This item was submitted to Loughborough's Research Repository by the author.

Items in Figshare are protected by copyright, with all rights reserved, unless otherwise indicated.

Fault Tree Analysis and Binary Decision Diagrams

PLEASE CITE THE PUBLISHED VERSION

PUBLISHER

© Institute of Electrical and Electronics Engineers (IEEE)

LICENCE

CC BY-NC-ND 4.0

REPOSITORY RECORD

Sinnamon, Roslyn M., and J.D. Andrews. 2008. “Fault Tree Analysis and Binary Decision Diagrams”. figshare.
[Link]
 This item was submitted to Loughborough’s Institutional Repository by the
author and is made available under the following Creative Commons Licence
conditions.

For the full text of this licence, please go to:

[Link]
 Fault Tree Analysis and Binary Decision Diagrams
Roslyn M. Simalnoli Loughborough University Loughborough
John D. Andrews Loughborough University Loughborough

Key Words: Fault Trees, Biuwy Decision Diagrams

SUMMARY 8 CONCLUSIONS approach using a Binary Decision Diagram may provide a

h t e r :md more efticient means of aindysing tault trees.
Fault tree malysis is now commonly used to ;issess the
adequacy, in reliability terms, of inclustri:tl systems. For 2. NOTATION
complex systems ;in m;ilysis may produce thousands of
combinations of events which c;in c;iuse system failure P( Top) - Probability ot Top Even1 o f ;i hult tree.
(minimal cut sets). The deterinination of these minimal cut
sets c m be a very time consuming ,process even on inodeni Ci - Minirnal ('ut Set.
high speed digital computers. Also it the fault tree h:is inmy PRE(Top) - R:re Event Approximation of Top Event
minimal cut sets calculating the exact top event prob:ibility Prohability.
will require extensive calcul;ilions. For many complex fault
trees this requirement is beyond the capability of Lhe iivailaible
P ~ c , s u ~ ( T o p- )Minirnal Cut Set Upper Bound of Top
machines, thus appro xi in a t i on t ec hi1i y u es nee cl to be Event Probability.
introduced resulting in loss of accuracy. Xi - Boolean Variable.
f(x)/fl/f2 - Boolean Functions.
This paper describes the use of ;I Binary h c i s i o n 1)i:igr:un for ire - If-Then-Else structure for Binary
Fault Tree Analysis m c l soine ways ill which it c;m be Ikcision 1)Bcgcun.
efficiently iinpleiiiented on :L cornputer. The work to date <0p> - Boolem operation ( . 01-+).
shows :i substmtial improvement in computation;il effort for Fi - Nocles/Vertices in IIBinary 1)ecision
large, complex fault trees an;ilysetl with this method in I>ingr:un.
comp:rison to the tr:iditionad approach. The Binxy Decision - Probability of occurrence of top event of
Diagram method has the additional ndvantage that Q,sy
;ipproximalions are not required. exact cdcu1;itions lor the top frlult tree.
event pw:uneters C;LII be perfonned. W(0, t) -Expected number of top event
()ccU rrences.
1. INTRODUCTION ~"sy.v - System T TnconclitionalFailure Intensity.
Gi (Y) - (Iriticality Function for component i
The fault tree diagram itself is an excellcilt way of deriving the
failure logic for ;i system auld representing it in ;I form which is
ideal for communic~itioIi to other managers/ 3. ARRREVIATION,S
clesigners/operiitors etc. The f;iult tree is discussed in cletail iii
Andrews and Moss (Ref. 1). Since the method was first s.0.p - s u m of products expression
conceived in the early sixties, algorithms to dcrive the minimal BDI) - Binary 1)ecisioii 1)i:igram.
cut sets have worked directly with the fault tree diagram itself
using either bottom-up, Seinanderes (Ref. 2). or top-down, 4. FAULT TREE ANALYSIS
Fussell and Vesely (Ref. 3 ) , appro:ichcs. C'ompulerised
methods to conduct this :ui;tlysis w e iiow so well clevelopecl The analysis of the fault tree is generally unclert:ken in two
that further refinement is unlikely to resul~in v x t rcductions stages: qualitative analysis and quantitative analysis.
in computer time. Tackling this problem t o impi-ove Qualitative nnalysis involves obtaining the various
computational efficiency has heell the miiiii conceni over the combinations of events which c;iuse system ftiilure (minimal
yews for many fault tree rese:uchers, Bennetts (Ref. 4) and cut sets) and yu;uitificatioii then cleats with calculating the
Bengiamiii et al. (Ref. 5 ) hnve both aclclre probability or freyuency that system failure will occur.
I~Jsua11y by modify i 11g t h e est ab 1i sh ecl , con ve 11ti on a 1
approaiclies such :is MOCT JS (Ref. 3).

It is felt that substmtial improvement in computer utilisation The conventional :ipproach to obtain the minimal cut sets is to
will only result from ;I completely new :ippro:ich. Such :in take the Boolean logic expression for the Top Event and
approach would involve spccifyiilg the logic cyu:ilion in ;I transform it into ;i sum of products (s.0.p) form. (.)ne way of
form which is easier to 1nanipu1;iteth;tn ;I tau11 wee. A recent doing this is to use ;I Bottom-'I Jp procedure such ;is that of
p:tper by Riuzy (Ret. 6) tias iiiclicatecl that ;in nllcm;itive Sem:inderes (Ref. 2). To obtain the w . p form for the Top
Event of the f:tult tree, the inputs to the lowest gates ; r e

0-7803-3112-5/96/$5.00 0 1996 IEEE

1996 PROCEEDINGS Annual RELIABILITY and MAINTAINABILITY Symposium 215
represented ais logic equations. Once the lowcr gates have %e11
expressed in this way highei- gates :U-e then treated sirnil:u-ly.
The final s.0.p fonn should be in t e r m of basic events only.

If the fault tree contains repeated events then the resulting 5. BINARY IIECISION DIAGRAM MElHOII
s.0.p will not be m i n i n ~ and
l tlie minimal cut sets c;ui not be
directly obtained. If this is the case Boo1e;iu Recluction R I ~ ! The
~ Binary Decision Diognun (BL>I>)method, developed by
must first be applied to the s.0.p form to obtain the minirnal Rauzy (Ref. (71, first convei-ts the fault tree to ;I binary decision
cut sets. The task of obtaining the minimal cut sets of ;i ftiult diagram which encodes ;in If-Then-Else (ire) structure. The
tree can become computationally intensive. if the logic attractive thing :&out the BIN> method is that the ite structure
equations produce many cut sets, due to the number of derives from Shannons' formula (Ref. lo), such that if f(x)is
compzu-isons that 2u.e needed to make the expression minirnal. the Boolean Function for the top event of a lault tree theu the
Also the expansion procedure c;ui rnake extensive dernands on Shannon formula c m he written its:
memory space.

To overcome these problems various techniques h w e been XI. ,f'l+ XI. f 2 (4)
employed to reduce the uurnber of compnrisons (Ref. 7). Some
methods only produce the most important minirnal cut sets. and the corresponding ite structure is ite(X1, f l , f2), for ;I
(hie of these techniques is refeired to ;is culling, which means detailed account of this procedure refer to Ref. 11 and Ref. 12.
that cut sets of a certain order, s:iy 4 and above, ;re ignored or From this diagram both the qualitative x i c l quantitative
deleted from the expression, Rasmuson and Marshall (Ref. X) analysis can be achievecl.
employ this technique in their paper. The .iustificntion for
doing this is that cut sets of ;I high order tend t o have ;Ilow The size of the resultkg BDD is cleterminecl by the ordering
probability of occurrence ancl therefore do not 1nake ;I that has to be given to the basic events in the fault tree before
significmt contribution to the Top Event prolxibility. However the BDD is constructecl. This ordering has further implications
the dis;idvantage of this is that when coininon cause failures for the analysis. If the BDD is not in ;I minimal fonn, then the
:re involved this method results in consider;ible inaccuracies. BDD must first undergo ;I rniniinising algorithm before the
Probabilistic culling GLII also be applied, in this c x e ;i cut set minimal cut sets c:ui he obtained, this minimising technique is
whose probability of occuirence is below some thresholcl limit discussed in section 6. The quantitative analysis must be
will ag:un be ignored. performed 011 the unminirnised diagram. The reason being that
the minimising procedure produces ;i new BDD which only
4.2 Qurr n tit( i tivc An (I [Link] encodes the minimal cut sets. However if the ordering of the
basic events produces ;I minimal BDI> then both the
The convenlional ;ipproach (see Henley and Kurn:unoto in yunntitutive and qualitative analysis is straight forw:ud. It is
Ref. 9) to obtain the exact probability o f the 'Top Evcnt is to therefore beneficial t o achieve x i ordering which is optimal in
use the formula: terms of the resulting size of the BI>I>.The ordering of basic
events to produce ;i minimal diagram is consiclered in (Ref. 11)
:ind discussed in section 7.

To illustrate the method of obtaining the miniinad cut sets and

probability of occuirence o f tlie top event using the BI)I)
method refer to the example fault tree 111 tiguic 1.

Where ci,
i=l,.....,.... uc we the minini~ilcut sets of the Top
Event, i.e. product term.

Clearly if the fiiult bee has many minirnal cut sets calculating
P(Top) will require extensive calcu1:itions to evaluate each
Q
term in the expression, for rnaiiy complex fault trees the
requirement is beyond the capability of the available
machines. To simplify the calculation the Rare Eveut
Approximation, P RI TO^), ~ c m he used which is:
9 G1

111'

(2)
i=l Figure 1. Example Fault Tree.
However a more ;iccut';ite approxi~nationis the Mininixl (:ut Assume xi ordering for the basic events which is derived by
Set T Jpper Bound, P M ~ - , ~ ~ ~which
(TO is: ~ ) , considering those events at higher levels in the tree structure
first:

216 1996 PROCEEDINGS Annual RELIABILITY and MAINTAINABILITY Symposium

 XI<X2<X3<X4 F1

To obtain the ire structures for each gate in the f:iult tree the
following procedures are usccl:

(1) Taking X<Y;

Let J=ite(X, F1, F2) and II=ice(Y, (;I, (i2) then:

J<op>Ii=ite(X, Fl<op>lI, 1:2<op>I I) (5)

(2) Taking X=Y:

i.e., J=ite(X, F1, F2) :md II=ite(X, (;I, (;2) then:

J<op>H=ite(X, Fl<op>(il, F2<op>(;2) (6)

where <op> corresponds t o the Boolexi operation ot the logic

gates in the fault tree. For :ui AND gate <op> will he the dot
or product symbol aiid tor ;in OR gate <op> will hc the
addition symbol.
Figure 2. BDD for ite(X1. ite(X2. itebX3. 1, ite(X4. 1. O)),
Also it is eviclent that; ite(X3. 1. 0)). tb

To oht:un the cut sets ~t the hult tree the pnths through the
BDD :re traced from the top or root vertex to ;I terminal 1
vertex. Only the basic e v e m that lie oil ;I 1 hriuich (indicating
the friilure of that basic event) on the way to ;I terminal 1
vertex are included in ;I path. Thereforr the p:iths through the
Therefore the BDT) calculations for the fault tree in iigure 1
BDD which correspond to the cut set\ o f tlie lault tree :re:
are tlie following:
(1) Xl.X2.X3
G2 = ite(X3, 1,O)+itc(X4, 1, 0 ) (2) Xl.X2.X4
= ite(X3, 1, ite(X4, 1,O))
(3) XI.X3
Ci1 = ite(X2, l,O)+ite(X3, 1, 0 )
= ite(X2, 1, ite(X3, I, 0 ) )
Cllearly the resulting BDI) for this orclering is not minimum ;is
Top = [Link] it procluces one reclundant cut set. The minimising procedure
= ite(X2, I, ite(X3, I , O)).ite(X3, 1, ite(X4, 1,O)).
for the BDI) which will produce the miiiimal C u t sets directly
ile(X1, 1, 0)
is discussed in section 6.
= ite(X2, itc(X3, 1, ite(X4, 1, O)), ite(X3, 1, 0).
ite(X3, 1, ite(X4, 1, O)).ite(Xl, 1, 0 )
= ite(X2, ite(X3, 1, itc(X4. 1,O)). ite(X3, 1, 0))
.ite(Xl, 1, 0)
To obtxin the prohahility o f occurrence oi the top evellt ot the
Top =ite(X1, ite(X2, ite(X3, 1, itc(X4. 1, O ) ) , ite(X3, 1, O ) ) , 0 )
e,,,,,,
fault tree ( 1 the prohahihty ot (lie m n of the clisloiiit
This top event ire swucture coi-respontls to the B1)I:) shown in paths through the BDL) iu-e culcul:wd. The clis,joint paths
figure 2. through the BDD are iouncl hy simply iriclucling in a path the
basic events thar lie oil ;I 0 branch m c l inclicathig ~hese;isxi,
_.

i.e., 'Not' Xi, meiuiing basic event i cloes 1101 occur. 1)isjoiiit
paths through the BIN) ire:

(1) Xl.X2.X3
(2) Xl.X2.X1.X4
(3) Xl.X2.X3

Before continuing with the calculation of Q,vy,s the basic

emits iu the f d t tree ireecl to he assigned prohahilities, which
for this example :re given in table 1.

1996 PROCEEDINGS Annual RELIABILITY and MAINTAINABILITY Symposium 217

 e(l;,q ) - is die probability of system f%ilurewith q i= 1.
e(oi,y) - is the probability of system failure with yi = 0 .

Evaluating each o f the two terms Q(l, ,y) m c l Q(0,,y) to1

each cornponent coulcl be :ichievecl by t n \ t \ub\titutiiig q, = 1
Talde 1. Basic Event Data. aiid then yi = 0, i.e., the probability that coinpoiieiii i ecluals 1
and 0 respectively, and re-running tlie system failure
Where;
probability calculations. This would require the equivalent of
qi - I Jnavailability of compoiienr i. 211evalu:ttions of the top event probability to cleduce all terms
hi - Conditional failure intensity of coinpoileiit i.
in eq (X),
required in the expression for I.v,~,,,~
wi - 1Jiicoiiditioiialfailure intensity of component i.

Since e,(?,,
c m be obtained from the probability of the sum of Coiisicler the vat-iable Xi which occurs at two nodes in tlie
BDD (Figure 3 ) then:
the [Link] paths through the BIII) then:

Q,,,,s = P(x1. x 2 . x3 + x1. x 2 . m . x 4

-
+Yxl-Yx2.(1-Yx?)-'1x4+
- Yxl-qx2.Yx3

Yxl-(l-Yx2)-'Ix3
+ X 1 . E . X3)

/
/
/A\
/ \

\
\
+
= 0.0 1(0.02)((). 03) 0 . 0I((). ( )2)( 1 - ( ). (13)
/

(0.04) + [Link](1- 0.02)(0.03) Node ;I @ @ Node b

1,' \o l/
Q,v,s = 3.0776E - 4 / \
/
/ \.
\
/ \
/ /
\ / \\

The algorithm used by R X ~ for

J calculating the probability is
given in Ref. 6.
Fipure 3. Considerinlr varialde Xi.

For some systems it is Uie unreliability which IS required tor

the top event i.e., the probability it will not work continuously
over ;I given time period. An upper bound lor this is the
Expected number of top event tccurrences W(0, t):

t where:
W ( 0 ,t ) = J W,s,,,dt (7)
0 ) is
1 1 7 - ~ ~ ( q- tlie probability of the path section from tlie 1-oot
node to node xi.
1
w , ~ is, ~the
, ~system unconditional f;iilure iiiteiisi(y: p,.i (q) - is the probability of the path section from
node xi to the terminal 1 nocle after the 1 branch
from node xi.
0
poV,(y) - is Uie probability of the path sectioii from nocle xi
to the termin:il I node after the 0 branch from
where Gi (y) is tile criticality tunction lor e:tc~icomponeri( node xi.
Z(q) - is the probability of paths from the root node to
The criticality function Gi((1) is defined a s the prob;ibility the terminal 1 nodes which clo not go through ;i
node for variable xi.
that the system is in it critical state with respect to componeiit i I1 - All Iiocles for variable xi on the BDD.
and that the failure of coinpoileiit i will then cause the system
to go from the working t o the f;ulecl state, i.e., the probability Therefore:
that the system fails only if component i liiils. Therefore:

Where;

218 1996 PROCEEDINGS Annual RELIABILITY and MAINTAINABILITY Symposium

A more etficient way to calculate w,,,is to rnake one pass ot
I
Node Variable 1 Iwanch 0 Iwanch
the BDD to calculate p ~ .((I),
,~ J X ) , , ~ ((I) :111il l ~0 ((1)
) , tor
~ L,al,el pointer pointer
each node. With this irifoirnntion each G, ( q ) C;III hc easily F1 XI F2 0
F2 x2 F3 F4
eva1u;ltecl from eq (12) md w,y, !onnecl F3 X3 1 13
F4 X3 1 0
1 0 FS x4 1 0
The algorithm Prohpost to c;dcul:ite p , ,(q) ;uid I X I , (y)
~ is
given in ligure 4. The calculation of IN.,,,((I) c;ui he achieved 1
Performing one pass of the BL)D to evaluate po,vi((1) and
by the algorithm Probprev given in figure 5. The cntic:tlity
0
function G, ((I) for caicli hasic event is cadcuIated as sIiown in p O X i ((1) for each node using Prohpost givcs:
figure 6.
Probpost(FS)
FS=ite(X4, 1,O)
RtProht;ihle(X4, 1, 0 )
Ql cp(X4)=0.04
Q2=0
Prohpos t(F4)
F4=ite(X3, 1,O)
RtProhtahle(X3. 1. 0)
Q 1+p(X3)=0.03
Q2=0
Probpost(F3)
F3=ite(X3, 1, FS)
RtProhtahle(X3. I , proh(€?i))t(X?. I. 0 04)
Ql tp(X3)=0.03
Q 2 t ( l-p(X3))(0.04)=0.03~~
Prohpos t (F2)
F2=ite(X2, F3, F4)
RcPi-obt~thlt.(X2,pr~b(F3),proh(F4))t(X2,[Link].0.03)
Qltp(x2)(o.o6xx)=1.376E-3
Q 2 t ( 1-1)(X2))(0.03)=O.O2‘~4
Prohpost(F1)
Fl=ite(Xl, F2, 0)
IitProhtable(X1, proh(F2), O)t(X 1, 0.030776, 0)
~lt~(X1)(0.0.?0776)~3.0776~-4
y2=0

In performing this one pass. the top event prohahility caii he

cadcu1;itecl by:

Figure 5. l’rol~~wev
Algorithm.
for the top event node

The value\ of Piohpo4t 1 h i a “ and Piohpost 0 h m c h tor

each node are eiiteicd into the nocle piohahility tahle,
PROBTABLE ( w e figuie 7)
Next calculating the piohahility of the BD1) path t o
tach 11ode 14 established using Piohpiev and cnteied into the
4111 column ot the PR( )BTABLE
Figure 6. Algorithm for C;ilculatine the Criticality
Prohprev:
Function, G,,-.
l’rohprev(Fl)= 1
Ex;unde
Fl=ire(Xl, F2, 0)
Applying thew :IlgoriUiins t o the ex;irnple BDD given in tigurc Prohprev(F2)=P(X1 ).l’robprev(k 1 1
2 illustrates the application ot this methocl. =[Link]( 1 )=[Link]
The ite table for the B1)D in figure 2 is: H2=0
F2=ite(X2, F.?,F4)
Prohprev(F3)=p(X2).Pi.0hprev(F2)

1996 PROCEEDINGS Annual RELIABILITY and MAINTAINABILITY Symposium 219

 =0.02(0.01)=2.0E-4
Prohp1~v(F4)=( 1-p(X2)).P1.0bprev(F2)
=( 1-0.02)(0.01)=O .XE-3
F3=ite(X3, 1, FS)
Hl=l
Pl-ohprev(FS)=(1-p(X3)).Prohprev(F3)
=(1-O.O3)([Link]-4)=1.04E-4
F4=ite(X3, 1,O)
Hl=l
H2=0 1Jsing eq (7) the expected nuinher of top event O C C U I T ~ I ~hiC ~ ~
FS=ite(X4, 1,O) tiuie, t, c;ui he obtauned.
Hl=l
H2=0 6. MINIMISING THE RDl)

In the example fault tree (figure 1) the resulting BDI) (figure

Node I Variable post ‘1’ post ‘0’ prohprev 2) was not rniniinu~nas it procluced ;I redundant cut set. To
Label I PR( )BTABLE obtain only minimal cut sets the BDD must first undergo ;I
minimising procedure. From the unrninirnised BDD the
F1 I X1 0.030776 0 1
minimising algorithm of Rauzy (Ref. 6) creates ;I new BI>I)
O.06XX 0.03 0.01
0.04 [Link]-4 that symbolises only the minimal cut sets of the fault tree. If
[Link]-3 F=ite(x, G, H) then let 6 he ;I minimal solution of G which is
1.0454 not ai minitnal solution of H, then clearly the intersection of S
and x will be ai minimnl solution of F. Lastly, the set 0 of all
Proht;ible(i, l)=Basic event ot node Fi the miniinad solutions of F, YO/,,^^^ ( F ), will also include the
Prohtable(i, 2)=Prohability of post ‘1’brauich miniinal solutions of H so:
Probtable(i, 3)=Prohahility of post ‘0’hr:uich
Probtaible(i,4)=Prob~ihilityot previous

Calculation of tlie criticality function is then straight forw:rd

using the algorithm providcd in figure 6.
R u z y (Ref. 6) h s defined a ‘without’operator which removes
Criticality Algorithm:
from G,l,in:ill the paths included in ;i path of H. Applying this
c;(x1)=ci(x2)=(:~(x3)=~;(x4~=0 algorithm to the BDI) in figure 2 where ex11 node is
considered in tum:
Fl=Probtable(Xl, 0.030776, 0, 1)
(~(x1)=0+1(0.030776-0) FI=ite(Xl, F2, 0)- Here there are 110 solution\ 011 the 0 brainch
=0.030776 so the paths o k F2 remain unchmgecl.
F2=Probtable(X2, 0.06XX. 0.03, 0.0 1)
(;(x2)=0+0.0 1(0.06xx-0.03) F2=ite(X2, F3, F4) - Here X3 is iIicluded in a path 011 both the
=[Link]-4 1 branch (F3) xicl the 0 branch (F4), therefore X3 is removecl
F3=Prt)bt&le(X3, 1, 0.04, [Link]-4) trom the 1 brauich by replacing the tenninal 1 veriex with a 0.
(;(X3)=0+2.0E-4(1-0.04) [Refer to figure X I
=1.02E-4
F4=Probtahle(X3, 1, [Link]-3) F3=ite(X3, 0, FS) - FS does not contain any paths that :we
G(X3)=1.02E-4+[Link]-3(1-0) included in the 1 hixich a s this is a terminal vertex.
=O .00 2E - 3
FS=Probt&le(X4, 1, 0,1.04E-4) F4=ite(X3, 1, 0)- The without operator does not :ipply a s both
C;(X4)=1.94E-4(1-0) 0 and 1 brmiches w e tcrinin;l.
=1.04E-4
FS=ite(X4, 1,O) - S a n e applies ;is F4.
Since we have calculated the criticality function for each The minimised BI)I> is drawn in figure X
component, w , c m ~ now
~ he ~ evaluaietl
~ using the trequency
data from table 1 using eq (X).

220 1996 PROCEEDINGS Annual RELIABILlTY and MAINTAINABILITY Symposium

 F1 8 . MOI~UULARISING

Further improvements in terms oi coinputational efficiency

ciui he made for the more complex fiiult trees by modukrisiiig
the fault tree hefore the milysis takes place. Kliocla et al. (Ref.
14) define it module of ;I fiiult tree ~ L YIiaving no inputs which
appear elsewhere in the tree and iio outputs to the rest of the
tree except from its output event. For example consider the
fmlt tree in figure 0. Modules wliicli have the properties
defined above ru-e gates (i2, ( 3 iuicl Top.

Q
Figure 8. h4iiiimiserl BDD.

Tracing the paths through the minimised BDD we obtain the

minimal cut sets:
x
G1

Q
(1) Xl.X2.X4
(2) x 1. x 3

1. VARIABLE ORDERING SCHEME

The ordering of basic events will cletermine the size of the Firrure 9. A Fuult Tree which can be iiiodularised.
resulting BDD. BI)IYs produced using ;t simplc "top-down"
ordering of the v:u-i:tblcs are trequently inelficieiit since they
produce it large iiurnber of non-minimal cut sets. An
alternative ordering scheme is preseiited here which focuses
on those basic events which are repeated in the 13tult tree
structure. It is the repeattccl events which c;tusc the problem of
non-minimal cut sets, and by considering these events first
simplifies the resulting BDI) structure :tiid therefore inakes it
0
more optimal.
The alternative ordering scheme again considers the basic
events in it top-down ordering (after the fault tree structure is
contracted into a11 alteniating sequence o f AND and ()R
gates). However a s each gale is considered the basic events
which are inputs to the gatc are t;Lken in order of Uiosc which
occur most frequently in the tault tree and placed in the
ordering list. When gate input everits ;ue encountered which
are already entered in the ordering list cluc to the occurrence at Figure 10. Modularised h u l t tree.
;I higher level in the tree then they iire iyiored ~ i the d
rernauiiing input events are ordered. The modularised fault tree is shown in figure 10. By then
using the BDD method to mtlyse this tree in terins of the
Applying the new ordering heie to the examplc tault tree moduks and then each module in tuiii the results cxi he
(figure 1) with repeated event X7, we get the orrlcring combined to provide iui efficient means ot ;uialysing the whole
X l < X k X 2 < X 4 . The resulliiig BDI) toi tiits allernalive fialt tree.
ordering is miniiiiuin s o the minimising technique is not
needed, this is aclvantageous in terms ot reduced comput;ition 0. CONCLUSION
t h e . Work cmied out to ditte inclicates tha( the new orrlering
atppears to produce more optimal BDTI's comp;u-ecl to other ('onventional top-down m c l bottom-up techniques c;ui lead to
orderings. Bryant (Ret. 13) recogiii\ed the prohlcin of many redundant cut sets ancl calculating exact top event
computing iui ordering that iniiiutiises Uie size ot the Br)D m d probability can become impossible. To uiiprove these aitalysis
for some trees it may not be possiblc lo pioduce ;i miiiim;il procedures the aim has been to represent the system fiiilure
BDD whatever the ordering. logic in a mode which lends itsell to the mathernatical
oiaiiipulnt ion.

1996 PROCEEDINGSAnnual RELIABILITY and MAINTAINABILITY Symposium 221

 BIOGRAPHIE,\'
Representing the Boolean failure logic eciu:itioii in the form of
a B1)I) provides an alternative techiiique which gives Roslyn M. Sinn:unon, BSc
significant savings in the computational efficiency and lends Department of Mntlieinatical Sciences
itself to m;iiiipulation. Also the BDI) produces exact [Link] of Technology
quantified results ancl top event parmeters such as failure Loughborough, Leicestershire LE1 1 3TI.J, [Link].
probability and tlie system unconditional frulure intensity and Zn rein et (e-nzriiO : [Link];unon 1CC lu t .;[Link]
the expected number of occuixnces c m be obtained with e.:ise.
Ros Simiarnoa is a second year PhI) student working with I>r
To simplify the analysis even further the liiult tree may be .Tohi1 Andrews on Reliability Theory, iminly Fault Tree
modularised prior to tlie aiialysis. An alleimitive ordering of Analysis. She also tutors in Reliability, Statistics atid
the basic event variables has also shown itself to signillcantly Mathematics for Engineers at Loughhorough 1Jiiiversity. Her
improve efficiency . BSc Joint Honours is in Matliernatics, Physical Education xicl
Sports Science auld was gained ;it Loughborough. She is :I
The trade off for the advantages described is the effort taken to member of the Sakty ancl Reli:ihility Society.
convert the logic from tlie fault tree structure to the BDL)
form. However early work indicates that for large, complex John D. Andrews, PhD, BSc
trees this can produce ;I substantial recluction iii coniputational Depautinent of MathernaticA Sciences
effort. Loughborough [Jniversity of Technology
Loughborough, Leicestershire, LE1 1, 3TI.J, [Link]
10. REFERENCES I n tmwt (e-nuril):[Link](4) lu t .;IC. uk

1 J.1). Andrews and T.R. Moss, "Rcliahility r i n d Risk Dr Andrews currently lectures in Risk ancl Snfety Assessrneiit
A.r,ses,strzcn.r, Longinan Scientific aiid Techniail, 1003.
" Techniques in the Ikpartmeiit of Mathernat ical Sciences at
2 S.N. Sernanderes, "'ELRAFT,' A computer program for Loughborough IJniversity of Technology. Prior to this
the efficient logic reduction analysis of fault trees," IEEE appointment he was ;I Senior Lecturer in the 1)epartment of
TrLins. Nticlcrir. Scirnce, vol NS-18. 1071 Feh, pp 481-487. Mechanical iuid Production Engineering at Biriningliam
3 J.B. Fussell, W.E. Vesely, " A new rnetlioclology for Polytechnic ancl has also h;d two periods of employment ;is a
obtaining cut sets for fault trees," Ti-rin.~.Aiii. Nircl. Soc.. vol Senior Scientist Engineer in the Resexch aiid 1)evelopment
15, 1072 J u ~pp, 262-263. Division at British G a s ,
4 R.G. Bennetis, "(.hi the an;ilysis 0 1 h u l t trees." IEEE
Turns. Relitihility, vol R-24. No. 3 . 1075 Aug. pp 175-185. His industrial work has involved research into methods of
5 N.N. Bengiarnin. B.A. Bowen, K.F. Schenk, "An awessing the safety and risk of potentially haz~u-dousindustrial
efficient algoritlun for reducing the complexity o f computation activities. This research is tiow continuing at Loughborough
iu fiiult tree analysis," IEEE Ti-([Link]~l(!trr-ScYcircc, vol NS- [Link].I>rAndrews is currently ;I member of-committees
23, NO. 5 , 1076 (.)ct, pp 1442-1446. of tlie Institution of Mechmic:l Engineers :mcl tlie Safety and
6 A. Rauzy, "New :ilgorithms for t';iult trce an;ilysis," Reliability Society which focus on Risk, Sxfety and Reliability
Relitrhility Engineci-ing i i n d Systcin S(!/i,t~, vol 40. 1003, issues.
pp203-211.
7 N. Lirninios, R. Ziaini, "Ail algori thin lor reducing the
minimal cut sets in f a u l t tree :iiialysis," IEEE Ti-tins.
Reliuhilrty, vol R-35, No. 5, 1986 Dee, pp 550-561.
8 D.M. Ramuson, [Link]. "FATRAM-A core
efficient cut-set algorithm," IEEE TI-([Link])ility, vol 11-27,
NO. 4, 1078 Oct, pp 250-253.
9 E.J. Henley and [Link];unoto, "Rclirrhility En,qincer-ing
ancl Risk [Link]~nt,"Englewood Cliffs, 1081.
10 W. (3. S chnee wei ss, "Roolct in Fitn ct iolr s with
Enginecrin.g Applictrtionx t i i d Coiiipritcr Pi~o,yi-(ii/is,"
Springer-
Verlag, 1989.
11 R.M. Sinn:unon ;md J.D. Andrcws, "Ncw Approaches
to Evaluating Fault Trees," Pi-ococriings c!f E ~ i - i ~ I ' 9 . 5
Coqfki-cmx, JUIIG, 1005 pp24 1-254.
12 A. Rauzy et al., "Computation of prime iinplicants of
;I fault tree within Aralia," P1-0c~c~c~dit7,qs c!f' Esi-elYS
Confei-cwo, June, 1005 ppl00-202.
13 R. E. Bryant, "(Iiraph-B;ised algorithins for Boolean
function manipulation," IEEE [Link]~[Link]~pi~rci-.~, vol (3-35,
No. 8, 10x6 Aug, pp677-601.
14 T. Khocla, E. J. Henley x i c l K . Inoue, "Fiiiding
Modules in Fault Trees," IEEE Ti-trnx R<,/itrlJility,vol 38, No.
2, 1089 Jun, pp165-176.

222 1996 PROCEEDINGS Annual RELIABILITY and MAINTAINABILITY Symposium