Structure Function
Reliability of Systems of Independent Components
Bounds on the Reliability Function
System Life as a Function of Component Lives
Expected System Lifetime
1. 1
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Structure Function
Reliability of Systems of Independent Components
Bounds on the Reliability Function
System Life as a Function of Component Lives
Expected System Lifetime
Reliability Theory9
2. 2
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Structure Functions
• Consider a system consisting n components, each component is
either functioning or has failed.
=
otherwise0
functionscomponentif1 i
xi• Indicator variable:
( )nxxx ,...,,x 21=• State vector:
• Suppose that: whether the system is functioning or nor depends on x.
There exists a function Φ(x) such that:
( )
=
x
x
x
isvectorstatewhenfailssystemif0
isvectorstatewhenfunctionssystemif1
φ
( ) SystemtheofFunctionStructureThe:xφ
3. 3
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Example: Serial Structure
• System functions only if all components function. Therefore:
( ) ( ) ∏=
==
n
i
in xxxx
1
21 ,...,,minxφ
4. 4
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Example: Parallel Structure
• System functions only if at least one component functions. Hence:
( ) ( ) ( )∏=
−−==
n
i
in xxxx
1
21 11,...,,maxxφ
5. 5
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Example: k-out-of-n Structure
• System functions if at least k of the n components function.
( )
<
≥
=
∑
∑
=
=
kxif
kxif
n
i
i
n
i
i
1
1
0
1
xφ
The 2-out-of-3 system
6. 6
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Example: four-component Structure
( ) ( )[ ] ( )( )[ ]
[ ]434321
43114321 111,max
xxxxxx
xxxxxxxx
−+=
−−−==xφ
7. 7
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Monotone system
• Assumption: replacing a failed component by a functioning one will
not lead to a deterioration of the system, i.e., Φ(x) is an increasing
function of x
( ) ( ) ( )yx φφ ≤⇒=≤ niyx ii ,...,2,1
• We consider only systems with the above property as monotone
systems
8. 8
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Minimal Path Set
• x is called a path vector if Φ(x) = 1. Furthermore, if Φ(y)=0 for all y
< x then x is said to be a minimal path vector.
<
=≤
⇔<
isomeforxy
nixy
Note
ii
ii ,...,2,1
: xy
• If x is a minimal path vector then the set A{i, xi = 1} is called a
minimal path set.
• ⇒ A minimal path set is a set of components which ensures the
functioning of the system.
9. 9
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Example
( ) ( )[ ] ( )[ ]
[ ][ ]5435432121
54321 ,max,max
xxxxxxxxxx
xxxxx
−+−+=
=xφ
There are four minimal path sets: {1,3,4},{1,5},{2,3,4},{2,5}
10. 10
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Minimal Path Set (cont.)
• Indicator function αj(x) of the minimal path set Aj{j = 1, 2…,s} is defined as:
( ) ( ) ∏∈
∈
==
j
j
Ai
ii
Ai
j xxThen minxα
• The system functions if all components of at least one minimal path set
function. Hence,
( )
=
otherwise0
functionofcomponentsallif1 j
j
A
xα
( )
( )
( )
( ) ( ) ∏∈
==⇒
=
=
=
jAi
i
j
j
j
j
j
xα
jα
jα
maxmax
allfor0if0
somefor1if1
xx
x
x
x
φ
φ
Note that αj(x) is a serial structure function of the components of the jth
minimal
path set
⇒ Any system can be considered as a parallel arrangement of serial systems
11. 11
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Example
( ) ( )
( )( )( )( )5251432431
5251432431
11111
,,,max
xxxxxxxxxx
xxxxxxxxxx
−−−−−=
=xφ
The minimal path sets are: A1={1,3,4},A2={1,5}, A3={2,3,4}, A4={2,5}
This is exactly the same function as before, isn’t it?
12. 12
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Example: The bridge system
( ) ( )
( )( )( )( )5241531432
5243253141
11111
,,,max
xxxxxxxxxx
xxxxxxxxxx
−−−−−=
=xφ
• minimal path sets : A1={1,4}, A2={1,3,5}, A3={2,5}, A4={2,3,4}
• The structure function can be expressed as:
13. 13
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Minimal Cut Set
• x is called a cut vector if Φ(x) = 0. Furthermore, if Φ(y) =
1 for all y > x then x is said to be a minimal cut vector.
• If x is a minimal cut vector then the set C{i: xi = 0} is
called a minimal cut set.
• ⇒ A minimal cut set is a minimal set of components whose
failure ensure the failure of the system.
14. 14
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Minimal Cut Set (cont.)
• Indicator function βj(x) of the minimal cut set Cj{j = 1, 2…,k} is defined as:
( ) ( ) ( )∏∈
∈
−−==
j
j
Ci
ii
Ci
j xxThen 11maxxβ
• The system is not functioning only if all components of at least one minimal cut
set are not functioning. Hence,
( )
=
otherwise0
functionsofcomponentsoneleastatif1 j
j
C
xβ
( )
( )
( )
( ) ( ) ( )∏∏ =
∈
=
==⇒
=
=
=
k
j
i
Ci
k
j
j
j
j
x
j
j
j
11
max
allfor1if1
somefor0if0
xx
x
x
x
βφ
β
β
φ
Note that βj(x) is a parallel structure function of the components of the jth minimal cut
set
⇒ Any system can be considered as a serial arrangement of parallel systems
15. 15
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Example: The bridge system
• minimal cut sets :
C1={1,2}, C2={1,3,5}, C3={4,5}, C4={2,3,4}
• Equivalent structure:
( ) ( ) ( ) ( ) ( )
( )( )[ ] ( )( )( )[ ]
( )( )( )[ ] ( )( )[ ]54432
53121
5443253121
1111111
1111111
,max,,max,,max,max
xxxxx
xxxxx
xxxxxxxxxx
−−−−−−−×
−−−−−−−=
=xφ
16. 16
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Reliability of Systems of Independent Components
• Suppose that the state of the ith component, Xi, is a random variable
such that:
{ } { }011 =−=== iii XPXPp
pi is the reliability of component i.
• Reliability of the system:
( ){ } ( )nXXXPr ,...,,where1 21=== XXφ
Because the random variables Xi; i = 1,2,…,n are independent, r can be
expressed as a function of component reliabilities:
{ } ( )npppprr ,...,,where 21== p
• Note: { } ( ){ } ( )[ ]
( ) variable.random)(Bernoulii10aisBecause
1
−
===
X
XXp
φ
EPr φφ
17. 17
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Example: Serial System
( ) ( ){ }
{ }
∏=
=
===
==
n
i
i
i
p
niXP
Pr
1
,...,2,1allfor1
1Xp φ
18. 18
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Example: Parallel System
( ) ( ){ } ( ){ }
{ }
{ }
( )∏=
−−=
==−=
===
====
n
i
i
i
i
i
i
p
niXP
niXP
XPPr
1
11
,...,2,1allfor01
,...,2,1somefor1
1max1Xp φ
19. 19
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Example: 2-out-of-3 System
( ) ( ){ }
{ } { } { } { }
( ) ( ) ( )
321323121
321321321321
2
111
)1,1,0()1,0,1()0,1,1()1,1,1(
1
ppppppppp
pppppppppppp
PPPP
Pr
−++=
−+−+−+=
=+=+=+==
==
XXXX
Xp φ
20. 20
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Example: k-out-of-n System
• Suppose that pi = p for all i = 1,2,…,n
( ) ( ){ }
( ) in
n
ki
i
n
i
i
pp
i
n
kXP
Pr
−
=
=
−
=
≥=
==
∑
∑
1
1
1
Xp φ
21. 21
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Reliability of Systems of Independent Components
Proposition: If r{p}is the reliability of a system of independent
components, then r{p}is an increasing function of p.
{ } ( )[ ]
( )[ ] ( ) ( )[ ]
( )[ ] ( ) ( )[ ]XX
XX
Xp
,01,1
011
Proof
iiii
iiii
EpEp
XEpXEp
Er
φφ
φφ
φ
−+=
=−+==
=
( ) ( )
( ) ( )
{ } ( ) ( )[ ] ( )[ ]
( ) ( )[ ]
{ } ipr
E
EEpr
XXXX
XXXXwhere
i
ii
iiii
niii
niii
allforinincreasingis
0,0,1:functionincreasinganisSince
,0,0,1
,...,,0,,...,,0
,...,,1,,...,,1
111
111
p
XX
XXXp
X
X
⇒
≥−
+−=⇒
=
=
+−
+−
φφφ
φφφ
φ
φ
22. 22
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Reliability of Systems of Independent Components
• Consider the following problem: A system of n different
components is to be built from a stockpile containing exactly two of
each type of component. The question is whether:
1. To build two separate systems with the probability of functioning is:
{ }
{ }
( )( ) ( )( )[ ]p'p rr
P
P
−−−=
−=
111
functionsystemsofneither1
functionsystemstwotheofoneleastAt
23. 23
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Reliability of Systems of Independent Components
2. Or to build a single system whose ith component functions if at
least one of the number i components function. In this case the
probability that the system will function is:
{ }
{ }
( )( ) ( )( )[ ]p'p rr
P
P
−−−=
−=
111
functionsystemsofneither1
functionsystemstwotheofoneleastAt
24. 24
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Reliability of Systems of Independent Components
Theorem: For any reliability function r and vectors p, p’
( )( )[ ] ( )[ ] ( )[ ]'111'111 pppp rrr −−−≥−−−
“Replication at the component level is more effective
than replication on the system level”
25. 25
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Reliability of Systems of Independent Components
Proof:
Let X1, X2,…, Xn ; X1’, X2’,…, Xn’ be mutually independent 0 - 1 random variables with
{ } { }
( ){ } ( )( )
( )( )[ ] ( )( )[ ]
( )( ) ( ) ( )
( )( )[ ] ( ) ( )( )[ ]
( ) ( )( ){ }
( ) ( ){ }
( )[ ] ( )[ ]'111
0,01
1,max
,maxHence,
max:oftymonotoniciBy
max
1111,max
1'1
''
'
pp
X'X
X'X
X'Xp'1p11
X'XX'X,
X'X,p'1p11
rr
P
P
Er
and
Er
ppXXP
XPpXPp
iiii
iiii
−−−=
==−=
==
≥−−−
≥
=−−−⇒
−−−==
====
φφ
φφ
φφ
φφφφ
φ
( ) ( ) ( )
( ) ( ) ( ) ( ){ }nn
nnnn
yxyxyx
yxyxyxyyyxxx
Notation
,max,...,,max,,max,max
,...,,,...,,;,...,,
:
2211
22112121
=
=⇒==
yx
xyyx
26. 26
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Example
• Consider the case of two types of component with
2
121 == pp
• Replication at the system level:
• Replication at the component level:
( )[ ] ( )[ ]
16
7
2
1
2
1
1
2
1
2
1
11'111 =
−
−−=−−−= pp rrrs
( )( )[ ]
16
7
16
9
4
3
2
1
11,
2
1
11
222
>=
=
−−
−−=−−−= rrrC p'1p11
27. 27
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Bounds on the Reliability Function
Method of Inclusion and Exclusion
•Formula for the probability of the union of the events E1, E2, . . . , En:
( ) ( ) ( ) ( ) ( )n
n
kji
kji
ji
ji
n
i
i
n
i
i EEEPEEEPEEPEPEP ...1... 21
1
11
+
<<<==
−+−+−=
∑∑∑
• Hence: ( )
( ) ( )
( ) ( ) ( )
...
...
11
11
11
≤
≥
+−≤
−≥
≤
∑∑∑
∑∑
∑
<<<==
<==
==
kji
kji
ji
ji
n
i
i
n
i
i
ji
ji
n
i
i
n
i
i
n
i
i
n
i
i
EEEPEEPEPEP
EEPEPEP
EPEP
28. 28
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Bounds on the Reliability Function
Method of Inclusion and Exclusion (cont)
•Let A1, A2, . . . ,As denote the minimal path sets of a given structure φ, and define the
events E1, E2, . . . , Es by
Ei = {all components in Ai function}
•Since the system functions if and only if at least one of the events Ei occur, we have
( ) ( ) ( ) ∏∏∏ ∪∪∈∪∈∈
===
kjijii AAAl
lkji
AAl
lji
Al
li pEEEPpEEPpEP ;;
• where:
( ) ( )
( ) ( )
...
1
11
≤
−≥
≤
=
∑∑
∑
<=
=
ji
ji
s
i
i
s
i
i
s
i
EEPEP
EPEPr p
29. 29
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Example: The bridge system
ippi allfor=
• minimal path sets : A1={1,4}, A2={1,3,5}, A3={2,5}, A4={2,3,4}
• Because exactly five of the six unions of Ai and Aj contain four components (the
exception being A2 ∪ A4, which contains all five components), we have
( ) ( ) ( ) ( ) 3
42
2
31 ; pEPEPpEPEP ====
( ) ( ) ( ) ( ) ( )
( ) 5
42
4
4332413121 ;
pEEP
pEEPEEPEEPEEPEEP
=
=====
• Hence, the first two inclusion–exclusion bounds yield
( ) ( ) ( )325432
252 pprpppp +≤≤−−+ p
30. 30
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Bounds on the Reliability Function
Second Method for Obtaining Bounds on r(p):
•Let A1, A2, . . . ,As denote the minimal path sets of a given structure φ, and define the
events D1, D2, . . . , Ds by
Di = {at least one component in Ai has failed}
• since the system will have failed if and only if at least one component in each of the
minimal path sets has failed we have:
( ) ( ) ( ) ( ) ( )12112121 .........1 −==− sss DDDDPDDPDPDDDPr p
• We have: ( ) ( )iii DPDDDP ≥−11...
• Hence, ( ) ( )
( ) ( ) ∏ ∏∏
∏
−−=−≤⇔
≥−
∈i Aj
j
i
i
i
i
i
pDPr
DPr
111
1
p
p
31. 31
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Bounds on the Reliability Function
Second Method for Obtaining Bounds on r(p) (cont):
•let C1, . . . ,Cr denote the minimal cut sets and define the events U1, . . . ,Ur by
Ui = {at least one component in Ci is functioning}
• since the system will function if and only if all of the events Ui occur, we have:
( ) ( )
( ) ( ) ( )
( ) ( )∏ ∏∏
−−=≥
=
=
∈
−
i Cj
j
i
i
rr
r
i
pUP
UUUUPUUPUP
UUUPr
11
......
...
121121
21p
• Finally,
( ) ( ) ∏ ∏∏ ∏
−−≤≤
−−
∈∈ i Aj
j
i Cj
j
ii
prp 1111 p
32. 32
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
System Life as a Function of Component Lives
• Letting F(t) denote the distribution of system lifetime
( ) ( ) { }
{ }
( ) ( )( )tPtPr
tP
tPtFtF
n,...,
at timegfunctioninissystem
lifesystem1
1=
=
>=−=
Where:
( ) { }
{ }
( )tF
tiP
tiPtP
i
i
=
>=
=
oflifetime
at timegfunctioniniscomponent
Hence, ( ) ( ) ( )( )tFtFrtF n,...,1=
33. 33
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Example
( ) ( ) ( )∏∏ ==
=⇒=
n
i
i
n
i
i tFtFpr
11
p
• Serial System:
( ) ( ) ( ) ( )∏∏ ==
−=⇒−−=
n
i
i
n
i
i tFtFpr
11
111p
• Parallel System:
34. 34
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
The Failure Rate Function
• the failure rate function λ(t) of continuous distribution G represents
the probability intensity that a t-year-old item will fail.
( ) ( )
( )
( ) ( )
dt
tdG
tgwhere
tG
tg
t ==λ
• G is an increasing failure rate (IFR) distribution if λ(t) is an
increasing function of t.
• G is a decreasing failure rate (DFR) distribution if λ(t) is a
decreasing function of t.
35. 35
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Example: The Weibull Distribution
( ) ( )
0,1 ≥−= −
tetG t α
λ
• A random variable is said to have the Weibull distribution if its
distribution is given, for some λ > 0, α > 0, by
• The failure rate function for a Weibull distribution:
( )
( )
( )
( )
( ) 1
1
−
−
−−
==
α
λ
αλ
λαλ
λλα
λ α
α
t
e
te
t t
t
• the Weibull distribution is IFR when α ≥ 1, and DFR when 0 < α ≤1.
• when α = 1, G(t) = 1 − e−λt
is the exponential distribution.
36. 36
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
The Hazard Function
( ) ( )
( )
( ) ( )
( )
( )( )
( ) ( )
( ) ( ) .ondistributitheofthe:
,
ln
1
1
0
00
Fctionhazard fundsstwhere
etFHence
tFds
sF
sf
dss
tF
tf
t
t
t
tt
∫
∫∫
=Λ
=
−=
−
=⇒
−
=
Λ−
λ
λ
λ
• A distribution F is said to have increasing failure on the average
(IFRA) if:
( )
( )
0forinincreases0
≥=
Λ ∫
tt
t
dss
t
t
t
λ
37. 37
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Expected System Lifetime
• We have:
{ } ( )( ) ( ) ( ) ( )( )tFtFtwheretrtP n,...,lifesystem 1==> FF
• And:
{ } ( ) ( ) ( ) [ ]XEdyyyfdxdyyfdydxyfdxxXP
y
x
====> ∫∫∫∫∫∫
∞∞∞ ∞∞
00 000
• Thus,
[ ] ( )( )∫
∞
=
0
lifesystem dttrE F
38. 38
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Examples
( )
( )( )
>
≤≤
−
=
=
>
≤≤
=
10,0
100,
10
10
Therefore,
3,2,1
10,1
100,10
3
t
t
t
tr
i
t
tt
tFi
F
• A Series System of Uniformly Distributed Components:
Consider a series system of three independent components each of which functions for
an amount of time (in hours) uniformly distributed over (0, 10). Hence, r(p) = p1p2p3
[ ]
2
5
10
10
10
lifesystem
1
0
3
10
0
3
==
−
=
∫
∫
dyy
dt
t
E
39. 39
Assoc. Prof. Ho Thanh Phong
Probability Models
International University – Dept. of ISE
Examples
( )
( )
>
≤≤
=
−++=
1,1
10,
2 321323121
t
tt
tF
pppppppppr
i
p
• A Two-out-of-Three System:
Consider a two-out-of-three system of independent components, in which each
component’s lifetime is (in months) uniformly distributed over (0, 1).
[ ] ( ) ( )[ ]
[ ]
2
1
2
1
1
23
1213lifesystem
1
0
32
1
0
32
=−=
−=
−−−=
∫
∫
dyyy
dtttE