Você está na página 1de 15

Microelectron. Reliab., Vol. 33, No. 4, pp. 583-597, 1993. 0026-271419356.00+.

00
Printod in Great Britain. © 1993 Pergamon Press lad

FUZZY FAULT-TREE ANALYSIS USING FAILURE POSSIBILITY


GIN-SHUH LIANG and MAO-JIUN J. WANG
Department of Industrial Engineering, National Tsing Hua
University, Hsinchu, Taiwan 300, R.O.C.
(Receivedfor publication 25 October 1991)

Abstract - In conventional p r o b a b i l i t y - b a s e d f a u l t - t r e e analysis~ the vague


property in many systems, e s p e c i a l l y , the man-machine system i s often ignored.

To overcome t h i s disadvantage, the fuzzy set theory is applied to f a u l t - t r e e

analysis. The f a i l u r e p o s s i b i l i t y defined by a t r i a n g u l a r fuzzy number on the

interval [0,1] is used to characterize the p o s s i b l e d e v i a t i o n of the basic

events. The fuzzy operations according to the extension p r i n c i p l e is used to

c a l c u l a t e the f a i l u r e p o s s i b i l i t y of the top event. Furthermore, a fuzzy


importance index is proposed to demonstrate the contribution of a basic event

to the s a f e t y improvement of the top event in a fuzzy environment.

1. I n t r o d u c t i o n

F a u l t - t r e e a n a l y s i s is an e f f e c t i v e t o o l to evaluate system s a f e t y and


reliability. A f a u l t - t r e e is a l o g i c a l and graphical d e s c r i p t i o n of various
combinations of f a i l u r e events. If the f a i l u r e r a t e s of system components are

given, the f a i l u r e rate of the top event can be c a l c u l a t e d by means of


a l g e b r a i c operations. In conventional f a u l t - t r e e a n a l y s i s , the f a i l u r e r a t e s

of system components are given in c r i s p values. In Terano et a l . [19] and

Henneman and Rouse [9] papers, they indicated that human performance is
a f f e c t e d by many Performance Shaping Factors [13,17], e.g. s t r e s s l e v e l , task

knowledge and experience, congnition a b i l i t y , and etc. In f a c t , because of


the u n p r e d i c t a b i l i t y of human performance [14~ as well as the vagueness of

system phenomena, it is d i f f i c u l t to evaluate the exact f a i l u r e r a t e s of


system components, e s p e c i a l l y in man-machine systems. In other words, the

crisp approach tends to have d i f f i c u l t y in conveying the imprecision or


vagueness nature in system modeling to represent the f a i l u r e r a t e of system
component.
To overcome t h i s d i f f i c u l t y , the fuzzy set theory is applied to evaluate
system r e l i a b i l i t y under fuzzy environment. Thus, instead of using the
p r o b a b i l i s t i c r e p r e s e n t a t i o n of f a i l u r e r a t e , the f a i l u r e p o s s i b i l i t y [14,23]

583
584 G.-S. LIANGand M.-J. J. WANG

is proposed to c h a r a c t e r i z e the f a i l u r e occurrence of system events. In t h i s

study, the t r i a n g e l a r fuzzy numbers which is defined on a i n t e r v a l [0,1] are

used to denote the f a i l u r e p o s s i b i l i t i e s . I t can not only convey the most

p o s s i b l e f a i l u r e r a t e of system components but also r e f l e c t the d i s p e r s i o n of

the e v a l u a t i o n data. The f a i l u r e p o s s i b i l i t y can e f f e c t i v e l y r e p r e s e n t the

v a r i a t i o n of human performance, the vagueness of system phenomena as well as

the degree of stability of machine components. In addition, the fuzzy

m u l t i p l i c a t i o n and s u b t r a c t i o n according to the extension p r i n c i p l e [22] are


used to calculate the failure possibility of the top event.

To effectively evaluate the importance of different basic events, a fuzzy


possibility-based importance index is used to represent the contribution of a

basic event to the occurrence of the top event. ~eanwhile, for the

calculation of fuzzy importance index, a ranking method of triangular fuzzy

numbers based on the concept of the maximizing set and minimizing set is

developed.

In the following sections, the notions and concepts of the fuzzy set

theory will be briefly introduced. Further, failure possibility, fuzzy fault

tree analysis [7,16] and fuzzy importance are presented. Finally, a numerical

example is given to illustrate the method.

2. Fuzzy set theory

Fuzzy set theory was introduced by Zadeh [22] to deal with problems in

which the fuzzy phenomena is present. In a universe of discourse X, a fuzzy

subset A of X is characterized by a membership function fA(x) which associates

with each element x in X a real number in the interval [0,I]. The function

value fi(x) represents the grade of membership of x in A. The larger the

fA(x), the stronger the degree of belongingness for x in A.

2.1. Fuzzy numbers

A fuzzy number :~5] is described as a subset of real line (~) whose

membership function f is a continuous mapping from ~ to a closed interval

[0,ii, which has the following characteristics: (I)f(x)=O for all x E (-m, c i

U [d, ®), (2) f(x) is strictly increasing on [c, a I and strictly decreasing on

[b, d], and (3) f(x)=l for all x E [a, b]. Eventually, it can be c=-® or c=a

or a=b or b=d or d=®. The fuzzy numbers are very useful in promoting the
representation and information processing under fuzzy environment. In here, a

special class of fuzzy numbers, called triangular fuzzy numbers, is applied.


Fuzzy fault-tree analysis 585

A fuzzy number A in 2 (real line) is a triangular fuzzy number, if its

membership function fk : 2 ~ [0,1] is equal to


(x- c ) l ( a - c), c<x<a,
fl(x) ={ (x- b ) / ( a- b ), a < x < b,
O, otherwise.
with c < a < b. The t r i a n g u l a r fuzzy number can be denoted by k = (c, a, b).

2.2. The operations of fuzzy numbers


Zadeh introduced the extension principle to find the membership function
after mapping fuzzy sets through a function. The extension principle is

defined in the following.


Definition I. Let f be a mapping from X I * X 2 to a universal set Y such
that y = f(xl,x2). Given fuzzy numbers A, and 12 are in X l and X 2 with
membership functions fA~ and fA2 respectively. The membership function of the

fuzzy number B = f(k I,A2) is defined as:

s~p min(fA (x~), fA2(x2)), if f"(y) ~ ¢,


Xl ,X2
fB(Y) = Y=f(xl,x2)
o, if f-'(y) -- ¢.
By the extension principle described as above, the following theorems are

true.

fheorem 1. The fuzzy addition of two t r i a n g u l a r fuzzy numbers k i = (ci,

ai, bi) and Aj = (cj, aj, bj), i.e. Ai ® Aj, is also a t r i a n g u l a r fuzzy

number. And I i m Aj = (ci+cj, ai+aj, bi+bj) [21].


For the subtraction of triangular fuzzy numbers, suppose that fA is the
membership function of a triangular fuzzy number A. By the extension

principle, the membership function g of the opposite of A is defined as:


g(y) = sup f(x) = f ( - y ) .
y=- X

Based on t h i s d e f i n i t i o n , the following theorem can be obtained.

fheorem 2. The negation of a t r i a n g u l a r fuzzy number Ai = (ci, ai, bi) is


also a t r i a n g u l a r fuzzy number. And e Ai = ( - b i , - a i , -c i ) .
By transforming the subtraction of Ai e Aj into Ai • ( e Aj) and using
theorem I , theorem 3 can be obtained.
fheorem 3. The fuzzy subtraction of any two t r i a n g u l a r fuzzy numbers Ai =
(Ci, ai, bi) and Aj = (cj, aj, bj), i.e. A i o Aj, is also a triangular fuzzy
number. And A i o Aj = (ci-bj, ai-aj, bi-c j ). Especially, 1 o Aj = (I, I, i)
O (Cj, aj, bj) = (1-bj, 1-aj, 1 - c j ) .
fheorem ¢. The fuzzy m u l t i p l i c a t i o n of any two t r i a n g u l a r fuzzy numbers Ai
586 G.-S. LIANGand M.-J. J. WANG

: (Ci, ai, bi) , c i _> O, and Aj = ( c j , a j , bj , cj > O, i.e. A i e Aj, gives the
following membership f u n c t i o n :

fA i ® Aj(X) = { -HI + [ H~ + ( x -

H2 - [ H~ + ( x - Z )
O,
Y ) Tt ] 1/2,
UI ] 1/2,
Y < x _< {],
O. _< x _< Z,

otherwise.
where

T1 = (a i - c i ) (aj-cj), T2 : c i ( a j - c j ) + c j ( a i - c i ) ,
Ut = (b i - a i) ( b j - a j ) , U2 = bj ( a i - b i ) + b i ( a j - b j ),
HI = T 2 / 2T~, H2 =- U2 / 2UI,
Y = CiCj, Q = aiaj, Z = bib j .
Proof: By the extension p r i n c i p l e , the membership f u n c t i o n of the fuzzy
number Ai ® Aj can be represented by

E~supmin (fAi(xi), fAj (xj))


fAi ® Aj (x) =xi,xj
X=XiXj

= xiE~sup min (fAi(Xi), fAj (X/Xi))

Using the same concepts in Dubois and Prade [5], we know t h a t V w E [ 0 , 1 ] , 3


xi, xj E ~, s.t. w = fAi(xi) = fAj(xj). Thus,

c o n s i d e r i n g the i n c r e a s i n g p a r t s ( i . e . x E [cic j , a i a j ] ), we can get

x = xix j = w 2 ( a i - c i ) ( a j - c j ) + wcj ( a i - c i ) + wci(aj-cj) +cic j (1)


S i m i l a r l y , f o r the decreasing p a r t s ( i . e . x E [aia j , bibj] ),

x = xix j = w 2 ( a i - b i ) ( a j - b j ) + wbj ( a i - b i ) + wbi(aj-bj) + bib j (2)


By s o l v i n g equations (1) and (2), the membership f u n c t i o n of Ai e Aj (w = fAi

® Aj (x)) is obtained.

I t is obvious t h a t Ai ® Aj is not a t r i a n g u l a r fuzzy number. However,


the approximation formula can be deduced by taking w2 _'2 w in equations (1) and
(2), i f w is in the neighbourhood of 1, and

Ai ® Aj ~ (CiCj, a i a j , bibj).
To make the c a l c u l a t i o n easy, the approximate fuzzy m u l t i p l i c a t i o n of two
t r i a n g u l a r fuzzy numbers is defined.

Definition 2. Let Ai = (ci, a i , bi) , c i _> O, and Aj = ( c j , a j , b j ) , cj _>


O, the approximate fuzzy m u l t i p l i c a t i o n of Ai and Aj is

Ai @ Aj : (CiCj' a i a j , bibj)"

2.3. R~.king f u z z y mmbers v i t h the maximizing and miaimizing s e t s


The ranking of fuzzy numbers is important for determining the fuzzy
importance of b a s i c events. Many methods of ranking fuzzy numbers have been
Fuzzy fault-tree analysis 587

proposed [ 1 , 2 , 3 , 8 , 1 0 ] . The disadvantages of some of the methods have been

reported in Bortolan and Degani [1], and Chen's [3] papers. For s i m p l i c i t y of

c a l c u l a t i o n and e f f e c t i v e n e s s in problem solving, the concept based on the


Chen [3] and Kim and Park [10] is used tO rank fuzzy numbers.
To consider the optimistic and pessimistic view points of the
decision-makers, the maximizing and minimizing sets are f i r s t defined [10].
Oefini~ion 3. Let Ai, i = 1, 2, . . . , n , be n fuzzy numbers with membership
functions fAi respectively. The maximizing set M is defined as I[ = {

(~,f~(x)) I x e ~ ) ~-ith
(x- ~) / (~2- x~), x~ [~,x~],
fM(x) = O, otherwise,
and the minimizing set G is defined as G = { (x,fG(x)) I x E 2 } with
(x- ~) / (~- x~), x~ [x~,x~],
fG(x) = O, otherwise,
I|
where x~ = in]I), x 2 = supD, D =xUxDi= and Di = { x I fAi(X) > 0 }, i = 1, 2,

..., n.

Definition ~. The right ranking value UM(Ai) and the l e f t ranking value
UG(Ai) of each fuzzy number Ai are defined as
~(A~) = s~p (fA~ (~) ^ f l (x))
X

and
UGCA~) = 1 - sup (fACx) ^ fcCx)/,
X

where a stands f o r the minimum operation and i = 1, 2, . . . , n .


Oe/ini~ioa 5. The ranking value of fuzzy number Ai, R(Ai) , is defined as

R(Ai) = k UM(Ai) + ( 1 - k ) UG(Ai) , i = 1, 2, . . , n ; 0 < k'< 1.


In d e f i n i t i o n 5, the value of k is an index of optimism. I t can be
manipulated to reflect the decision-maker's optimistic or pessimistic
attitude. I f k = 1, i t implies that the decision-maker is r a t h e r o p t i m i s t i c .
I f k = O, i t implies that the decision-maker is rather p e s s i m i s t i c .

Oefini~ion 6. Let Ai and Aj be two fuzzy numbers with membership functions

fAi(X ) and fAj (x) r e s p e c t i v e l y . And l e t

m~ = (~i.{XlfACX)=l}÷~.~{~lfACX)=l})/2,
and

mj = (mi.{xlfAj (x)=l}+max(x[fAj( x ) = l } ) / 2 ,
define the fuzzy ranking of Ai and Aj as:
(i) A£ > Aj ¢=# R(A£) > R(Aj).
(2) A i > Aj ¢~ a(Ai) = R(Aj), m i > mj.
588 G.-S. LtANG and M.-J. J. WANG

(3) Ai = Aj ¢=~ R(Ai) = R(Aj), m i = mj.

By definition 5, theorem 5 can be obtained.

Theorem 5. Let k i = (ci, a i , bi) , i : 1, 2, . . . , n, be n t r i a n g u l a r fuzzy

numbers, then the ranking values of these t r i a n g u l a r fuzzy numbers are:

R(ii) = k [(b i - xl) / (x= - x I - a i + bi) ] + (i - k) [i

-(X 2 - el) / (X2 - Xt + a i - Ci)],

where x 1 = min{ct, % , ..., %} and x~ = maz{bt, b~, . . . , b n } .

By theorem 5 and taking the s u i t a b l e k value (e.g. k = 0 . 5 ) , the ranking

values of the n t r i a n g u l a r fuzzy numbers can be c a l c u l a t e d . Then based on

d e f i n i t i o n 6, the ranking of the n t r i a n g u l a r fuzzy numbers can be e a s i l y

determined. Theorem 5 is useful in c a l c u l a t i n g the fuzzy importance i n d i c e s

of the basic events, and these w i l l be discussed in s e c t i o n 5.

3. F a i l u r e p o s s i b i l i t y

The concept of f a i l u r e p o s s i b i l i t y i s applied to replace f a i l u r e r a t e

( p r o b a b i l i t y ) in f a u l t - t r e e a n a l y s i s . The f a i l u r e p o s s i b i l i t y can be defined

by a fuzzy number on i n t e r v a l [0,1], with any shape of membership f u n c t i o n .

In here, the t r i a n g u l a r fuzzy numbers defined on the i n t e r v a l [0,1] is used to

represent the f a i l u r e p o s s i b i l i t i e s of the basic events. The reason of using

the triangular fuzzy numbers is because it is intuitively easy for the


decision-makers to perform e v a l u a t i o n . It is very u s e f u l in conveying the

v a r i a t i o n of human performance, the vagueness of system phenomena as well as

the stability of machine components in man-machine systems.

The failure possibility of a basic event E i denoted as PEi= (ci, ai, bi) ,

0 < c i < a i < b i < i (see Fig. I). The parameter a i gives the maximal grade

of fp (x), i.e. fe (ai)=l. It is the most possible value of failure rate.


Ei Ei
The c i and b i are the lower and upper bounds of the d i s p e r s i o n region of the

fPE~11..... ~ 1
0 ci ai bi

Fig. 1. The graph of membership f u n c t i o n fp •


Ei
Fuzzy fault-tree analysis 589

possible f a i l u r e rates. I t can be manipulated to r e f l e c t the f u z z i n e s s of the

e v a l u a t i o n data. For example, the f a i l u r e p o s s i b i l i t y of an operator is

approximately equal to 0.01 can be represented by (0.008, 0.01, 0.012). For a

f u z z i e r r e p r e s e n t a t i o n , i t can be (0.005, 0.01, 0.015). The non-fuzzy number

"a" can be expressed as (a, a, a).

Many p r e c i s i o n - b a s e d techniques for human f a i l u r e (error) r a t e p r e d i c t i o n

have been developed [6,12,13,15]. The human error rate estimated by the above

methods or obtained from the R e l i a b i l i t y Data Bank [4,20] can be considered as

the most p o s s i b l e y a l u e a i . In order to account for the u n p r e d i c t a b i l i t y of

human performance as well as the vagueness nature of system phenomena, the

lower and upper bounds (c i and bi) are s p e c i f i e d . Two approaches can be

applied to obtain the lower and upper bounds: (I)the use of guideline table
(e.g. Table I) [18], e.g. the failure rate ai=O.O06 , then the failure

possibility of basic event PEi = (0.002, 0.006, 0.018). (2) the direct

assignment based on expert's assessment.

In addition, the fuzzy-based technique for evaluating human reliability

developed by Liang and Wang [II] can also be used to estimate the human
failure possibility.
For the failure possibility of machine equipment, since the failure
probability of machine components can be estimated from the historical data,
the failure possibility of machine components can also be expressed as (c, a,
b) in which "a" is the most possible value, "c" and "b" are the upper and
lower bounds. The higher the degree of stability of machine components, the

smaller the values of Ic-a[ and [b-a]. In addition, if the reliability data
of machine components are not available, the failure possibility of the

Table I. The guideline for estimating the lower and


upper bound of the estimated failure rate.

Failure rate Lower bound Upper bound

0.01 < Fr Fr / 5 2 ,, F r - 5 ~ Fr

0.001 < Fr < 0.01 Fr / 3 3 ~ Fr

F r < 0.001 F r / 10 10 x F r
590 G.-S. LIANG and M.-J.J. W A N G

machine components can be subjectively evaluated by experts using the


triangular fuzzy numbers.

4. Fuzzy f a u l t - t r e e analysis
This section is mainly concerned with the fuzzy evaluation of the failure
possibility of the top event in a f a n l t - t r e e . By expressing the logical
operators AND and OR in fuzzy addition and multiplication, the failure
p o s s i b i l i t y of the top event can be calculated.
Theorem 8. Let PEi = (ci, ai, bi) and PEj = (cj, aj, bj), and take k=0.5,
the ranking values of the fuzzy numbers PEi ® PEj and PEi ® PEj are:

R(PEi® PEj) = [(xR+xL-2Y)/(Z-Y)]/2'

and
R(PEi® PEj) = [(x~+x~-2Y)/(Z-Y)]/2,

vhere
XR={2Y+2H~(Z-Y)+(Z-Y)2/Ut-(Z-Y)[(2~2+(Z-Y)/Ut)2+4(Y-Z)/U1]L/2}/2,
XL={2Z+2Ht(Z-Y)+(Z-Y)2/Tt-(Z-Y)[(2HI+(Z-Y)/TI)2+4(Z-Y)/T~II/2}/2,
x~t=(Z2- Yq) / (2Z- Y-q),
x~=(Y2- Zq)/(2Y- Z- q),
Tt=(ai-ci)(aj-cj), T2=Ci(aj.cj)+cj (ai-ci) ,
Ut=(bi-ai) (bj-aj ), U2=bj (ai-b i )+b i (aj-bj ),
Hx=T~/2T 1 , ll~--- U~/2UI,
Y=Ci cj , q=a i aj , Z=bib j .

t
~I'-~ .... rr--.~ ....
', ~ e O , , ' / r ~ / ;/ PE~ePE]
f

I
0 Y
!is . ,, - ,,~ "

Q
x
z
, PEi ® PE i

1
L

Remark: --e- : 1 - UG(PEi0 PEj)


--o-- : 1 - UG(PEi® PEj)
--oo-- : UM(PEi®PEi)
--®®-- : UM(PEi®PEj)

Fig. 2. The graphical i l l u s t r a t i o n of PEi® PEj < PEi ® PEj •


Fuzzy fault-tree analysis 591

According to theorem 6 and definition 6, theorem 7 can be obtained.

Theorem 7. PEi ® PEj < PEi ® PEj .

In here, a simple graphical presentation to i l l u s t r a t e the relation in


theorem 7 is given in Fig. 2. Since, UM(PEi® PEj ) > UM(PEi® PEj ) and

1-UG(PEi® PEj) > 1-UG(PEi® PEj)' therefore R(PEi® PEj) > R(PEi® PEj)' and thus

PEi ® PEj < PEi ® PEj

Since by Theorem 4, PEi ® PEj is not a triangular fuzzy number, the

implementation of fuzzy multiplication of two triangular fuzzy numbers is


going to be tedious and time-consuming. On the other hand, by Theorem 7, the
approximate fuzzy multiplication ® tends to over-estimate the failure
possibility, thereby resulting in a conservative estimation of system safety.
For simplicity of calculation, and without losing the effective estimation of
system safety, the fuzzy operation ® is used to calculate the failure
possibility of the top event.
By u t i l i z i n g the fuzzy operation ®, the failure p o s s i b i l i t i e s under AND
and O& operators can be obtained:
(1) the f a i l u r e possibility under the n-ary AND operator (Fig. 3) is

P~ = PE1 ® PE2 ® . . . ® PEn


(2) the f a i l u r e possibility under the n-ary OR operator (Fig. 4) is
P~ = ~ e (~ e PEI ) • (~ ~ PE~) ° "'" ° (~ ~ PE°)
where 1 = (1, 1, 1).
Fig. 5 i l l u s t r a t e s a simple f a u l t - t r e e which involves two types of event
symbols and two types of gates. The top event (T) can be expressed as:
T = E t n A o E5

= El n [ E 2 u E 3 u E~ ] n E S.

Assume that the failure possibilities of the five basic events are:

Fig. 3. 'Uh~D" g a t e .
592 G.-S. LLANOand M.-J. J. WANG

Fig. 4. "OR" g a t e .

PEt,PE,PE3,PE4 and PEs. Then the failure possibility of the top event T is:

PT = PEt ® [ie (1 e PE2 ) ® (i e PEa ) ® (i e PE4)] ® PEa (31

5. Fuzzy importance
In order to evaluate the contribution of different basic events to the
top event, a possibility-based importance index, called fuzzy importance
index, i s proposed.

Let PT denote the failure possibility of the top event (T) of a

fault-tree which involves n basic events. And PTi denotes the failure

possibility of the top event (T) in which the failure possibility of the i-th
basic event is eliminated. Since the failure possibility of the top event is
an increasing function with respect to the failure possibility of any basic

Fig. 5. An example of a f a u l t - t r e e .
Fuzzy fault-tree analysis 593

event, hence PT > PTi" Based on this property, the importance index

D(PT,PTi), which assesses the difference between PT and PTi , is used to

indicate the degree of improvement a f t e r eliminating the f a i l u r e occurrence of


the i - t h basic event Ei.
Definition 7. The fuzzy importance index of a basic event Ei, i : 1, 2,

. . . , n, D(PT,PTi), is determined by the ranking value of PT e PTi:

D(pT,PTi ) = R(P T e PTi ).


Since PT e PTi is also a t r i a n g u l a r fuzzy number, by taking a suitable

optimistic index k (e.g. k = 0.5), the fuzzy importance indices D(PT,PTi ) of

a l l the basic events can be obtained. And the p r i o r i t y of importance of the


basic events can be determined. The larger the D(PT,PTi), the more important

the basic event.

6. Numerical example
To demonstrate the fuzzy f a u l t - t r e e analysis method developed in this
paper, a numerical example is presented. The f a u l t tree is shown in Fig. 5.
Assuming that the f a i l u r e possibilities of the basic events are evaluated.
They are shown in Table 2.

Table 2. The f a i l u r e p o s s i b i l i t i e s of the basic events

Basic events Failure possibility

( Ci, a i , b i )

E1 ( 0.460, 0.500, 0.550 )

E2 ( 0.008, 0.010, 0.015 )

E3 ( 0.030, 0.050, 0.070 )

E4 ( 0.030, 0.050, 0.070 )

E5 ( 0.370, 0.400, 0.420 )

By Eqn (3), the f a i l u r e p o s s i b i l i t y of the top event i s :


PT = (0.0113,0.0213,0.0342). And the f a i l u r e p o s s i b i l i t y of the top event (T)
when the f a i l u r e p o s s i b i l i t y of the f i r s t basic event is eliminated, can be
computed as:

MR 33/4---J
594 G.-S. LIANG and M.-J. J. WANG

PTI = 0 ® E 1 e ( I e PE2 ) ® ( 1 e PE3 ) ® ( 1 e PE )I ® PEs

= (0,0,0), where 0 = (0,0,0).


Further, the faihre possibility of the top event (T) when the failure
possibility of the second basic event is eliminated, can be computed as:

PT2 = PE1 ® [ 1 e ( 1 e PEz ) ® ( 1 e PE4)] ® pEs

= (0.46,0.5,0.55)®[(1,1,1)e(0.S649,0.9025,0.9409)]®(0.37,0.4,0.42)

- (0.46,0.5,0.55)®(0.0591,0.0975,0.1351)®(0.37,0.4,0.42)

= (0.0101,0.0195,0.0312)o

Similarly, the remaining PTi,i = 3,4,5, can also be obtained:

PT3= PT4 = (0.0064,0.0119,0.0194), PTs = (0,0,0).

By d e f i n i t i o n 7 and taking k = 0.5 in theorem 5, the fuzzy importance

indices of the five basic events can be obtained. Table 3 presents the
calculated results. And the ranking order of the five basic events is

{E1,Zs} , {E3,E,} and E 2.

Table 3. The fuzzy importance index of basic event

Ei ( i = 1, 2, 3, 4, 5 )

Basic event Fuzzy importance index

EI 0.7251

E2 0.4311

E3 0.5336

E4 0.5336

Es 0.7251

Based on the values of D(PT,PTi ) in Table 3, it i s suggested t h a t the

basic event E1 or Es w i l l have the l a r g e s t c o n t r i b u t i o n to the improvement of

f a i l u r e p o s s i b i l i t y of the top event. I f the f a i l u r e p o s s i b i l i t y of E1 or Es

is changed to O, then the failure possibility of the top event can be

eliminated (i.e. PTI=PTs=(O,O,O)).

Similarly, i f the f a i l u r e p o s s i b i l i t y of E3 or E4 is e l i m i n a t e d , the

failure possibility of the top event is reduced to


Fuzzy fault-tree analysis 595

PT3=PT4=(O.O064,0.OI19,O.0194). And if the failure possibility of E 2 is

eliminated then the failure possibility of the top event becomes


FT2=(0.0101,0.0195,0.0312).

7. Conclusion
From the results of this study, the following conclusions are drawn:
(1) Instead of using the failure probability, the failure possibility is

used to characterize the failure occurrence of the system events. It can


effectively express the vagueness nature of system phenomena, the human
performance differences as well as the degree of stability of machine
components.
(2) The use of triangular fuzzy numbers is very. effective in representing
the failure possibility of the basic event under fuzzy environment.
(3) The fuzzy importance index which is based on the level of failure
possibility reduction of the top event after eliminating the failure

occurrence of the basic event, can provide useful information to effectively


improve the overall system safety.
(4) When the fault-tree is constructed, the calculation of the failure
possibility of the top event and the importance evaluation of the basic event

can be computerized. Thus after providing failure possibilities of all basic


events, the decision-makers can obtain the failure possibility of the top
event and the priority of importance of the basic events automatically.
leferences

1. G. Bortolan and R. Degani, A review of some methods for ranking fuzzy


subsets. Fuzzy Sets and Systems, 15, 1-19 (1985).
2. L.M. Cumpos and A. Gonzalez, A subjective approach for ranking fuzzy
numbers. Fuzzy Sets and Systems, 29, 145-153 (1989).
3. S.H. Chen, Ranking fuzzy numbers with maximizing and minimizing set. Fuzzy
Sets and Systems, 17, 113-129 (1985).
4. M.K. Comer, M.D. Donovan and C.D. Gaddy, Human r e l i a b i l i t y data bank :
Evaluation results (General Physics Corporation and Sandia National
Laboratories, NUREG/C~4009). Washington, DC : U.S. Nuclear Regulatory
Commission (1985).
5. D. Dubois and H. Prade, Operations on fuzzy numbers. International Journal
of Systems Science, 9, 613-628 (1978).
6. D.E. Embrey, P. Humphreys, E.A. Rosa, B. Kirwan and K. Rea,, SLIM-MAUD :
An approach to assessing human error p r o b a b i l i t i e s using s t r u c t u r a l expert
596 G.-S. LIANGand M.-J.J. WANG

judgment, vol. i, overview of SLIM-MAUD, March; vol. 2, Detailed analysis


of the technical issues (draft), March; (NUREM/CR-3518). Washington, DC :
U.S. Nuclear Regulatory Commission (1984).
7. H. Furuta and N. Shiraishi, Fuzzy importance in fault tree analysis. Fuzzy
Sets and Systems, 12, 205-213 (1984).
8. A. Gonzalez, ~ study of the ranking function approach through mean values.
Fuzzy Sets and Systems, 35, 29-41 (1990).
9. R.L. Henneman and W.B. Rouse, Human performance in monitoring and
controlling hierarchical large-scale systems. IEEE trans. Syst. Man
Cybern., 14, 184-191 (1984).
10. K. Kim and K.S. Park, Ranking fuzzy numbers with index of optimism. Fuzzy
Sets and Systems, 35, 143-150 (1990).
II. G.S. Liang and M.J.J. Wang, Evaluating human reliability using fuzzy

relation. Microelectronics and Reliability (accepted).


12. N.C. Lind, Models of human error in structural reliability. Structural

Safety 1, 167-175 (1983).


13. D.P. Miller and A.D. Swain, Human error and human reliability in G.
Salvendy (Ed:), Handbook of human factors, John Wiley ~ Sons, Inc.,

New York (1987).


14. T. Onisawa, &n approach to human reliability in man-machine systems using

error possibility. Fuzzy Sets and Systems, 27, 87-103 (1988).


15. D.A. Seaver and W.G. Stillwell, Procedures for using expert judgment to
estimate human error probabilities in nuclear power plant operations
(Decision Science Consortium and Sandia National Laboratories NUREG/CR
2743). Washington, DC : U.S. Nuclear Regulatory Commission (1983).
16. D. Singer, A fuzzy set approach to fault tree and reliability analysis.

Fuzzy Sets and Systems, 34, 145-155 (1990).


17. A.D. Swain, Field calibrated simulation. In proceedings of the symposium
on human performance quantification in systems effectiveness. Washington
DC: Naval Material Command and the National Academy of Engineering (1967).
18. A.D. Swain and H.E. Guttmann, Handbook of human reliability with emphasis
on nuclear power plant applications, NUREG/CR-1278 (2nd draft) (1980).
19. T. Terano, Y. Murayama and N. Akiyama, Human reliability and safety

evaluation of man-machine system. Automatica, 19, 719-722 (1983).


20. D.A. Topmiller, J.S., Eckel and E.J. Kozinsky, Human reliability data bank
for nuclear power plant operations, vol. i: A review of existing human
reliability data banks (General Physics Coporation and Sandia National
Fuzzy fault-tree analysis 597

Laboratories, NUREG/CR-2744). Washington, BC : U.S. Nuclear Regulatory

Commission (1982).

21. P.J.M. van Laarhoven and W. Pedrycz, A fuzzy extension of Saaty's p r i o r i t y

theory. Fuzzy Sets and Systems, 11, 229-241 (1983).


22. L.k. Zadeh, Fuzzy sets. Information and Control, 8, 338-353 (1965).
23. L.A. Zadeh, Fuzzy sets as a basis for a theory of p o s s i b i l i t y . Fuzzy Sets

and Systems, 1, 3-28 (1978).

Você também pode gostar