds on a plant's lifetime probability of aircraft accidents?

-- Above all, what more illuminating procedures can be trained on this problem-and how can they be applied to study of other types of hazards?

o., e s ?_

a s.../.a.n E *

.;c.

l Table 4 - ESTIMATES OF TAKE0FF-LANDING-SCHEDULED-tion-SCHEDULED AREAL CRASH :

DENSITIES FOR THREE MILE ISLAND-2 (See Section IV-E of text for i

notational conventions.)

N = 15 N = 40 T

L X

=3 X

=8 T

L l

l Y

=3 Y

=3 T

g

'N 1

3 1

8

.gT(2.7) = 1.5

• IT =.133 g ( 2. 7)' = 1. 5

• 4g =.133 L

l' /(34) = 180

'N 1

180 3

3 T

15r *l5 =.764 h (34) = 7 x 15n x g =.143 h

D (2 /, 34) = (.133)(.764)

D (2.7, 34) = (.133)(.143)~

T 2.7 L

2.7

=.0376

=.0070

-6

-6 P

= 0.127 x 10 p

= 0.290 x 10 TS lS

-6

-6 P

= 0.847 x 10 P

= 5.508 x 10 TN g

-9

/

LS(2.7, 34) = 2.0 x 10-9 CTS (2.7, 34) = 4.8 x 10 C

~

o

-9 CTN(2.7, 34) = 32 x 10 '

CLN(2.7, 34) = 39 x 10 i

[See Table 5 for assessments of precisions of these estimates.]

oin-;y u x e l. s t ) ')

-26 Table 5 - ESTIMATES A;D CONFIDEfiCE LIMITS FOR TAKEOFF-LANDIf4b-SCHEDULED-NON-SCHEDULED AREAL CRASH DEt;SITIES FOR THREE MILE ISLAND-2 (See Table 4 and Section IV-E of the text for bases.)

SCHEDULED Confidence Limit

( x 10-9) j ESTIVATE 70%

85%

97%

-9 TAKE0FFS 4.8 x 10

<36

<53

<105

-9 LANDINGS 2.0 x 10

<10

<13

<23 NONSCHEDULED j

Confidence Limit (x 10-9)

ESTIPATE 70%

85%

97%

-9 TAKE0FFS 32 x 10

<420

<670

<l 510 LANDINGS 39 x 10-9

<210

<290

<530 SL.'C:>B

- Acknowledcements We and this paper have benefitted from numerous discussions with many members of the NRC staff, particularly Lawrence Chandler, Jacques Read, David Rubinstein, and Stuart Treby.

We are indebted also to Terry Barnhart for her unfailing good humor when faced with yet another in a (possibly non-denumerable) sequence of textual adjustments.

(i.y <', *

• n v.h. a

.. s > _'

Peferences ALAR-486, U. S. Nuclear Pegulatory Ccmnission, Atomic Safety and Licensing Appeal Board (July 19, 1978), Decision in the Patter of Petropolitan Edison Company, et al.

(Three Mile Island Nuclear Station, Unit No. 2),

Washington, D.C.

Apostolakis, George, and Mosleh, Ali (1979), " Expert Opinion and Statistical Evidence: An Application to Reactor Core Melt Frequency,' Nuclear Science and Engineering, 70, 135-149.

Buehler, Robert J. (1957), " Confidence Intervals for the Product of Two Binomial Parameters," Journal of the Anerican Statistical Association, 52, 482-493.

Vuhn, Thomas S. (1970), The Structure of Scientific Pevolutions, Second Edition, Enlarged, Chicago: The University of Chicago Press.

Moore, Roger H., and Easterling, Robert G. (1978), " Statistical Problems in t'uclear Regulation:

Introduction and Overview," paper presented at the 138th Annual Peeting of the American Statistical Association in San Diego, California.

Morrison, Donald F. (1976), Pultivariate Statistical Methods, Second Edition, New York: McGraw-Hill Book Company.

NUREG/CR-0400, U. S. Nuclear Regulatory Commission (1978), Pisk Assessrent Review Group Reoort to the U. S. Nuclear Regulatory Commission, Springfield, VA: National Technical Information Service.

NUREG-75/087, U. S. Nuclear Regulatory Commission (revised periodically),

Standard Review Plan, Washington, D.C.

NUREG-75/094, U.S. Nuclear Regulatory Commission Regulatory Guide 1.70, Revision 2 (1975), Standard Format and Content of Safety Analysis Reports for Nuclear Power Plants, LWR Edition., Springfield, VA: National Technical Information Service.

WASH-1400 (NUREG-75/014), U. S. Nuclear Regulatory Commission (1975), Reactor Safety Study - An Assessment of Accident Risks in U. S. Commercial Nuclear Power Plants, Springfield, VA: National Technical Information Service.

10 CFR, General Services Administration Office of the Federal Register (updated January 1 of each year), Code of Federal Regulations, Title 10:

Chapter I __

Nuclear Regulatory Commission, Washington, DC:

U. S. Government Printing Office.

Not referenced in text:

Scherer, F. W. (1979), " Statistics for Government Regulation," The American Statistician, 33,1-5.

31.L'i.60