SYSTEM RELIABILITY

EVALUATION
1
 System reliability evaluation using
probability distributions
Series systems
The reliability of a 2-component series system

RS (t )  R1 (t )  R2 (t )

RS (t )  exp   1 (t )dt   exp   2 (t )dt 

t t

 0   0 
for an n-component series system
n
RS (t )   exp   i (t )dt 
 t

 
2

i 1  0 
In the special case of the exponential distribution,
For a 2-component series system

RS (t )  exp  1t  exp  2t 

 exp  1  2  t 
for an n-component series system

n
  n
RS t    exp  i t   exp    i t 
i 1  i 1 
3
 for the general case
n
  
RS t   exp   e t dt   exp   i t dt
t t

 0  i 1  0 
and for the exponential case

  n
RS t   exp  et   exp    i t 
 i 1 
The effective failure rate of a series system containing
components whose reliabilities follow exponential distributions is
n
e   i 4

i 1
Example
A simple electronic circuit consists of 6 transistors each having a failure rate of
1O-6f/hr, 4 diodes each having a failure rate of O.5x 1O-6 f/hr. 3 capacitors each
having a failure rate of 0.2x 10-6 f/hr,10 resistors each having a failure rate of 5
x 10-6 f/hr and 2 switches each having a failure rate of 2 x 10-6 f/hr. Assuming
connectors and wiring are 100% reliable (these can be included if considered
significant), evaluate the equivalent failure rate of the system and the probability
of the system surviving 1000 hr and also 10 000 hr if all components must
operate for system success.

e  6 x1x10  4 x0.5x10  3x0.2 x10 

6 6 6

 10 x5 x10   2 x2 x10   6.26 x10 f / hr

6 6 5

R 1000  exp  6.26 x10 x1000  0.9393

S
5

R 10000  exp  6.26 x10 x10000  0.5347 5 5

S
Parallel systems
 
n
 
QP (t )   1  exp   i (t )dt 
t

i 1  
 0 


RP t   1   1  exp   i (t )dt  
n
 t

i 1 
 0  
The special case of a constant hazard rate, i.e., the exponential
distribution, we obtain, for a 2-component parallel system

QP (t )  1  exp  1t 1  exp  2t 

exp  1t   exp  2t 
 1  
 exp  1  2 t   6
RP (t )  exp  1t   exp  2t 
 exp  1  2  t 

for an n-component parallel system

n
QP t    1  exp  i t 
i 1
n
RP t   1   1  exp  i t 
7
i 1
Reconsider the previous Example and evaluate the probability of
surviving 1000 hr and 10 000 hr if (a) two, (b) three identical
circuits to those described in the previous Example are used in
parallel and it is assumed that the system operates successfully if
only one of the circuits is successful.
In this example, all the parallel components (each being
equivalent to one of the series circuits) are identical and each
having a failure rate given by λe in the previous Example, i.e.,
i  6.26 x10 5 f / hr
RP 1000  exp  6.26 x10 5 x1000
  5

 exp  6.26 x10 x1000  exp 
2 x 6.26 x10
5


 x1000 
 0.9963 8
 
RP 10000  2 x exp  6.26 x10 x10000 5


 exp  2 x6.26 x10 x10000 5

 0.7835
Partially redundant systems
Rt   e  t

and Qt   1  e t

The binomial expression

e t
 1  e t
 n 9
 Consider a system comprising 4 identical units each having a
failure rate of 0.1 f/yr. Evaluate the probability of the system
surviving 0.5 yr and 5 yr if at least two units must operate
successfully.
Using the binomial expression for n =4 gives

Rt   Qt  4
 R t   4R t Qt   6R t Q t 
4 3 2 2

 4Rt Q t   Q t 
3 4

  0.1 f / yr , t  0.5 yr
R0.5  0.9996
R5.0   0.8282
10
Number of Probability of system success
units required
for success
4 4  t
e
3
e 4 t
 4e 3t
1  e   t

2 e 4 t
 4e 3t
1  e   t

 6e  2 t
1  e  t 2

1  e 
1
4 t 3t  t
e  4e
 6e  2 t
1  e   4e 1  e 
 t 2  t  t 3
11
In the more general case of non-identical units, the probability of
each system state

R1 t   Q1 t R2 t   Q2 t ........Rn t   Qn t 

Ri t   exp  i t  for i  1,2,...., n

Qi t   1  exp  i t  for i  1,2,...., n

Mean time to failure (MTTF)

E t    t f t  dt
t
12
0
f (t ) is the distribution of times to failure, this expected value
was the mean time to failure or MTTF, which can be
designated as m


m   t dRt 
0

Integration by parts gives

m   t Rt    Rt dt   Rt dt

  
0 0 0

The MTTF of any series or parallel system can therefore be

evaluated by using the appropriate system expression for R(t) and
integrating between the limits (0,∞ ). 13
Consider the particular case of exponential distributions.
For a series system

  n 
m   RS t dt   exp    i t   dt
0 0
 i 1 
1 1
 n 
1  2  ...  n

i 1
i

For a 2 parallel system


m   RP t  dt 14
0
m   exp  1t   exp  2t   exp 1  2 t dt



 1 

  exp   t  
1
exp   t 

1 2
 1 
 2

exp  1  2 t 
 1 
 1  2 0
1 1 1
  
1 2 1  2
15
For an n-component parallel system

1 1 1 
m     ...  
 1 2 n 
 1 1 1 
   ...   ...
       
 1 2 1 3 i f 
 1 1 1 
   ...   ... ...
           
 1 2 3 1 2 4 i f k 
  1
n 1 1
n


i 1
i 16
 Standby systems
B

A simple and most basic standby system is shown. A represents

the main operating component, B the standby component and S
the sensing and changeover switch.
One assumption which is used through out is that the
components, both normally operating and standby, have a
constant hazard rate, i.e., that failures are described by
exponential distributions.
17
Perfect switching

t  e
(a) 2- component system x  t
P t  
x
x!
Pno components fail   P0 t   e  t

Pexactly one component fails   P1 t   te t

Therefore, the reliability of the system is

Rt   P0 t   P1 t   e  t
1  t 
18
 (b) 2 standby components

Rt   P0 t   P1 t   P2 t 
 t
 t  
2
e 1  t  
 2! 

(c) n standby components

Rt   e  t
 t  t 
2 3
t  
n

1  t    ...  
 2! 3! n! 


n
t  e x  t
19

x 0 x!
 (d) Mean time to failure

m e  t
1  t dt  1

1

2
0   
and for n standby components (case (c)),

 n t  e
x  t
n 1
m
0

x 0 x!


Example
Compare the reliability of a 2-component system each having a
failure rate of 0.02 f/hr after a time of 10 hr if they are (a) parallel
redundant and, (b) standby redundant with a 100% reliable
sensing and changeover device. Also, compare the MTTFs of20the
two systems,
(a) Parallel system.

R10  2e 0.02 x10

e 2 x 0.02 x10
 0.967141
1 1 1
m    75 hr
0.02 0.02 0.02  0.02

(b) Standby system.

R10  e 0.02 x10

1  0.02 x10  0.982477
2
m  100 hr
0.02 21
 Imperfect switching
The value of PS can be established in practice from a data
collection scheme which records the number of successful
and failed operations of the device since

number of successful operations

PS 
total number of requested operations

Returning to the 2-component standby system in which the

switching and sensing device is less than 100% reliable, system
success requires that either no failures occur or one failure occurs
and the switching device operates.

Pno components fail   P0 t   e  t 22

 exactly one component fails and 
P 
the sensing and switching device operates
 P1 t   PS  PS te  t

Rt   e  t
1  PS t 
 1  PS
m e  t
1  PS t  dt 
0 

23
Effect of spare components
Consider a system comprising N identical components all of which
must operate for system success and consider that there are it
spares available to the operating personnel as standby
components. It follows that n failures in the system can be
tolerated and only the (n + 1) failure causes system failure. This
logic also indicates that the spare components are themselves not
replaced following their use as a normally operating component.
Again, consider the case of exponential distributions and assume
that the failure rate of each component is λ.

N
The failure rate of the system   i  N
i 1
24
 
Rt   e 1  Nt 
 Nt  N t 
2
 ... 
 N t  
n


 2! n! 

n 1
m
N

25
A system contains 50 identical components each of which has a
failure rate of 0.001 f/hr. Assuming that system failure occurs
when any one component fails, evaluate the system reliability for
an operating period of 20 hr and the MTTF when no spares are
available and also when a varying number of spares between 1
and 6 are carried as immediate replacements. If the system is to
have a minimum reliability of 0.9950, what is the minimum number
of spares that must be carried as immediate replacements?

System failure rate  N

 50 x0.001  0.05 f / hr

Rno spare available   e0.05x 20  0.367879

26
   0.05 x 20 
2

1  0.05 x 20  
Rn spare available   e 0.05 x 20 
2! 
 0.05 x 20 
n

 ...  
 n! 

Number System Reliability MTTF m,hr

of Spare R(20)
0 0.367879 20

1 0.735759 40
2 0.919699 60
3 0.981012 80
4 0.996340 100
5 0.999406 120
27
6 0.999917 140
mno spare available  
1 1
  20 hr
N 0.05
n 1 n 1
mn spares available   
N 0.05
for n  0,1,2,....,6

28
 Non-identical components
In order to illustrate the joint density function approach, consider
two components A and B in a standby system and having failure
rates λa and λb respectively. Let A be the normal operating
component and B be the standby one. Assume A fails at time t1
when B takes over instantaneously. Assume B fails at time t. Time
to failure of B is then t2 = t - t1

Failure density function of A, f a t1   a exp  a t1 

Failure density function of B, f b t 2   b exp  bt 2 

The joint density function of both components operating is given

by
f t   f a t1   f b t2  29
 f t   a exp  a t1   b exp  bt2 
 a exp  a t1   b exp  b t  t1 

There are two time functions, t1 and t. In order to obtain the joint
density function in terms of t only, f(t) is integrated with reference
to t1 This gives

f t    a b exp  at1  exp  b t  t1 dt1

t

t1 0

a b
 exp  bt   exp  at 
a  b

30
The system reliability can now be evaluated from
 
Rt    f t dt
t

a b 
  exp  bt   exp  a t 
a  b t
a b
 exp  bt   exp  a t 
a  b a  b

By adding and subtracting exp (-λat) and rearranging to give

a
Rt   exp  a t   exp  at   exp  bt 
b  a
31
The MTTF is given by


m   Rt  dt 
1 1

0 a b
If the sensing and changeover device is not 100% reliable,

PS a
Rt   exp  a t   exp  at   exp  bt 
b  a
This approach can be extended to include any number of
standby components. Consider a standby system consisting
of one normally operating component having a density
function f1(t1) and (n-1) components in standby having
density functions f2(t2), f3(t3), …….fn(tn). 32
The joint density function is
 f t    f1 t1  f 2 t 2 ,...,
t t3 t2
,...,  
t n1  0 t 2  0 t1  0

f n t n  dt1dt 2 , , dt n 1

where
t1 represents the time of failure of component 1
t2 represents the time of failure of component 2
tn represents the time of failure of component n
and tn = t - tn-1

33
Failures in the standby mode
(a) Case 1
Consider the case of two non-identical components A and B forming the standby system
described in Section 7 in which the standby component B cannot fail in the standby
mode. The events leading to the system success are either
(i) Component A does not fail for an interval of time 0 to t, or
(ii) Component A fails at time t1< t and component B does not fail in the interval t1 to t.
Let R1 and R2 be the reliabilities associated with these two events respectively, then

R1  exp  at 

 function for one failure  dt1

t
R2   density
t 0
1

 probabilit y of one failure in t1 

t

t1  0

 probabilit y of no failure in dt1  dt1

34

t

t1  0

a exp  a t1 exp  b t  t1 dt
1

 a exp  bt1  exp  a  b t1 dt1

t

a
 exp  bt 1  exp  a  b t 
a  b
a

a  b
exp  a t   exp  bt 

35
Since the events (i) and (ii) are mutually exclusive

Rt   R1  R2
a
 exp  a t   exp  at   exp  bt 
b  a

36
Graphical representation of Case 1. (a) Contribution R1. (b) Contribution R2
The contribution R1 is simply given by the area under the exponential
curve associated with component A for times greater than t, i.e., R1 = exp (-λat)

The contribution R2 for a given value of t1 is the area under the exponential curve
associated with component B [RB(t1-t2)] weighted by the probability that component A
fails in the time interval dt, at t1.
That is,

R2 t1   RB t  t1   a exp  a t1 dt1

 exp  b t  t1  a exp  a t1 dt1
The value R2 can then be determined by integrating R2(t1) over all values of t1

37
(b) Case 2
Consider now the case of 2 components in a standby system for which λ1 is the failure
rate of the normal operating component, A is the failure rate of the standby component
when operating and λ2 is its failure rate when in a standby mode.
Using a similar logic to that used in Case 1, the mutually exclusive events leading to
system success are
(i) Component 1 does not fail during the time (0-t),
(ii) Component 1 fails at time t1 and component 2 is not failed at time t1 (failure rate λ3)
and component 2 does not fail in the time (t-t1) (failure rate λ2).

Event Mode of operation in time domain of

Component 1 Component 2
1 Good for time t1
2 Fail at time t1 Good for time t1
Good for time t1 to t
38
Using the same logical derivation as in case 1

R1  exp  1t 
1 exp  1t1  exp  3t1  exp  2 t  t1 dt1
t
R2  
t1  0

 1 exp  2t1  exp  1  3  2 t1 dt1

t

 exp  1  3  2 t 
 1 exp  2t1 
1
 
 1  3  2 1  3  2 
1
 exp  2t1   exp  1  3 t 
1  3  2

39
And the system reliability is

Rt   R1  R2
1
 exp  1t   exp  2t   exp  1  3 t 
1  3  2
(c) Case 3
Consider the system shown below in which components 1 and 2 operate as a parallel
redundant system and component 3 is used when both 1 and 2 have failed.

40

System used for Case 3

In this case assume that the following system data is
available:

all components operate in their useful life period and wear-out can be
neglected
failure rate of component 1 when energized= λ1e
failure rate of component 2 when energized = λ2e
failure rate of component 3 when energized = λ3e
failure rate of component 3 when on standby = λ3s
failure rate of the sensing device = λs
failure rate of changeover device when on standby = λcs
failure rate of changeover device after switching = λce
probability of successful changeover = Ps

41
 Mode of Operation in time domain of component

Event 1 2 3 3 Sensing Changeover Device

Standby Energized device Stand Energized After

by Switching
1 Good/t Good/t

2 Good/t Bad/t

3 Bad/t Good/t

4 Bad/t1 Bad/t2 Good/t2 Good/t-t2 Good/t2 Good/ Good/one Good/t-t2

t2 cycle

5 Bad/t2 Bad/t3 Good/t2 Good/t-t2 Good/t2 Good/ Good/one Good/t-t2

t2 cycle

42
 t  t2  t1
5
Rt    Ri t 
i 1

R1 t   exp  1et  exp  2 et 

R2 t   exp  1et 1  exp  2 et 
R3 t   exp  2 et 1  exp  1et 
R4 t    1e exp  1et1   2e exp  2et 2 
t t2

t 2  0 t1  0

x exp  3 s t 2   exp  3e t  t 2 

x exp  s t 2   exp  cs t 2   Ps  exp  ce t  t 2 dt1dt 2


  1  exp  q t      
1 1 

 2 e Ps exp  3e  ce t  q q 1e 


 x1  exp     t   
43

 q 1e 
where q  2e  3s  s  cs  3c  ce
similarly ,
 1 
R5 t   1e Ps exp  3e  ce t   
1  exp  q ' t  
 q ' 


1
q '2e

1  exp  q '2e t  
where q '  1e  3s  s  cs  3e  ce
and Rt   R1 t   R2 t   R3 t   R4 t   R5 t 
44
 Problems

1.

4
S

1. For the standby system shown, mission success requires at least two
components. Components 1, 2, 3 are in parallel and component 4 is in stand-by.
Assuming 100% reliable sensing and changeover arrangement, develop the
expression for the system reliability. Assume constant failure rates λ1, λ2, λ3 and
λ4 for components 1, 2, 3, and 4 respectively.
45
 2 out of
2. 1 out of 3 requirement
2 requirement
4
2

5
1
2
6
1

Redundant
paths

3 7

Calculate the reliability of the system shown for a 100 hr mission.

1  12 x10 f / hr 2  2 x10 f / hr
5 4

3  1x10 5 f / hr 4  5  10 x10 5 f / hr 46

6  5 x10 5 f / hr 7  10 x10 5 f / hr
3.
1 2 3 4
input

output
5 6

input
7 8 9 10

Consider the system shown. Assume that the signal can flow only in the
directions shown.
(a). Develop an equation for the reliability of this system.
(b). Use your equation to calculate the system reliability if all components have a
reliability of 0.9.
47