iaea common cause failure (ccf) analysis and generic ccf ... · ng o f componen t s t a t es, d ue...

31
International Atomic Energy Agency International Atomic Energy Agency Common Cause Failure (CCF) Analysis and Generic CCF Data ~US Experience IAEA Technical Review Meeting November 06-08, 2013 See Meng Wong, PhD U.S. Nuclear Regulatory Commission See-Meng Wong@nrc gov 1 See Meng.Wong@nrc.gov

Upload: others

Post on 01-Sep-2019

2 views

Category:

Documents


0 download

TRANSCRIPT

Inte

rnat

iona

l Ato

mic

Ene

rgy

Age

ncy

Inte

rnat

iona

l Ato

mic

Ene

rgy

Age

ncy

Com

mon

Cau

se F

ailu

re (

CC

F) A

naly

sis

and

Gen

eric

CC

F D

ata

~U

S E

xper

ienc

e

IAE

A T

echn

ical

Rev

iew

Mee

ting

Nov

embe

r 06-

08, 2

013

See

Men

g W

ong,

PhD

U.S

. Nuc

lear

Reg

ulat

ory

Com

mis

sion

See

-Men

gW

ong@

nrc

gov

11

See

Men

g.W

ong@

nrc.

gov

Pres

enta

tion

Out

line

Pres

enta

tion

Out

line

O

bjec

tives

P

rovi

de a

n ov

ervi

ew o

f Com

mon

Cau

se F

ailu

re (C

CF)

m

odel

ing

D

iscu

ss th

e ge

nera

l app

roac

h of

CC

F an

alys

is in

risk

g

yas

sess

men

t

C

CF

Mod

els

CC

F M

odel

s

Bas

ic P

aram

eter

Mod

el

Bet

a-Fa

ctor

Mod

el

Mlti

lG

kL

ttM

dl

M

ultip

le G

reek

Let

ter M

odel

A

lpha

-Fac

tor M

odel

A

naly

sis

Proc

ess

22

Com

mon

Cau

seFa

ilure

s(C

CFs

)C

omm

on C

ause

Fai

lure

s (C

CFs

)

Su

bset

ofD

epen

dent

Failu

res

inw

hich

two

orm

ore

Subs

et o

f Dep

ende

nt F

ailu

res

in w

hich

two

or m

ore

com

pone

nt fa

ult s

tate

s ex

ist a

t the

sam

e tim

e, o

r w

ithin

a s

hort

tim

e in

terv

al, a

s a

resu

lt of

a s

hare

d ca

use

caus

e

Th

e sh

ared

cau

se is

not

ano

ther

com

pone

nt s

tate

b

hdi

ft

tt

dbe

caus

e su

ch c

asca

ding

of c

ompo

nent

sta

tes,

due

to

func

tiona

l cou

plin

gs, a

re a

lread

y us

ually

mod

eled

ex

plic

itly

in s

yste

m m

odel

s

R

esid

ual d

epen

dent

failu

res

who

se ro

ot c

ause

s ar

e no

texp

licitl

ym

odel

edin

the

PSA

not e

xplic

itly

mod

eled

in th

e PS

A

33

Why

isC

CF

Mod

elin

gIm

port

ant?

Why

is C

CF

Mod

elin

g Im

port

ant?

U

.S.c

omm

erci

alnu

clea

rpow

erpl

ants

(NPP

s)U

.S. c

omm

erci

al n

ucle

ar p

ower

pla

nts

(NPP

s)

are

desi

gned

with

saf

ety

as fo

rem

ost p

riorit

y

Red

unda

ncy

D

iver

sity

D

efen

se in

dep

th

Saf

ety

mar

gins

N

PPde

sign

sar

eef

fect

ivel

ysi

ngle

failu

re

NPP

des

igns

are

effe

ctiv

ely

sing

le fa

ilure

“p

roof

O

nly

com

bina

tions

of f

ailu

res

can

serio

usly

ch

alle

nge

reac

tor i

nteg

rity

4

Exam

ples

ofC

CF

Exam

ples

of C

CF

H

uman

Inte

ract

ions

M

aint

enan

ce te

chni

cian

inco

rrec

tly s

ets

setp

oint

son

mul

tiple

co

mpo

nent

s

Inco

rrec

t or i

ncor

rect

ly a

pplie

d lu

bric

ant

Ph

ysic

al o

r env

ironm

enta

l con

ditio

ns

Bio

-foul

ing

(e.g

., cl

ams,

mus

sels

, fis

h, k

elp,

etc

.)g

(g

p)

D

esig

n or

man

ufac

turin

g de

fect

C

onta

min

atio

n in

lubr

ican

t or f

uel

N

ot re

pres

ente

d ex

plic

itly,

onl

y pa

ram

etric

ally

5

Dep

ende

nt F

ailu

res

p

Com

bina

tions

of i

ndep

ende

nt fa

ilure

s ar

e ex

trem

ely

rare

even

tsex

trem

ely

rare

eve

nts

D

epen

dent

failu

res

pose

maj

or c

halle

nge

to

safe

ty

Sha

red

equi

pmen

t and

sup

port

syst

em

depe

nden

cies

depe

nden

cies

•E

xplic

itly

mod

eled

in P

SA

logi

c

Fa

ilure

s of

mul

tiple

com

pone

nts

from

a c

omm

on (o

r sh

ared

) cau

se•

Cau

se n

ot e

xplic

itly

mod

eled

py

•Tr

eate

d pa

ram

etric

ally

–C

CF

mod

els

6

Def

initi

onof

Dep

ende

ncy

Def

initi

on o

f Dep

ende

ncy

Ev

ents

Aan

dB

are

said

tobe

depe

nden

t

Even

ts A

and

B a

re s

aid

to b

e de

pend

ent

even

ts if

P(A

*B) =

P(A

|B) *

P(B

)=

P(B

|A) *

P(A

)≠

P(A

) * P

(B)

Ty

pica

lly if

eve

nts

are

depe

nden

tP

(A*B

) > P

(A) *

P(B

)(

)(

)(

)

Th

is is

why

dep

ende

nt e

vent

s ar

e a

safe

ty c

once

rn

7

Type

s of

Dep

ende

nt E

vent

s B

ased

on

Thei

rI

tPS

AM

dl

Impa

ct o

n a

PSA

Mod

el

8

CC

FM

odel

sC

CF

Mod

els

B

asic

Par

amet

er

Bet

aFa

ctor

Bet

a Fa

ctor

M

ultip

le G

reek

Let

ter (

MG

L)

Alp

haFa

ctor

A

lpha

Fac

tor

9

(1)B

asic

Para

met

erM

odel

(1) B

asic

Par

amet

er M

odel

1010

(1)B

asic

Para

met

erM

odel

(1) B

asic

Par

amet

er M

odel

QQQQ

QQAA

QQ33

QQ11 Q2Q2

1111

(1)B

asic

Para

met

erM

odel

(1) B

asic

Par

amet

er M

odel

M

otiv

atio

n fo

r Par

amet

ric M

odel

s

D

ata

need

ed to

est

imat

e Q

kin

bas

ic

para

met

erm

odel

are

notg

ener

ally

avai

labl

epa

ram

eter

mod

el a

re n

ot g

ener

ally

ava

ilabl

e−

Gen

eric

failu

re p

roba

bilit

ies/

rate

s fo

r com

pone

nts

(i.e.

, Qt)

−C

ompi

latio

nsof

depe

nden

tfai

lure

s(w

ithou

t−

Com

pila

tions

of d

epen

dent

failu

res

(with

out

dem

and

data

)

A

ltti

dl

ltt

if

tit

A

ltern

ativ

e m

odel

s us

e la

tter i

nfor

mat

ion

to

deve

lop

rela

tive

fract

ions

of d

epen

dent

failu

re

even

ts

1212

(2) β

-Fac

tor M

odel

O

rigin

ally

dev

elop

ed fo

r 2-c

ompo

nent

sys

tem

s; la

ter e

xten

ded

to

hdl

lt

hand

le la

rger

sys

tem

s

B

ased

on

notio

n th

at c

ompo

nent

failu

res

can

be d

ivid

ed in

to tw

o gr

oups

grou

ps

Inde

pend

ent f

ailu

res

D

epen

dent

failu

re o

f all

com

pone

nts

A

lloca

tion

mod

el:

Qt=

Q1+

Qm=

(1 –β)

Qt+

βQ

t

Inde

pend

ent c

ontri

butio

n D

epen

dent

con

tribu

tion

Ther

efor

e: β≡

Qm/(Q

1+Q

m)

13

(2)β

-Fac

torM

odel

(2) β

Fact

or M

odel

β-

Fact

or E

stim

atio

n ~

Exa

mpl

e p

C

onsi

der a

sys

tem

with

two

com

pone

nts:

A a

nd B

C

ompo

nent

A h

as fa

iled

3 tim

es in

50,

000

hour

s of

ser

vice

; ou

t of t

hose

3 fa

ilure

eve

nts,

1 e

vent

was

a c

omm

on c

ause

fa

ilure

(invo

lvin

gco

mpo

nent

B)

failu

re (i

nvol

ving

com

pone

nt B

)

Com

pone

nt B

als

o ha

s 50

,000

hou

rs o

f ser

vice

, and

it h

as

faile

d 2

times

(inc

ludi

ng th

e jo

int f

ailu

re e

vent

with

A)

P

oint

est

imat

es fo

r λtan

d β

are

calc

ulat

ed a

s fo

llow

s:

λ t=

5 fa

ilure

s/ 1

00,0

00 h

r= 5

.0 x

10-

5 /hr

β=

2/(3

+2) =

0.4

λ C

CF=

λt* β=

5.0

x 1

0-5 /h

r* 0

.4

λ C

CF=

2.0

x 1

0-5 /h

r

In

the

abse

nce

ofpl

ants

peci

ficda

taba

seco

mpo

nent

failu

rera

te

In th

e ab

senc

e of

pla

nt-s

peci

fic d

ata,

bas

e co

mpo

nent

failu

re ra

te

(λt)

is o

btai

ned

from

gen

eric

failu

re ra

tes

14

(3)M

ultip

leG

reek

Lette

r(M

GL)

Mod

el(3

) Mul

tiple

Gre

ek L

ette

r (M

GL)

Mod

el

•β-

fact

or e

xten

sion

to tr

eat m

ultip

le le

vels

of C

CF

βp

•D

efin

ition

s:D

efin

ition

s:β

= co

nditi

onal

pro

babi

lity

that

the

caus

e of

a s

peci

fic c

ompo

nent

fa

ilure

will

be s

hare

d by

one

or m

ore

addi

tiona

l com

pone

nts

ϒ=co

nditi

onal

prob

abilit

yth

atco

mm

onca

use

failu

reof

asp

ecifi

cϒ= co

nditi

onal

pro

babi

lity

that

com

mon

cau

se fa

ilure

of a

spe

cific

co

mpo

nent

that

has

faile

d tw

o co

mpo

nent

s w

ill b

e sh

ared

by

one

or

mor

e ad

ditio

nal c

ompo

nent

s∆

=co

nditi

onal

prob

abili

tyth

atco

mm

onca

use

failu

reof

asp

ecifi

c∆

= co

nditi

onal

pro

babi

lity

that

com

mon

cau

se fa

ilure

of a

spe

cific

co

mpo

nent

that

has

faile

d th

ree

com

pone

nts

will

be

shar

ed b

y on

e or

mor

e ad

ditio

nal c

ompo

nent

s

15

(3)M

ultip

leG

reek

Lette

r(M

GL)

Mod

el(3

) Mul

tiple

Gre

ek L

ette

r (M

GL)

Mod

el

DDBB

CC

1616AA

(3)M

ultip

leG

reek

Lette

r(M

GL)

Mod

el(3

) Mul

tiple

Gre

ek L

ette

r (M

GL)

Mod

el

1717

(4)A

lpha

Fact

orM

odel

(4) A

lpha

Fac

tor M

odel

S

impl

eex

pres

sion

sfo

rexa

ctdi

strib

utio

nsof

Sim

ple

expr

essi

ons

for e

xact

dis

tribu

tions

of

MG

L pa

ram

eter

s (a

ccou

ntin

g fo

r unc

erta

intie

s)

are

not a

lway

s ob

tain

able

A

ppro

xim

ate

met

hods

lead

ing

to p

oint

es

timat

ors

prov

ided

earli

erun

dere

stim

ate

estim

ator

s pr

ovid

ed e

arlie

r und

eres

timat

e un

certa

inty

ά-

fact

or m

odel

dev

elop

ed to

add

ress

this

issu

e

1818

(4)A

lpha

Fact

orM

odel

(4) A

lpha

Fac

tor M

odel

1919

(4)A

lpha

Fact

orM

odel

(4) A

lpha

Fac

tor M

odel

2020

Ana

lysi

sPr

oces

sA

naly

sis

Proc

ess

G

ener

alS

teps

Gen

eral

Ste

ps

Sta

rting

with

sys

tem

logi

c m

odel

, ide

ntify

co

mm

onca

use

com

pone

ntgr

oups

com

mon

cau

se c

ompo

nent

gro

ups

D

evel

op C

CF

mod

el

Gat

hera

ndan

alyz

eda

ta

Gat

her a

nd a

naly

ze d

ata

Q

uant

ify C

CF

mod

el p

aram

eter

s

Qua

ntify

CC

Fba

sic

even

ts

Qua

ntify

CC

F ba

sic

even

ts

2121

Ana

lysi

sPr

oces

sA

naly

sis

Proc

ess

Id

entif

y “C

omm

on C

ause

Com

pone

nt G

roup

s”

D

efin

ition

:Agr

oup

ofco

mpo

nent

sth

atha

sa

sign

ifica

ntlik

elih

ood

ofex

perie

ncin

ga

D

efin

ition

: A g

roup

of c

ompo

nent

s th

at h

as a

sig

nific

ant l

ikel

ihoo

d of

exp

erie

ncin

g a

com

mon

cau

se fa

ilure

eve

nt

C

onsi

der s

imila

rity

of:

•C

ompo

nent

type

pyp

•M

anuf

actu

rer

•M

ode

of o

pera

tion/

mod

e of

failu

re•

Envi

ronm

ent

•Lo

catio

n•

Mis

sion

•M

issi

on•

Test

and

Mai

nten

ance

Pro

cedu

res

“C

omm

on C

ause

Com

pone

nt G

roup

s” S

cree

ning

Pro

cess

D

iver

sity

(eg

inop

erat

ion

mis

sion

s)is

apo

ssib

lere

ason

fors

cree

ning

D

iver

sity

(e.g

., in

ope

ratio

n, m

issi

ons)

is a

pos

sibl

e re

ason

for s

cree

ning

ou

t •N

ote:

div

erse

com

pone

nts

can

have

com

mon

pie

ce p

arts

(e.g

., co

mm

on

pum

ps, d

iffer

ent d

river

s)

2222

Ana

lysi

sPr

oces

sA

naly

sis

Proc

ess

Iffi

iFl

Dev

elop

men

t of C

CF

Mod

elIn

suffi

cien

t Flo

w

From

2/3

EC

I Tr

ains

pE

xplic

it re

pres

enta

tion

exam

ple

Spe

cific

com

bina

tions

of

com

pone

nts

are

expl

icitl

y

Inde

pend

ent

Har

dwar

e Fa

ilure

O

f Pum

p Tr

ains

Com

mon

-Cau

se

Failu

re O

f Pu

mps

show

n on

faul

t tre

e

Com

mon

-Cau

se

Failu

re O

f Pu

mps

A a

nd B

Com

mon

-Cau

se

Failu

re O

f Pu

mps

B a

nd C

Com

mon

-Cau

se

Failu

re O

f Pu

mps

A a

nd C

Com

mon

-Cau

se

Failu

re O

f Pu

mps

A, B

, and

C

2323

Ana

lysi

sPr

oces

sA

naly

sis

Proc

ess

Im

plic

itm

odel

ing

exam

ple

(3tra

ins)

Impl

icit

mod

elin

g ex

ampl

e (3

trai

ns)

P

{top

even

t due

to C

CF}

= 3

Q2

+ Q

3

P

roba

bilit

ies

ofdi

ffere

ntco

mbi

natio

nsar

e“r

olle

d-up

Pro

babi

litie

s of

diff

eren

t com

bina

tions

are

rol

led

up

into

the

CC

F te

rm.

Insu

ffic

ient

Flow

Insu

ffic

ient

Flo

w

From

2/3

EC

I Tr

ains

Inde

pend

ent

Har

dwar

e Fa

ilure

O

f Pum

p Tr

ains

Syst

em F

ails

du

e to

CC

F

2424

Dat

aA

naly

sis

Proc

ess

Dat

a A

naly

sis

Proc

ess

D

ata

Sou

rces

D

ata

Sou

rces

G

ener

ic ra

w d

ata

com

pila

tions

(e.g

., Li

cens

ee E

vent

R

t(L

ER

)LE

Ri

NP

E)

Rep

orts

(LE

Rs)

, LE

R s

umm

arie

s, N

PE

)

Pla

nt-s

peci

fic ra

w d

ata

reco

rds

(e.g

., te

st a

nd

mai

nten

ance

reco

rds

wor

kor

ders

oper

ator

logs

)m

aint

enan

ce re

cord

s, w

ork

orde

rs, o

pera

tor l

ogs)

G

ener

ic e

vent

dat

a an

d pa

ram

eter

est

imat

es (e

.g.,

NU

RE

G/C

R-2

770,

EP

RI N

P-3

967)

N

RC

/INL

CC

F da

taba

se (N

UR

EG

/CR

-626

8)

2525

Dat

aA

naly

sis

Proc

ess

Dat

a A

naly

sis

Proc

ess

E

xam

ines

failu

re e

vent

s(no

t all

dem

ands

or

tsu

cces

s ev

ents

)

R

elat

ivel

yfe

wfa

ilure

sar

ecl

ear-

cutC

CFs

Rel

ativ

ely

few

failu

res

are

clea

rcu

t CC

Fs

Dem

ands

on

redu

ndan

t com

pone

nts

do n

ot a

lway

s oc

cur

sim

ulta

neou

sly

“F

ailu

res”

are

som

etim

es n

ot d

emon

stra

ted

failu

res

•S

econ

d co

mpo

nent

insp

ecte

d an

d re

veal

ed s

imila

r de

grad

atio

n/co

nditi

ons

In

terp

reta

tion

and

judg

men

tuse

dto

“fill-

in”t

heIn

terp

reta

tion

and

judg

men

t use

d to

fill

in th

e ga

ps in

the

data

D

egra

datio

n Va

lue

tech

niqu

e•

Ass

igns

prob

abili

ties

forl

ikel

ihoo

dan

even

twas

anac

tual

CC

FA

ssig

ns p

roba

bilit

ies

for l

ikel

ihoo

d an

eve

nt w

as a

n ac

tual

CC

F ev

ent

2626

Dat

aA

naly

sis

Proc

ess

Dat

a A

naly

sis

Proc

ess

C

lass

ifica

tion

exam

ple

Cla

ssifi

catio

n ex

ampl

ePl

ant T

ype

(Dat

e)Ev

ent D

escr

iptio

nIm

pact

Vec

tor

Com

pone

ntG

roup

Siz

eG

roup

Siz

e

To

mot

ordr

ien

AFW

pm

ps

PP 11

PP 22

00

PP

PWR

((12/

73))

Two

mot

or-d

riven

AFW

pum

psw

ere

inop

erab

le d

ue to

air

inco

mm

on s

uctio

n lin

e22

0000

11

D

ata

typi

cally

col

lect

ed in

clud

e

Com

pone

nt g

roup

siz

e

Num

ber o

f com

pone

nts

affe

cted

S

hock

type

(let

hal v

s. n

on-le

thal

)

Failu

re m

ode

2727

Dat

aA

naly

sis

Proc

ess

Dat

a A

naly

sis

Proc

ess

2828

Dat

aA

naly

sis

Proc

ess

Dat

a A

naly

sis

Proc

ess

C

CF

Bas

icE

vent

Qua

ntifi

catio

nC

CF

Bas

ic E

vent

Qua

ntifi

catio

n

M

onte

Car

lo m

etho

ds c

an b

e us

ed to

pro

paga

te

unce

rtain

ties

N

ote

that

sta

te o

f kno

wle

dge

depe

nden

ce is

lost

if th

e C

CF

basi

cev

ents

{AB

}{A

C}

{BC

}an

d{A

BC

}are

CC

F ba

sic

even

ts {A

B},

{AC

}, {B

C},

and

{AB

C} a

re

treat

ed a

s in

depe

nden

t bas

ic e

vent

s

2929

CC

F Ev

ent C

lass

ifica

tion

dA

li

San

d A

naly

sis

Sum

mar

y

3030

The

End

Que

stio

ns &

Ans

wer

s…..

3131