Reliability Engg Lect 3
Reliability Engg Lect 3
(p+q)
2
= p
2
+ 2pq + q
2
(p+q)
3
= p
3
+ 3p
2
q + 3pq
2
+ q
3
=
n
C
r
r)! - (n r!
n!
Required conditions:
- number of trials (n) must be known
- all trials are independent
- each trial results in either a success (p) or failure (q) i.e. p + q = 1
- values for p and q are constant for each trial
Probability of exactly r failures (and n-r successes),
P
r
=
n
C
r
p
(n-r)
q
r
Coefficients of Binomial Distribution
No. of combinations of r successes (or failures) in n trials =
=
n
C
r
r)! - (n r!
n!
If n is small, Pascals Triangle:
1
1 1 n=1 (p+q)
1
= 1p + 1q
1 2 1 2 (p+q)
2
= 1p
2
+ 2pq + 1q
2
1 3 3 1 3 (p+q)
3
= 1p
3
+ 3p
2
q + 3p
2
q + 1q
2
1 4 6 4 1 4
1 5 10 10 5 1 5
Example
Consider the case in which the probability of success in a single
trial is and four trials are to be made. Evaluate the individual
and cumulative probabilities of success in this case.
n = 4, p = , q = 1 =
(p+q)
4
= p
4
+ 4p
3
q + 6p
2
q
2
+ 4pq
3
+ q
4
Number of
Cumulative
successes failures
Individual probability probability
0 4 q
4
= (3/4)
4
= 81/256 81/256
1 3 4pq
3
= 4(1/4)(3/4)
3
= 108/256 189/256
2 2 6p
2
q
2
= 6(1/4)
2
(3/4)
2
= 54/256 243/256
3 1 4p
3
q = 4(1/4)
3
(3/4) = 12/256 255/256
4 0 p
4
= (1/4)
4
= 1/256 256/256
E = 1
Probability Distributions
Probability Distribution Function
0.0
0.2
0.4
0.6
0.8
1.0
0 1 2 3 4
# of Heads, x
C
u
m
.
P
r
o
b
a
b
i
l
i
t
y
,
F
(
x
)
Probability Density Function
0.0
0.1
0.2
0.3
0.4
0 1 2 3 4
# of Heads, x
P
r
o
b
a
b
i
l
i
t
y
,
f
(
x
)
Moments of Binomial Distribution:
Expected # of successes, E(x) = n.p
where,
n = # of trials or samples
p = probability of success
q = probability of failure
Std. Deviation, (x) = n.p.q
Example
A product is claimed to be 90% free of defects. What is the
expected value and standard deviation of the number of defects
in a sample of 4?
Here, n = 4
p(defect) = 0.1
E(defects) = n.p
= 4 x 0.1 = 0.4
(defects) = npq
= 4 x 0.1 x 0.9 = 0.6
Binomial Distribution Application
Binomial Distribution can be used to determine the probability of all
possible outcomes in a system. Only a restricted number of outcomes
(and their probabilities) are usually of interest in engineering
applications, e.g. outcomes that lead to system failure.
Example:
It is known that, in a certain manufacturing process, 1% of the products are
defective. If the a customer purchases 50 of these products selected at random,
what is the probability that he receives 2 or less defective products?
n = 50
p(defect) = 0.01 p(no defect) = 1 0.01 = 0.99
P(2 or less defects) = 0.0756 + 0.3056 + 0.6050 = 0.9862
P(2 defective) =
50
C
2
(0.01)
2
(0.99)
48
= 0.0756
P(1 defective) =
50
C
1
(0.01)
1
(0.99)
49
= 0.3056
P(0 defective) =
50
C
0
(0.01)
0
(0.99)
50
= 0.6050
Mutually exclusive events
Example:
The manufacturing company has a policy of replacing, free-of-charge,
all defective products that are purchased.
If the product manufacturing cost is $10 per unit and each product is
sold for $15, how much profit is made?
p(defect) = 0.01
For 1000 products, n = 1000
Expected # of defects, E(defects) = n.p = 10
Therefore, 1010 products must be manufactured to sell 1000 products.
Manufacturing cost = $10 x 1010 = $10,100
Income = $15 x 1000 = $15,000
Profit = $15,000 - $10,100 = $4900
Profit per unit = $4900/1000 = $4.90
Example:
If the company decides to increase the manufacturing cost to $10.05
per unit in order to decrease the probability of defects to 0.1%,
p(defect) = 0.001
For 1000 products, n = 1000
Expected # of defects, E(defects) = n.p = 1
Therefore, 1001 products must be manufactured to sell 1000 products.
Manufacturing cost = $10.05 x 1001 = $10,060.05
Income = $15 x 1000 = $15,000
Profit = $15,000 - $10,060.05 = $4939.95
Profit per unit = $4939.95/1000 = $4.94
Effect of Redundancy
Consider a system consisting of 4 identical components,
each having a failure probability of 0.1.
q = 0.1 (p = 0.9) n = 4 (p+q)
4
= p
4
+ 4p
3
q + 6p
2
q
2
+ 4pq
3
+ q
4
System state Individual probability
all components working
p
4
= (0.9)
4
= 0.6561
3 working, 1 failed
4p
3
q = 4(0.9)
3
(0.1) = 0.2916
2 working, 2 failed
6p
2
q
2
= 6(0.9)
2
(0.1)
2
= 0.0486
1 working, 3 failed
4pq
3
= 4(0.9)( 0.1)
3
= 0.0036
all components failed
q
4
= (0.1)
4
= 0.0001
E = 1
Consider 4 criteria
System reliability, R
all components required for success
(no redundancy)
0.6561
3 components required for success
(partial redundancy)
0.6561+0.2916
= 0.9477
2 components required for success
(partial redundancy)
0.6561+0.2916+0.0486
= 0.9963
1 component required for success
(full redundancy)
0.6561+0.2916+0.0486+0.0036
= 0.9999
Effect of Redundancy
A pump has a failure probability of 0.1. Compare the reliability
of the cooling system with and without redundancy.
q = 0.1 (p = 0.9)
cooling
System
pump 1
cooling
System
pump 1
pump 2
pump 1
cooling
System
pump 1
pump 2
Reliability, R = p = 0.9
Unreliability, Q = q = 0.1
R = p
1
+ p
2
+ p
1
x p
2
= 0.9 + 0.9 - 0.9 x 0.9 = 0.99
Q = q
1
x q
2
= q
2
= 0.1
2
= 0.01
Q = q
1
x q
2
x q
3
= q
3
= 0.1
3
= 0.001
R = 1 - Q = 0.999
System with Derated States
Two-state systems: (1) success or up state
(2) failure or down state
Many systems can also reside in derated (or partial output) states,
e.g. process plant, power generation plant
One of the most useful application of Binomial Distribution is to
evaluate engineering systems with de-rated states.
- build COPT using Binomial Distribution
- use COPT to evaluate system reliability
COPT: capacity outage probability table
System with Derated States
Consider a generation plant with two 10 MW units, each having a
forced outage rate (probability of failure) of 10%.
p = 0.9, q = 0.1, n = 2
(p + q)
2
= p
2
+ 2pq + q
2
Binomial Distribution:
Units Out Cap Out (MW) Cap In (MW) Probability Cum. Prob
0 0 20 0.81 1
1 10 10 0.18 0.19
2 20 0 0.01 0.01
Capacity Outage Probability Table:
If the generation plant operates to supply a 15 MW load,
what is the probability of load loss (system failure)?
Probability of load loss = Loss of Load Probability (LOLP) = 0.19
Expected # of days of load loss = 0.19 x 365 = 69.35 days/yr
Loss of Load Expectation (LOLE)
System with Derated States
Consider a generation plant with two 10 MW units, each having a
forced outage rate (probability of failure) of 10%.
If the generation plant operates to supply a 15 MW load,
what is the Expected Load Loss (ELL)?
p = 0.9, q = 0.1, n = 2
Capacity Outage Probability Table:
Expected Load Loss (ELL) = 1.05 MW
Units Out Cap Out (MW) Cap In (MW) Load Loss (MW) Probability Col.4 x Col.5
0 0 20 0 0.81 0
1 10 10 5 0.18 0.90
2 20 0 15 0.01 0.15
1.05
Example
A generating plant is to be designed to satisfy a constant 10 MW load.
Four alternatives are being considered:
a) 1 x 10 MW unit
b) 2 x 10 MW units
c) 3 x 5 MW units
d) 3 x 3.33 MW units
The probability of unit failure (unavailability or FOR) is assumed to be
0.02.
FOR or U = 0.02
Availability of each unit, A = 0.98
Capacity Outage Probability Tables
units
capacity
(MW) Binom. Individual cum.
out out in Distr. prob prob
(a) 1 x 10 MW
0 0 10 A 0.98 1.00
1 10 0 U 0.02 0.01
(b) 2 x 10 MW
0 0 20 A
2
0.9604 1.0000
1 10 10 2AU 0.0392 0.0396
2 20 0 U
2
0.0004 0.0004
(c) 3 x 5 MW
0 0 15 A
3
0.941192 1.000000
1 5 10 3A
2
U 0.057624 0.058808
2 10 5 3AU
2
0.001176 0.001184
3 15 0 U
3
0.000008 0.000008
(d) 4 x 3.33 MW
0 0 13.33 A
4
0.92236816 1.00000000
1 3.33 10 4A
3
U 0.07529536 0.07763184
2 6.66 6.66 6A
2
U
2
0.00230496 0.00233648
3 10 3.33 4AU
3
0.00003136 0.00003152
4 13.33 0 U
4
0.00000016 0.00000016
LOLP = 0.01
LOLE = 365 x 0.01
= 3.65 d/yr
LOLP = 0.0004
LOLE = 0.15 d/yr
LOLP = 0.001184
LOLE = 0.43 d/yr
LOLP = 0.00233648
LOLE = 0.85 d/yr
Expected Load Loss
ELL = 0.2 MW
= 200 kW
Capacity (MW) Load Loss Prob L
i
x p
i
Out In L
i
(MW) p
i
(MW)
(a) 1 x 10 MW
0 10 0 0.98 0
10 0 10 0.02 0.2
0.2
(b) 2 x 10 MW
0 20 0 0.9604 0
10 10 0 0.0392 0
20 0 10 0.0004 0.004
0.004
(c) 3 x 5 MW
0 15 0 0.941192 0
5 10 0 0.057624 0
10 5 5 0.001176 0.00588
15 0 10 0.000008 0.00008
0.00596
(d) 4 x 3.33 MW
0 13.33 0 0.92236816 0
3.33 10 0 0.07529536 0
6.67 6.67 3.33 0.00230496 0.00768320
10 3.33 6.67 0.00003136 0.00020907
13.33 0 10 0.00000016 0.00000160
0.00789387
ELL = 0.004 MW
= 4.00 kW
ELL = 0.00596 MW
= 5.96 kW
ELL = 0.00789 MW
= 7.89 kW
Comparative Analysis
On the basis of the expected load loss only the alternative b is
most reliable
System LOLE D/YEAR INVESTMANT
1X10MW 3.65 1 P U
2X10MW 0.15 2
3X5MW 0.43 1.5
4X3.33MW 0.85 1.33
On the basis of investment alternative c can be adopted though
it have higher LOLE
Effect of unit unavailability
System ELL (kW) at Different Unit FORs
2% 4% 6%)
(a) 1 x 10 MW 200 400 600
(b) 2 x 10 MW 4.00 16.00 36.00
(c) 3 x 5 MW 5.96 23.68 52.92
(d) 4 x 3.33 MW 7.89 31.12 69.09
These result indicate that the ELL and any reliability index is
very Sensitive to an increase in value of unavailability (FOR)
System with Non-identical Components
All components must be identical to apply the Binomial Distribution.
If components of a system have non-identical capacities:
- Units with identical capacities are grouped together
- COPT is developed for each group
- COPT for different groups are combined, one at a time
- Final COPT for the system is used for reliability evaluation
A pumping station has 2 x 20 t/hr units, each having an unavailability
of 0.1, and 1 x 30 t/hr unit with an unavailability of 0.15.
Calculate the capacity outage probability table for this plant.
COPT for 2 x 20 t/hr units:
COPT for 1 x 30 t/hr unit:
Units
Out
Cap Out
(t/hr)
Cap In
(t/hr)
Prob
Units
Out
Cap Out
(t/hr)
Cap In
(t/hr)
Prob
0 0 40 0.81 0 0 30 0.85
1 20 20 0.18 1 30 0 0.15
2 40 0 0.01
System with Non-identical Components
Combining COPTs:
Each cell contains: Capacity In / probability
40 / 0.81 20 / 0.18 0 / 0.01
30 / 0.85 70 / 0.6885 50 / 0.153 30 / 0.085
0 / 0.15 40 / 0.1215 20 / 0.027 0 / 0.0015
1 x 30
t/hr unit
2 x 20 t/hr units
Overall System COPT:
Cap In
(t/hr)
Cap Out
(t/hr)
Prob
70 0 0.6885
50 20 0.1530
40 30 0.1215
30 40 0.0850
20 50 0.0270
0 70 0.0015
Example
A 60 MW generation plant consists of one 20 MW
and four 10 MW units. The plant must satisfy a
load demand of 45 MW. What is the Loss of Load
Probability (LOLP) and the Expected Load Loss
(ELL)? Each 10 MW unit has an unavailability (or
Forced Outage Rate, FOR) of 0.1, and the 20 MW
unit has a FOR of 0.15.
Mixture of the exponential Distribution
Let
i
= denotes the failure rate of the items
coming from the plant I for i=1,2 . . The items
are mixed up before they are sold . The fraction
of p is coming from plant 1 and rest (1-p) from
plant 2
I f we pick one item randomly the survivor
function of the item is
Mean Time to Failure MTTF
Mixture of the exponential Distribution
The Failure Rate function
The failure rate function is seen
to be decreasing . If 1> 2 early
failure should have failure rate close
to 1. After a while all the week
components have failed we are left
with components with lower failure
rate 2
Reliability Block Diagram
A reliability block diagram is success- oriented network
describing function of the system .
It shows the logical connection of the components needed
to fulfill the specified system function
If the system has more than one function , each function
must be considered individually and separate reliability
block diagram has to be established for each system
function
Reliability block diagram are suitable for the system of
non repairable components and where the order in which
failures occurs doesnt matter
When the system are repairable and the order in which
failures occurs are important Markov Method are more
suitable
Reliability Block Diagram
For the system functioning i
th
component must
functioning . This doesnt mean component i
function in all accepts . It mean one ore
specified set of the function is achieved
1 2 3 i n
Structure Function
The state of component i (i= 1,2, n) can be described
by binary variable x
i
where
X={x
1
, x
2
, x
3
..x
n
} is called state vector .By knowing
the state of n components we knom whether the system
is functioning or not
State of system |(x) =|( x
1
, x
2
, x
3
..x
n
)
This |(x) is called the structure function
Series Structure
The structure function of a series structure
is
A system that is functioning if and only if
all of its n components are functioning is
called series structure
The survivor function of a series structure is
1 2 3 i n
Series Structure
The failure rate function z
s
(t) of a series
structure( of independent components) is equal to
the sum of the failure rate functions of the
individual components:
Series Structure
Consider a series structure of n independent
components with constant failure rates i
for i = 1, 2, . . . , n. The survivor function is
Failure rate of the system
s
=
1
+
2
+
n
Parallel Structure
A system is functioning if at least one of its n
components is functioning is called parallel
structure
The structure function of the Parallel structure
is
n
i
1
2
Parallel Structure
The survivor function of a non repairable
parallel structure of independent
components is
When all the components have constant failure
rates zi(t) = i, for i = 1, 2, . . . , n, then
Parallel Structure
Consider a parallel structure of two independent
components with failure rates 1 and 2
respectively. The survivor function is
Parallel Structure
K out of N system
A system that is functioning if at least k of n
components are functioning is called k out of N
structure(KOON)
The structure function of a k-out of n structure may
be written as
When all the n components have identical
reliabilities p
i
(t) = p(t) , the variable is binomially
distributed (n,p(t)
In this case
K out of N system
A system has three components 1, 2 ,3
2003 system can be represented as
Structure Function
Since x
i is
binary variables so x
i
k
=x
i
1 2
2
3
1 3
MTTF of k-out-of-n Structures
Pivotal Decomposition
Conditional Probability approach
p(system success or failure )= p(system success or
failure if component x is good).p(x is good)+
p(system success or failure if component x is
bad).p(x is bad)
By pivotal decomposition the structure function
may be written
|(X(t)) = Xi(t) | (1i,X(t)) + (1 Xi(t)) | (0i,X(t))
= Xi(t) [| (1i,X(t)) | (0i,X(t))] + | (0i,X(t))
Pivotal Decomposition
Consider the system represented by the reliability
block diagram.
Assume that the components of the system are independent with the
following function probabilities (reliabilities): p1 = 0.90, p2 = 0.95, p3
= 0.85, p4 = 0.90, p5 = 0.80.
Find the system reliability ps using Pivotal
Sol
Considering i
th
component is good and bad both the structure
function is expressed as
Assume component 2 doesnt work then the reliability can be
reduced as
( )
( ) ( ) ( )
1 1 0
i i i i
X x x x x | | | = +
Pivotal Decomposition
Assume component 2 works then the component 1 becomes
irrelevant component the reliability can be reduced as shown in below
If the system components have independent probabilities
(reliabilities p1 = 0.90, p2 = 0.95, p3 = 0.85, p4 = 0.90, p5 = 0.80.)
Hence substituting the values of reliability =0.9949
( ) ( ) { }
0 3 1. 4 5
i
x x x x x |
(
=
( ) ( )
1 3 4 5
i
x x x x | =
( )
( ) ( ) ( ) { }
2
2
3 4 5 1 3 1. 4 5 X x x x x x x x x x |
(
= +
( )
( ) ( ) ( ) { }
2
2
3 4 5 1 3 1. 4 5 p X p p p p p p p p p
(
= +
Critical Component
The i
th
component is critical for the system if the
system is in such a state that system is functioning
if and only that i
th
component is functioning
43
Let R=P [Success] Q=P [Failure]
R+Q=1
Series Systems
[
=
=
=
n
1 i
i
2 1 s
R
R R R
2 1 2 1
2 1
2 1
S s
Q Q Q Q
) Q )(1 Q - (1 - 1
R R 1
R 1 Q
+ =
=
=
=
2 1
product rule of reliability
Reliability Network Modeling
44
Example 4.1:
A system consists of 10 identical components, all of which must
work for system success. What is the system reliability if each
component has a reliability of 0.95?
Component reliability, R = 0.95
Number of components, n = 10
Using Product Rule of Reliability,
System Reliability, R
s
= R
n
= (0.95)
10
= 0.5987
45
Series System
Assume each component has a reliability of 0.9.
Number of Components Reliability
1
2
3
4
5
10
20
50
0.9
0.81
0.729
0.6561
0.59049
0.348678
0.121577
0.005154
46
Parallel Redundant Systems
2 1 s
Q Q Q =
2 1 2 1
2 1
2 1
S s
R R R R
) R )(1 R - (1 - 1
Q Q 1
Q 1 R
+ =
=
=
=
2
1
product rule of
unreliability
47
Example 4.6:
A system is to be designed with an overall reliability of 0.999 using components
having individual reliabilities of 0.7. What is the minimum number of
components that must be connected in parallel?
System Reliability, R
s
= 0.999
Component reliability, R = 0.7
Number of components, n = ?
System Unreliability, Q
s
= 1 - R
s
= 1 0.999 = 0.001
Component Unreliability, Q = 1 R = 1 0.7 = 0.3
Using Product Rule of Unreliability, Q
s
= Q
n
i.e. 0.001 = (0.3)
n
therefore, n = ln (0.001) / ln(0.3) = 5.74
since n is an integer, n = 6
48
Parallel System
Number of Components Reliability
1
2
3
4
5
0.9
0.99
0.999
0.9999
0.99999
49
Series/Parallel Systems
4
2
3
1
Redundant
Rs = ] R R R R R [R R
4 3 2 4 3 2 1
+
network reduction technique
50
Evaluate the unreliability of the system if all components have
a reliability of 0.8.
Q
8
= Q
7
. Q
5
= (1 - R
7
). Q
5
= (1 - R
1
. R
2
. R
6
). Q
5
= [1 - R
1
. R
2
. (1 - Q
6
)]. Q
5
= [1 - R
1
. R
2
. (1 - Q
3
. Q
4
)]. Q
5
Q
8
= [1 0.8 x 0.8 (1 0.2 x 0.2)]x 0.2 = 0.07712
51
Network Modelling complex systems
A C
B D
E
System success requires continuity from input to output.
input
output
Evaluation Techniques:
- Conditional probability approach
- Cut set method
- Tie set method
- Connection matrix technique
- Tree diagrams
52
Cut Set Method
Cut Set:
set of system components which, when failed, causes system failure
set of components which if removed from the network separate the input from the output
Cut Sets:
AB, CD - 2
nd
order
ABE, ABC, ABD, ABE, AED, BEC, CDA, CDB, CDE
ABCD, ABCE, ABDE, ACDE, BCDE
ABCDE 5
th
order
A C
B D
E
Minimal Cut Set:
any cut set which does not contain any other cut set as a subset
all components of a minimal cut set must fail in order to cause system failure
Minimal Cut Sets:
AB, CD - 2
nd
order
AED, BEC 3
rd
order
53
Cut Set Method contd..
P(C
1
) = Q
A
.Q
B
P(C
2
) = Q
C
.Q
D
P(C
3
) = Q
A
.Q
D.
Q
E
P(C
4
) = Q
B
.Q
C.
Q
E
Q
S
= P(C
1
C
2
C
3
C
4
)
= P(C
1
) + P(C
2
) + P(C
3
) + P(C
4
)
- P(C
1
C
2
) - P(C
1
C
3
) - P(C
1
C
4
)
- P(C
2
C
3
) - P(C
2
C
4
) - P(C
3
C
4
)
+ P(C
1
C
2
C
3
) + P(C
1
C
2
C
4
)
+ P(C
1
C
3
C
4
) + P(C
2
C
3
C
4
)
- P(C
1
C
2
C
3
C
4
)
Minimal Cut Sets:
C
1
-> AB A and B in parallel, since both must fail for system failure
C
2
-> CD
C
3
-> AED C1, C2, C3 and C4 in series, since all 4 must be successful for system success
C
4
-> BEC
C
D
A
B
A
D
E
B
C
E
C
1
C
2
C
3
C
4
P(C
1
C
2
) = P(C
1
).P(C
2
) = Q
A
.Q
B
Q
C
.Q
D
P(C
1
C
3
) = P(C
1
).P(C
3
) = Q
A
.Q
B
Q
D
.Q
E
P(C
1
C
4
) = P(C
1
).P(C
4
) = Q
A
.Q
B
Q
C
.Q
E
..
P(C
3
C
4
) = P(C
1
C
2
C
3
)
= P(C
1
C
2
C
4
) = P(C
1
C
3
C
4
)
= P(C
2
C
3
C
4
) = P(C
1
C
2
C
3
C
4
)
= Q
A
.Q
B
Q
C
.Q
D
.Q
E
54
Cut Set Method contd..
Q
S
= Q
A
.Q
B
+ Q
C
.Q
D
+ Q
A
.Q
D.
Q
E
+ Q
B
.Q
C.
Q
E
- Q
A
.Q
B
Q
C
.Q
D
- Q
A
.Q
B
Q
D
.Q
E
- Q
A
.Q
B
Q
C
.Q
E
- Q
A
.Q
C
Q
D
.Q
E
- Q
B
.Q
C
Q
D
.Q
E
- Q
A
.Q
B
Q
C
.Q
D
.Q
E
+ 4 Q
A
.Q
B
Q
C
.Q
D
.Q
E
- Q
A
.Q
B
Q
C
.Q
D
.Q
E
Q
S
= Q
A
.Q
B
+ Q
C
.Q
D
+ Q
A
.Q
D.
Q
E
+ Q
B
.Q
C.
Q
E
- Q
A
.Q
B
Q
C
.Q
D
- Q
A
.Q
B
Q
D
.Q
E
- Q
A
.Q
B
Q
C
.Q
E
- Q
A
.Q
C
Q
D
.Q
E
- Q
B
.Q
C
Q
D
.Q
E
+ 2Q
A
.Q
B
Q
C
.Q
D
.Q
E
If Q
A
= Q
B
= Q
C
= Q
D
= Q
E
= Q, then
Q
S
= 2Q
2
+2Q
3
5Q
4
+ 2Q
5
If Q = 0.01, Q
S
= 0.00020195
R
S
= 0.99979805
55
Advantages of Cut Set Method:
- cut sets identify ways in which a system may fail
- approximation can be used to simplify evaluation
- can be easily programmed on a computer
1st Approximation
Q
S
= P(C
1
C
2
C
3
C
4
)
= P(C
1
) + P(C
2
) + P(C
3
) + P(C
4
)
- P(C
1
C
2
) - P(C
1
C
3
) - P(C
1
C
4
)
- P(C
2
C
3
) - P(C
2
C
4
) - P(C
3
C
4
)
+ P(C
1
C
2
C
3
) + P(C
1
C
2
C
4
)
+ P(C
1
C
3
C
4
) + P(C
2
C
3
C
4
)
- P(C
1
C
2
C
3
C
4
)
~ P(C
1
) + P(C
2
) + P(C
3
) + P(C
4
)
~ Q
A
.Q
B
+ Q
C
.Q
D
+ Q
A
.Q
D.
Q
E
+ Q
B
.Q
C.
Q
E
If Q
A
= Q
B
= Q
C
= Q
D
= Q
E
= Q, then
Q
S
= 2Q
2
+2Q
3
If Q = 0.01, Q
S
= 0.000202, R
S
= 0.999798
n
For n cut sets, Q
S
= E P(C
i
)
i=1
2nd Approximation
Neglect higher order cut sets
(events with very low probabilities)
1
st
order: none
2
nd
order: AB, CD
3
rd
order: ADE, BCE
Neglecting cut sets higher than 2
nd
order
Q
S
~ P(C
1
C
2
)
~ P(C
1
) + P(C
2
)
~ Q
A
.Q
B
+ Q
C
.Q
D
If Q
A
= Q
B
= Q
C
= Q
D
= Q
E
= Q, then
Q
S
= 2Q
2
If Q = 0.01, Q
S
= 0.0002, R
S
= 0.9998
56
Consider:
2
3
1
Cuts Min Cuts Probability
1,3
2,3
1,2,3
1,3
2,3
---
Q
1
Q
3
Q
2
Q
3
-
Q
s
~ Q
1
Q
3
+Q
2
Q
3
Complete Equation:
3 2 1 3 2 3 1
2 1 2 1 3 S
Q Q Q Q Q Q Q
] Q Q Q [Q Q Q
+ =
+ =
57
Example:
Calculate the reliability of the system below if all
the individual components have a reliability of 0.95.
58
Probability Density Function, f(x)
(see p. 43) - plot of Probability vs. Random Variable x
(Cumulative) Probability Distribution Function, F(x) = f(x).dx f(x) = dF(x)
dx
For discrete distribution, F(x) = f(x)
In reliability evaluation, the random variable is usually time (t)
Failure Density Function, f(t)
(Cumulative) Failure Distribution Function, Q(t) -- Probability of Failure
Survivor Function R(t) = 1 Q(t) -- Probability of Success
Hazard Rate, (t) = f(t) / R(t) = # of failures per unit time
# of components exposed to failure
Probability Distributions
59
Total number of samples = N
0
Number of failures at time t = N
f
(t)
Number of survivors at time t = N
s
(t)
Number of failures in interval At = AN
f
(t)
Failure Density Function, f(t) = AN
f
(t) / N
0
Failure Distribution Function, Q(t) = N
f
(t) / N
0
Probability of Failure
Survivor Function R(t) = N
s
(t) / N
0
Probability of Success
Hazard Rate, (t) = f(t) / R(t)
Example using Discrete Distribution
60
Bathtub Curve
Typical Electric Component Hazard Rate as a Function of Age
De-Bugging
Normal operating
Or useful life
Wear out
Region 1 Region 2
Region 3
Operating Life
H
a
z
a
r
d
r
a
t
e
61
Exponential Distribution
System
Up
System
Down
R(t) = Q(t) = 1 - f(t) =
is the constant failure rate
t
e
t
e
t
e
=
=
=
=
+
t -
)t ( -
t t
2 1 s
i
2 1
2 1
e
e
e e
R R R
63
Parallel Systems
2
1
)t ( - t t
S
2 1 2 1 s
2 1 2 1
e e e R
R R R R R
+
+ =
+ =
64
Complex Systems
*Develop the basic equations
*Substitute
t
t
e 1 Q(t)
e R(t)
=
=
Redundancy
Two way to ensure higher system reliability
use components of higher reliability in critical places
Introduce redundancy in those places
Redundancy
In an entity, the existence of more than one means for performing
a required function [IEC50(191)]
Existence of means, in addition to the means which would be
sufficient for a functional unit to perform a required function or
for data to represent information [IEC61508, Part 4]
Redundant items may be classified as:
Active (warm) redundancy
Passive redundancy
Cold redundancy
Partly loaded redundancy
Passive Redundancy or perfect switching
T
i
time to failure item i for i=1,2n
Life time of the whole stand by system
Mean time to failure
For Two stand by system n=3
For Two stand by system n=3
Cold Standby, Imperfect Switching
Consider a system of n = 2 components with failure rates 1
and 2. Let the probability of successful switching be 1 p.
The system may survive (0, t] in two disjoint ways:
Item 1 does not fail in (0, t] (i.e., T1 > t).
Item 1 fails in a time interval (t, t + dt] where 0 < t < t.
The switch S is able to activate item 2. Item 2 is activated
at time t and does not fail in the time interval (t, t].
The probability of event 1 is Pr(T1 > t) = e
1
t