Reliability

Introduction
Reliability is the capability of a device to work without any breakdown. It is defined with respect to a
time horizon and combines the time dimension with the performance level of equipment. Formally, it
is defined as the probability of a device performing satisfactorily for a given period of time when it is
operated in the manner specified and for the purpose it is intended.
Use of the concept of reliability dates back to the time when multiengine aircraft emerged
between the World Wars I and II. Probability concepts were used to evaluate the chances of failure of
one out of two engines or one out of four engines during flights. A complex system such as an aircraft
or a computer system not only requires smooth functioning of the components but also depends upon
its design as well as on various interconnections and switchover devices. The environmental
conditions like pressure, humidity, vibrations and shocks also affect the reliability of a system.
Although originated from aeronautical and various defence equipments, the concept of reliability is
also being used in industry specially in manufacturing instruments.
The idea of reliability is of help to the manufacturer in copying with the occurrence of
malfunctions or failures of the products in service. While the quality of the components of a
manufacturing item is very important for the finished product, it has been observed that the
performance level as measured by reliability can be significantly improved by using an appropriate
design.
The reliability of a system is a function of its components. The occurrence of failure in a
component can not be predicted with certainty as it is a statistical variable. The study of reliability of
various components requires a significant amount of data collection and its analysis. The feedback of
information goes from the user to the repairer, the manufacturer right upto the designer.
The reliability of equipment is estimated from the reliability of its components. Statistical
procedures such as Life Testing can be used to estimate the reliability of the different components.
Sometimes failure rates under conditions of extreme stress tend to infer the rates under normal
working conditions. Reliability estimating and allied problems have thus become an important branch
of statistics.

Definition of Reliability
According to the Aeronautical Radio Inc. (1994), the definition of reliability is as follows:

“Reliability is the probability that a system will perform satisfactorily for at least a

given period of time when used under stated conditions”.

Therefore, the probability that a system successfully performs as designed is called “system

reliability” or the “probability of survival”. So, the reliability is defined as the probability of a device

performing its intended purpose adequately for the period of time intended under the operating

conditions encountered. The reliability is the probability with which the devices will not fail to

perform a required operation for a certain length of time. Such problem is known as the problem of

survival. This definition brings into the focus of four important factors, viz.

(i) The reliability of a device is expressed as a probability.

(ii) The device is required to give adequate performance.

(iii) The duration of adequate performance is specified.

(iv) The environmental or operating conditions are specified.

However, in practice, even the best design manufacturing and maintenance efforts do not completely
eliminate the occurrence of failure. During the life of a system we may experience three distinct types
of failures – early failures, random failures and wear out failures.

Early failures
Early failures are those which occur in the early life of a system operations.

Random or Chance failures

Chance failures are predominant during the actual operating of the system and occur at random,
irregularly and unexpectedly. The phase during which only chance failures occur is called the useful
life of the system.

Wear-out failure
Wear – out failures are caused due to ageing any wearing out of components. These failures occur if
the system maintain properly or not maintain at all. The frequency of such failures increases rapidly
with time.

Measures of reliability

The prediction of system reliability is based on a number of factors such as: life characteristics,
operating conditions and the failure distribution.
If a random sample of items are taken from a population and are put to test (or use) under a set
of fixed (or given) environmental or operating conditions, some number of sample will fail

2
successively in time. The data so obtain will represent the length of each item. The length of life can
be measured depending on whether the item is repairable (Radio, TV, Aeroplane etc.) or non-
repairable (bulb, fuse, missile, rocket etc). For repairable items, the life can be measured by failure
rate or mean time between failures whereas for non repairable items the life can be measured by mean
time to failure.
Failure rate: The failure rate  is defined as the number of failure in a given time interval
Number of failure f
 
Total unit of operating hour T

Failure density
This is the ratio of the number of failures during a given unit interval of time to the total number of
items at the very beginning of the test.

Example 1: Let the total number of items at the beginning of the test i.e., the total initial population
was 1000 and during the first unit interval, the number of components that failure 130 (say). The
n1 130
failure density during the first unit interval is given by fd1    0.13 .
N 1000
Similarly, the failure density during the second unit interval can also be calculated.

Generalization
Let n1 be the number of components that fail during the first unit interval and n2 be the number that
fail during the second unit interval and so on.
n1
Let N be the total population. Then the failure density during the first unit interval is fd1  ,
N
n2
failure density during the second unit interval is fd 2  and so on.
N
Let l be the last interval after which there is no survivor, then
l

 fd
i 1
i  fd1  fd 2  ...  fdl

n1 n2 n
   ...  l
N N N
n  n  ...  nl N
 1 2  1
N N

3
Remarks:
(i) The sum of the areas of the vertical bars in the bar chart for failure density is 1.
(ii) The total number of failed items till the end of the i-th time interval say, i-th hour
 n1  n2  ....  ni

The number of survivor at the end of i-th unit time interval  N  (n1  n2  ...  ni )
iii) Hence the reliability for the i-th hour = probability of survival till i-th hour
number of survivers till the i-th hour N-(n1  n 2  ...  n i )
=R(i) (say)  
total population N
i
n1 n2 n
 1   ....  i  1  fd1  fd 2  ...  fd i  1   fd l
N N N l 1

=1-(sum of the areas of the failure density bars from the beginning till the i-th hour)

(iv) The probability of failure for i-th hour is the ratio of the number of components that fail in i hours
n1  n2  ...  ni
to the total population i.e., .
N
If this probability of failure is denoted by Q(i) then Q(i)  fd1  fd 2  ...  fd i = sum of areas of the
failure density bars from the beginning till the i-th hour

Mean time between failure (MTBF)

During the operating period when the failure rate is fairly constant, the mean time between failures is
the reciprocal of the constant failure rate i.e., m=1/ where  is the failure rate.
MTBF is also required to as the average time of satisfactory operation of the system. In this
case, larger the MTBF, the higher of the reliability of the system.

Note: The failure rate and MTBF are made when the item is repairable.

Mean time to failures

If we have a life test information on n items with failure times t1 , t2 ,..., tn then the mean time to
failures (MTTF) is defined as
1 n
MTTF   ti
n i 1

4
Note: MTTF is used when the items are non repairable.

Derivation of reliability function

Let N0 be a fixed number of components put under test repeatedly at the beginning of the test. Also,
let N f be the number of components failed during time t and N s , the number of components till

surviving after the time t.

As the time of the test increases, N f will increase and N s will decrease in the same rate. Thus the

reliability measure from such test is a function of time t.

Ns Ns
The reliability or the probability of survival is defined as R(t )  
NO N s  N f

Nf Nf
 The probability of failure at any time t can be expressed as Q(t )  
NO Ns  N f

 At any time t, R(t)+Q(t)=1.

Remarks
Nf
Since R(t) and Q(t) are mutually exclusive, then R(t )  1  Q(t )  1 
No

dR d N 1 dN f
  (1  f )  - .
dt dt NO N O dt

Again, N f  N s  N 0

i.e., Nf = N0 - Ns
dN f dN s
Hence 
dt dt
dN f dR(t )
or,   N0
dt dt
dN f dN s
This gives the rate at which components fail at any time t. Again, as  i.e., the rate of
dt dt
components failure during time dt is equal to the negative rate of components survival during time dt .
dN f dR(t )
(ii) We have   N0
dt dt
which implies

1 dN f N dR (t )
 0.
N s dt N s dt

1 dR (t )  Ns 
 Since R(t)= 
R(t ) dt  N0 

5
 1 dR (t )
  (t )  
R(t ) dt
1 dN f
where  (t ) = which is the failure rate of the components.
N s dt

Now, integrating both sides of the above with respect to t with in the limit 0 to t, we have
t t
1 dr (t )
  dt   
0 t 0
R(t ) dt
dt

t t
1
  dt   
0 t 0
R(t )
dR (t )

R (t )  
R (t )
dR N s N0
     log R 1  At t = 0, R(t )    1
1
R  N0 N0 
  log R (t )
t
Hence, log R (t ) = -   dt
0

 t 
i.e., R(t) = exp -  dt  (1)
 0 
This gives a general formula for computing reliability.

Particular case

If (t) is constant over the time i.e.,  (t) =  (say)

Then equation (1) reduces to R(t) = exp-t  = exp[-t/m], where m = MTBF

Properties of reliability function

(i) Since the reliability is a probability function, 0  R(t)  1.
(ii) R(0) = 1 and lim R (t ) = 0
t 

(iii) R(t) is a decreasing function of time.

Remarks :
1 dN f N 0 dR (T )
We have    (2)
N s dt N s dt

6
 1 dN f dN f
If  is constant, then must be constant. Again, will remain constant if the number of
N s dt dt
failed components Nf increase with time from the beginning of the test.
dN f Nf
In that case, 
dt t
1 dN f 1 Nf
 
N s dt N0 t
When failed components are substituted immediately after failure. i.e., Ns = N0 for all the time.
Now, to calculate the failure rate we need to count the number of failures Nf and the hours of
operations i.e., t. We have from (2),
1 dN f N dR (t ) dR 1 dN f
 0 i.e., 
N s dt N s dt dt N 0 dt

1 dN f
When is plotted against time t, we obtain the distribution function of failures at time t on per
N 0 dt
component basis i.e., failure frequency curve per component. This curve is known as failure density
curve.
1 dN f dR (t )
i.e., f (t )   (3)
N 0 dt dt
The area under this carve is unity.

f (t)

f  t    e  t

t
Fig. 3
1 dR 1  dR  f (t )
Therefore, the failure rate can be defined as  (t )     
R (t ) dt R (t )  dt  R (t )
This means that the failure rate of any time t is equal to f(t)/R(t).
This is applicable for all possible distribution.
Particular case
When  (t) is constant over time t i.e.,  (t) =  (say)
dR d
Then f (t )     (e  t )   e  t
dt dt

7
 1 dN f
Again, Q(t)=Nf / N0 which implies dQ/ dt = = f (t ) [ by (3) ]
N 0 dt
t
Hence, Q (t )   f (t )dt
0

This means that the probability of failure Q(t) at time t is equivalent to the area under the density
curve taken from t = 0 to t. Thus Q(t) is the cumulative probability of failure function which increases
with time t.
Conversely, we have, R(t) =1- Q(t)
t  t
 

=1-  f (t )dt =  f (t )dt   f (t )dt  Since  f (t )dt  1
0 t 0 0  0


=  f (t )dt
t
This means that the probability of survivals decreases with time t and this equal to that area under the
density curve taken from t = t to .

If an item has a constant failure rate, say,  and the probability density function is exponential then
various function can be shown as follows:

Q (t)

f  t    e  t
f (t)
R (t)

Fig. 4

Hazarad rate

Hazarad rate or instantaneous failure rate is defined as the limit of the failure rate as the

interval length approaches to zero i.e.,

 R(t )  R(t  h) 
Z (t )  lim  
h0
 hR(t ) 

1  R (t  h) - R (t )  1 dR
=  lim   
R (t ) h  0
 h  R(t ) dt
f (t )  dR 
 Since, f (t )   
R (t )  dt 
8
Relationship between Z(t) and R(T)

f (t ) dQ(t ) 1
Now, Z(t) = = [Since Q(t) +R(t) = 1 ]
R(t ) dt R(t )
t t t
1 dQ 1
which implies  Z (t )dt  
0 0
R(t ) dt
dt 
0
1  Q(t )
dQ

  log 1  Q (t ) t 0   log R(t) 0   log R(t)

t t
[Since R(0) = 1]

 t
 R(t )  exp   Z(t)dt
0 
 
t
 Q(t )  1  R(t )  1  exp   Z (t )dt 
 0 
dR(t )  t 
 f (t )    Z (t ) exp   Z (t )dt 
dt  0 
Particular case
f (t )  e  t
When the failure rate is constant then Z (t )   
R (t ) e  t
This means that for constant failure rate, the hazard rate is also constant and is equal to the failure rate.

The expected value E(t) or MTBF of a continuous random variable T is given by


E (t )  MTBF   tf (t )dt
0

where f(t) is the failure density function .

dQ(t ) dR (t )
We have f (t )  
dt dt
  
dR(t ) 
Hence, MTBF   t .dt    tdR  tR (t ) t 0   R (t ) dt [ by parts ]
0
dt t 0 0


 0   R (t )dt
0


 MTBF  m   R (t )dt
0

Particular case
For a constant failure rate we have R (t) =e-t

9
  
1
 MTBF  m   R (t )dt   e t dt 
0 0


Example 2: A device with 1000 hours useful life at a constant failure rate 0.0001 per hour in a given
experiment. What is the reliability per 10 hours of operation of this device. Find the value of MTBF?
Solution: We know that the reliability per 10 hours is R(t) =e-t
Here  = 0.0001 and t =10

Hence, R(t )  e t  e.000110  e0.001  0.999 (Approx.)

1 1
 MTBF    10, 000 hours.
 0.0001

Example 3
For equipment, the reliability per 100 hours of operation has been estimated to be 0.999. What is the
failure rate of the equipment? Calculate MTBF?
Solution:
It is given that R(t) = 0. 999
-t -t - t
Here , e is constant , so ,e =0. 999. as R(t) = e
 1- t  0.999, neglecting square and higher powers of t.
 1- 100  0.999 as t = 100
  = 0.00001.
i.e., Failure rate = 0.00001 per hours.
=10 -5per hours.
Hence, MTBF = 1/  = 10 5 hours.

System Reliability

Reliability engineering came into existence in the late 1940’s and early 1950’s. It was first

applied to communication and transportation systems. Most of the early works on reliability were

limited to the analysis of system performance. During the last several decades, numerous well-written

books on reliability have been published. Among them one may refer to the books of Grosh(1989) for

an introductory understanding, Barlow and Proschan (1965) for a theoretical foundation and

O’conner(1995) and Kapur and Lamberson (1977 ) for a practical engineering approach, Kuo [Link].

10
 (2001) for fundamentals and applications. The primary objective of reliability optimization is to find

the best way to increase the system reliability. This can be done by the following principles:

(i) Keeping the system as simple as is compatible with the performance requirements.

(ii) Increasing the reliability of each component in the system.

(iii) Using parallel redundancy for the less reliable components.

(iv) Using standby redundancy which is switched to active components when failure occurs.

(v) Using repair maintenance where failed components are replaced.

(vi) Using preventive maintenance such that components are replaced by new ones whenever they

fail or at some fixed interval, whichever is earlier.

(vii) Using better arrangement for exchangeable components.

(viii) Using large safety factors or a product improvement management program.

To improve the system reliability, implementation of the above steps will normally result in

the consumption of resources. Hence, a balance between the system reliability of a system and

resource consumption is an important task.

Generally to determine the reliability factor of a system, the system is broken down to sub

systems and elements whose individual reliability factors can be estimated or determined. Depending

on the manner in which these subsystems and elements are connected to constituted the given system.

The combinatorial rules are applied to obtain the system reliability. Hence the basic steps are as

follows:

(i) First identify the elements and subsystems which constitute the given system and whose
individual reliability factors can be estimated.
(ii) Next draw a block diagram or a circuit diagram to represent the logical manner or configuration
in which these units are connected to form the system.
(iii) The condition for the successful operation of the system is then determined i.e., it may be
decided as to how the units should function.
(iv) Finally, the combinatorial rules of probability theory i.e. addition, multiplication and their
combinations are applied to arrival at the system reliability factor.

11
Reliability of series system

Series system
In this system, a large number of components of the system are connected in series which
means that if any one of the components fails, the system fails. In other words, if the system is
operated, each component connected in series should successfully be operated. The system
comprising of n – components in series is represented as

1 2 3 n

Fig. 5
Let the successful operation of these individual units be represented by X1, X2, …., Xn. and their
respective probabilities by p(X1), p(X2), …., p(Xn). For the successful operation of the system, it is
necessary that all n units function satisfactorily.
Hence if the units are not independent one another then the system reliability is
p(S) = p(X1 X2X3 ….. Xn )
= p(X1) p(X2/ X1) p(X3/X1X2)………p(Xn/X1 X2 ….. Xn-1)

If the successful operation of each unit is independent of the successful operation of the remaining
units then p(S) = p(X1) p(X2) …..p(Xn).
If Ri is the reliability of the i-th components in series in the system and Rs is the reliability of
n
the system having n components in series then Rs = R1 R2 …. Rn =  Ri
i 1
If R1=R2=R3= … =Rn =R (say) then
Rs= Rn = (1-Q)n where Q is the probability of failure of each unit.
When each component has an exponential time to failure density,
 1t
then RS = R1 R2 …..Rn = e e t .... e   t
2 n

n
  i t n
e where s   i
 ( 1  2  n ) t  s t
=e =e i 1

i 1

The mean time between failures for the system having n-components in series is given by

ms   Rs (t )dt
0

  e  ( 1  2  n ) t dt
0

1 1
 
1  2  ...  n s
 ( n ) t 1
if 1  2  ...  n   (say) then Rs (t )  e and ms 
n
12
If n=1, then Rs =e-t and m= 1/.

Example 4 :An electronic circuit consists of 5 silicon transistor , 3 silicon diodes ,10 composite
resistor and 2 ceramic capacitor in series configuration. The hourly failure rate of each component is
for transistor : t = 4  10-5

for diode : d = 3  10-5

for resistor : r = 2  10-4

for capacitor : c =2  10-4

Calculate the reliability of the circuit for 10 hours when the components follow exponential
distribution.
Solution :
Components Nos. (Ni) i((per hour ) Nii
Transistor 5 4 10-5 20 10-5
Diode 3 3  10-5 9  10-5
Resister 10 2 10-4 20 10-4
Capacitor 2 2  10-4 4 10-4

 Ni i  0.00269

Let λs and Rs be the failure rate and reliability of the system.

As each component follows exponential time to failure density, then
λs =  i =5λt +3λd +10λr+2λc =2010-5+ 910-5 +2010-4 +410-4

=0.00269 per hour

 st -0.00269t
 Rs = e =e
-0.00269  10
The estimated reliability of the circuit for ten hours is Rs(10) = e = 0.9735.
This means that the circuit is expected to operate on an average without failure 9735 times and would
fail (10000-9735) i.e., 265 times out of 10,000 operations of 10 hours each.
In this case, MTBF is m= 1/λ =1/0.00269 = 371.75 hours.
This means that the circuit is expected to operate without failure for 372 hours.

Example 5 :The system connected in series consists of three independent parts A, B and C which
have MTBF of 100, 400 and 800 hours respectively. Find the MTBF and reliability of the systems for
30 hours. How much MTBF of the parts A has to be increased to get and improvement of MTBF of the
system by 30%.

13
 A B C

Fig. 6
Solution :
Here, MTBF of the part A = m1 =100 hours
MTBF of the part B = m2 =400 hours
MTBF of the part C = m3 =800 hours
 1 = failure rate of A =1/m1 = 1/100 failure hours
 2 = failure rate of B = 1/m2 = 1/400 failure hours
 3 = failure rate of C =1/m3 = 1/800 failure hours
Hence failure rate of the system  s =  1+  2+  3
=1/100 + 1/400 +1/800
=11/800 failure/hour
Therefore, MTBF of the system =1/ S

=72.75 hours.

and reliability of the system = Rs(30) = es 30 =e-11/ 800  30 = e-33/ 80
Second part: MTBF of the system = 72.75 hors = ms (say)
Let ms be the new MTBF of the system.
800 30 800
 ms '  ms  30% of ms    = 1040/11
11 100 11
Again, failure rate of the system = s =1/ms’ = 11/1040 hours.
Let x hours will be the MTBF of parts A to improve the MTBF of the system by 30%.
Hence, 1/(100+x) +1/400 +1/800 = 11/1040
i.e., x = 46.5

6 Reliability of parallel system

In this system a large number of components of the system are connected in parallel which means that
the successful operation of the system depends on the satisfactory functioning of any one of the
components. A system comprising of n–components connected in parallel represented by Fig. 16.7.

14
n
 Fig. 7
The reliability of the system can be calculated considering the conditions for the system failure. Let X1

, X2,.., Xn represent the successful operation of units 1, 2, …, n respectively. Similarly, let X 1 ,

X 2 ,…, X n represent their unsuccessful operation respectively. If p(X1) is the probability of successful

operation of unit then p( X 1) is the probability of its failure.

Hence, p( X 1) = 1- p(X1).
As the system will fail if all n units of the system fail simultaneously. The probability of
failure of the system is given by
p( S ) = p( X 1 X 2 X 3 …, X n )

= p( X 1 ) p( X 2 / X 1 ) p( X 3 / X 1 X 2 ) … p( X n / X 1 X 2 … X n 1 )

As the failure of the units are independent of one another then p( S ) = p ( X 1 ) p( X 2 ) – p( X n ).

Hence, the probability of success of the system is given by
p(S) = 1- p( S )

= 1- p ( X 1 ) p( X 2 ) … p( X n )]
= 1-[{1-p(X1)}{1-p(X2)} … {1-p(Xn)}]
Hence the reliability of the system comprising of n components connected in parallel is given by
n
Rs(t) = 1-  Qi (t )
i 1

n
where qi(t) is the failure of the i-th component. =1- {1  R (t )}
i 1
i

If n units are identical and the unit failures are independent of one another then
n
Rs (t )  1   1  Ri (t ) 
i 1

 1  1  R (t )
n

If the failure rate of each unit is exponential time to failure distribution, then
n n
Rs (t )  1   1  Ri (t )  1   1  e it
i 1 i 1
 
where i is the failure rate of the i-th component.

 
n
For identical units RS(t)= 1- 1  e t

15
Mean Time between Failure (MTBF)
The mean time between failure of a system having two components connected in parallel can be
obtained by integrating the reliability function over the range of t from 0 to ,
 
i.e., ms   Rs (t )dt   1  1  R1 (t )1  R2 (t ) dt
0 0

 

    
  1  1  eit 1  e 2t  dt   e1t  e2t  e  (1 2 )t  dt
 
0 0

=1/1 + 1/2 - 1/(1+2)

If these two components are identical i.e., 1 = 2 = (say) then ms =1/ +1/ -1/2 = 1.5/. This
means that the mean time between failure of a parallel system consisting of two components of equal
failure rate is 1.5 times the MTBF of a single component.
The system reliability can be improved either by improving the design or by providing
redundancy of the system. The system with redundancy has a number of stand-by units which take
over if the other components of the system in fail. Thus a reserve stock of stand-by’s improves the
system reliability but at a higher cost. The compromise between the increase of cost due to stand-by
arrangement and the increase of reliability will help to decide the optimal redundancy of the system.

Example 6 : The failure rate of an electronic subsystem is 0.0005 failure/hour. If a operational period
of 500 hours with probability of success 0.95 is desired. What label of parallel redundancy is needed?
Solution : Given that  = 0.0005 failures/hour , RS. = reliability of the system = 0.95 .
Let there be n subsystems connected in parallel.
As the subsystems are connected in parallel, RS =1-(1-R)n where R is the reliability of the system.
- t
Again, R =e = e-.0005500 = e-0.25
-0.25 n
Therefore, 0.95 = 1-(1-e )
which implies n =log(0.05)/log(0.2214)  2
Hence the label of parallel redundancy is 2.

Example 7 : How many identical components each of which is 90% reliable over a period of 50 hours
be used to obtain a 99.99% parallel redundancy system over 50 hours. If we want to obtain the same
system reliability over a period of 100 hours, how many components should be added?
Solution : Given that R = reliability of each component = 90% = 0.9, Rs = reliability of the parallel
system = 99.99% = 0.999 and t = 50 hours.
Let there be n label of parallel redundancy.
 RS =1-(1-R)n  0.9999 = 1-(1-0.9)n  (0.1)n = .0001  n  4
Hence 4 identical components are required to be connected in parallel.
16
Second part :
As the reliability of the system per 100 hours is given, we have to evaluate the reliability of each
component per 100 hours.
Since the reliability of each component per 50 hours is 0.9,
0.9 =e-50 or 0.9 = 1-50 [neglecting higher powers of ]
  = 0.002.
Let R be the reliability of each component per 100 hours.
- t
R =e =e-0.002100 = e-0.2 = 0.81.
Rs = 1- (1-R')n where n is the label of redundancy .
=1- (1-0.81)n
 0.9999 = 1- (0.19)n  n  6
Hence for the second case, 6 - 4 i.e., 2 identical components are to be added.

n
1 1
Example 8 : Show that MTBF of the system of n identical units connected in parallel is 
 i i 1
where

 is the failure rate of each component.

Solution : Let the parallel redundant system be n units with failure rates 1 , 2 , …, n(say). Hence the
reliability of the system is given by
Rs = 1-(1-R1) (1-R2) … (1-Rn)

= 1- (1-e-1t )(1-e-2t ) ... (1-e-n t )

n
= 1   1  e it
i 1
 


Now, MTBF of the system =  Rs (t )dt   [1  (1  e1t )(1  e2t ) ... (1  e nt )]dt
0
0

  
   e1t  e2t  ...  ent  e(1 2 )t  e( 1  3 )t  ...  ...  dt
 
0
=(1/1 +1/2 + … +1/n) – {1/(1+2) + 1/(1+3)+ …}+ {(1/(1+2+3) + 1/(1+3+4) -…
1
}+ …+ (-1)n+1 n

i 1
i

For identical units, 1 =2 = … = n

n 1 n 1 1 1 n 1
In that case, MTBF =  n C2  C3  ...  (1) n  
 2 3 n  i 1 i

17
7 Mixed Configuration
Parallel series system RB
RA B
A
C

Fig. 8(a)

The above system can be represented by an equivalent system shown in Fig. 16.8(b).

A BC

Fig. 8(b)
As the units B and C are connected in parallel, these units may be replaced by a unit BC (say) with
reliability RBC where
RBC = 1- (1-RB) (1-RC)
and the changed configuration of the system will be as

A BC

Fig. 9
Now the units A and BC are connected in series. The reliability of the system is given by
Rs = RARBC = RA{1- (1-RB) (1-RC)}

Mixed configuration
R11 R21 R31
R12 R22 R32

R13 R23 R33

Fig. 10(a)

The above system has three subsystems connected in series in which each subsystem is parallel
system. Hence this system can be represented by an equivalent system shown in the following figure.

RS1 RS2 RS3

Fig. 10(b)
18
Let Rs1, Rs2 and Rs3 be the reliabilities of the subsystem.

Hence
RS1 = 1- (1-R11) (1-R12) (1-R13)
RS2 = 1- (1-R21) (1-R22) (1-R23)
RS3 = 1- (1-R31) (1-R32) (1-R33)
Therefore the reliability of the system will be Rs = RS1RS2RS3.
Series parallel system

R11 R12 R13 R14

R21 R22 R23 R24

R31 R32 R33 R34

Fig. 11(a)

The above system can be represented by an equivalent system shown in Fig. 16.11(b).

RS1

RS2

RS3

Fig. 11(b)
Here RS1 = R11 R12 R13R14 , RS2= R21 R22 R23 R24 and RS3 = R31 R32 R33 R34
Hence the reliability of the given system will be Rs= 1- (1-RS1) (1-RS2) (1-RS3).

Example 9 : A system consisting of 4 identical subsystems connected in parallel. Each subsystem

consists of 3 identical units connected in series. If for each unit the probability of each over a certain
period of time is 0.95 obtain the system of reliability.

19
 R R R The given system can be
represented by an equivalent system
shown in Fig. 12(b).

R R R

RSB
R R R

R R R
RSB

Fig. 12(a)

RSB

RSB

Fig. 12(b)
Here R = 0.95 (given)
Hence the reliability of each sub system is RSB = R3 = (0.95)3 = 0.857
Hence the system reliability (RS) = 1- (1-RSB)4 = 1-(1-0.857)4 = 0.999582

K-out-of-n systems
A k-out-of-n system is an n-component system which functions when at least k of its n components
function. It is also called as k-out-of-n:G system.
When all the n components are identical and independent then system reliability can be calculated by
the formula as follows:
n
RS   nci R i (1  R) n i
i k

Where R is the component reliability.

Note-1 An n-component series system is a n-out-of-n:G system.
n
Proof Put k=n we get, RS   nci R i (1  R) n i = RS  n cn R n (1  R) n  n = R n
in

Note-2 A parallel system with n components is a 1-out-of-n: G system.

20
 n n
Proof: Put k=1 we get, RS   nci R i (1  R) n i = RS   nci R i (1  R) n i - n c0 R 0 (1  R) n 0
i 1 i 0

= ( R  1  R )n  (1  R ) n =1  (1  R ) n .

Stand-by Redundancy
In parallel configuration, all the components operate simultaneously and the experiences wear and tear
during the operation of the system. In the stand-by-system, there is a primary active element which is
operating and one or more components are standing-by to take over the operation one after another
when the first one fails. The operation of stand-by components is sequential i.e., each of the duplicate
elements becomes active and is energized only after the failure of the previous active elements. The
system will survive until the required time t if either of the following two condition holds.
(i) Primary element is successful operating up to time t.
(ii) Primary element fails at t1 and the active stand-by element takes over and survives from t1 to t.

Active

Stand-by system
Fig. 13
Assuming the failure density function of the primal active and the stand-by unit as fp(t) and fd(t)
respectively, we have
t 
(a) probability of primary element surviving upto the time t is 1   f p (t )dt   f p (t )dt
0 t

(b) probability of primary element failing at time t1 is fp(t1)

t  t1 
(c) probability of stand-by element working successfully from t1 to t = 1  0
f d (t )dt  
t t1
f d (t )dt

 t  
Hence the reliability of the system is given by R2(t) =  f p (t )dt  f p (t1 )   f d (t )dt ]dt1 .
t 0 t t1 

21
 f(t) t
Q(t) = f
0
d (t )dt = 1

R(t)

0 t 

Fig. 14
For the exponential case, fp(t) = p e pt , fd(t) = de-dt
where p and d be the failure rates of the primarily active and stand-by elements respectively.

  pt
t
  p t1
  
Hence the reliability of the system R2(t)    p e dt    p e   d e d dt dt1
 t

t 0 t t1 

pt
t
  p t1  d ( t t1 ) pt
t
( d   p ) t1  d  pt   p   t
=e   pe e dt1 =e   p  e  d t e dt1 =   e    e d
  
0 0  d p   d   p 
t
For p = d = , R2(t) = e t    e t1 dt1
0

 dt
 t  t
=e  e 1 = e t   te  t
0

Now the failure density function is exponential. Number of failures follows Poisson distribution i.e.,

 t 
i
 t
p(Number of failures = i) = p(i) = e
i!
From this, p(0) = e-t , p(1) = te-t and so on.
So, R2(t) = e-t +te-t = p(0) + p(1)
Similarly it can be shown that

 t    
2 e  t  t 
2 i

R3(t) = p(0) + p(1) +p(2) = e  t

  te  t
 e  t
 As R3 (t )   
2!  i 0 i! 

Hence, for a stand-by redundant system with n components of equal failure rate, the reliability of the

e  t   t   t 
i i
n 1 n 1
system is given by Rn (t )   e  t

i 0 i! i 0 i!
MTBF

Let us consider the MTBF of a stand-by redundant system with one active and one stand-by. Then
   
1  2
m2   R2 (t )dt   1   t  e  t dt =  e  t dt    te  t dt =  2 
0 0 0 0
  

22
Hence the mean time between the failures in the case of a stand-by redundant system consisting of
two components is twice that of the single component.

Stand-by system with imperfect sensing over device

Let sensing over device be not perfectly reliable and hence its probability of the failure must be
considered. Let the primary and secondary components be not similar and higher different failure rate
and the system consists of two dissimilar components placed in stand-by redundancy. The failure rates
of both components follow the exponential failure time distribution. The primary components have a
failure rate 1(say) and the stand-by component 2 (say).
Let rss be the reliability of the sensing and switching over device.
    1t  2t
The reliability of such a system is given by R(t )  e  1t  rss 
  
 e e  
 2 1

When 1 = 2 =, then R (t )  e  t  rss   te  t  = e t 1  rss ( t ) 

In general for a stand-by system of n components which have equal failure rate and when one unit is
operating actively and the rest (n-1) units are standing-by to take over the operation in succession then
MTBF of a system is mn = n/.

Example 10: An industrial process is controlled by a computer and two similar computers are
operated in stand-by redundancy such that if a computer fails another is instantaneously brought into
use in its place. The failure rate of each computer is given by  = 0.01 failure/hour. Compare the
improvement in reliability over a single computer when one and then two computers are used in a
stand-by. The operating period is 100 hours and the switch is considered to be perfect.
(i) In the first case, only one computer is in the system.
Hence the reliability of the system having
single computer is given by

R1(t) = e-t = e-.01100 =1/e = 0.37

Fig. 15

23
(ii) For the system having one computer along with another one as stand-by, the reliability of the
system is given by
e   t (  t )i
2 1
R2(t) = 
i 0 i!

 e  t  ( t )e  t = 0.37+0.011000.37 = 0.37 + 0.37 =0.74

As R2(t) = 2R1(t), hence there is 100% improvement in reliability by using one stand-by computer.
(iii) Again, for system having one computer along with two computer as stand-by, the reliability of the
system is given by

e  t  t  e  t .   t 
i 2
31
0.37 1
R3 (t )  
i!
e  t

 t e  t
 2!
 0.74 
2
=0.74 +0.185 =0.925
i 0

Now, R3(t)/R1(t) = 0.925/0.37 = 2.5

Hence by using two stand-by computers, reliability is improved by 2.5 times i.e., there is a 150%
improvement in reliability over the single computer.

Example 11 : An electronic device has a failure rate of 500 failures per 106 hours. One identical
stand-by unit is added to increase the reliability of the basic device. The operating time is 1000 hours.
The failure rate of the sensing and switching element is 0.97. What is the system reliability? What will
be the system reliability if the sensing and switching element is 100% reliable?

Fig. 16

Solution :
Here, the failure rate  = 500 failures /106 hours
= 500/106 failure per hour
= 0.0005 failure per hours,
and the operating time t = 1000 hours.
Hence t = 0.0005  1000 = 0.5
Given that the failure rate of the sensing and switching element is 0.97.
Hence the reliability of sensing and switching element is rss = 1-0.97 = 0.03.
So, when the sensing and switching element is not perfect, the system reliability is given by
R2(t) = e-t[1+ rss (t)] = e-0.5(1+0.015) =0.6156
When the sensing and switching element is perfect, the system reliability is given by

24
 R2(t) = e-t(1+t) = e-0.5(1+0.5) = 0.909

Reliability optimization
Reliability optimization appeared in the late 1940s and was first applied to communication and
transportation system. Much of the early work was confined to an analysis of certain performance aspects of
system. The goal of the reliability engineer is to find the best way to increase the system reliability. As a system
has grown more complex, the consequences their unreliable behaviours have become severe in terms of cost,
effort and so on. The interest in accessing the system reliability and the need to improve the reliability of
products and system has become more and more important.
The objective of the redundancy allocation problem is to find the number of redundant components that
maximize the system reliability under several resource constraints. This problem is one of the most popular
ones in reliability optimization since 1950s because of its potentiality for broad applications. System reliability
can easily be enhanced by improving the reliability of unreliable components and/or by adding redundancies on
the components. Improving the component reliability has been generally preferred over by adding redundancy
in industry, because, in many cases, redundancy is difficult to add to real systems due to technical limitations
and relatively large quantities of resources, such as weight, volume and cost that are required.
Network reliability design problems have attracted many researchers, such as network designers,
network analysts and network administrators, in order to share expensive hardware and software resources and
provide the access of main systems from different locations. The problems have many applications in the areas
of telecommunications and computer networking and related domains in the electrical, gas and sewer networks.
During the designing network system, one of the important steps is to find the best layout of components to
optimize some performance criteria, such as cost, transmission delay or reliability. The optimal design problem
can be formulated as a combinatorial problem.
Recently developed advanced technologies such as semiconductor integrated circuit and Nano
technology, however, have revived the importance of the redundancy strategy. The current downscaling trend
in the semiconductor manufacturing has caused many inevitable defects and subsequent faults in integrated
circuits. It is widely accepted that there are certain limitations on enhancing reliability or yield in
semiconductor manufacturing by developing relevant physical technologies. Hence, various fault-tolerant and
self-repairable techniques have been studied. These approaches are mainly based on adding redundancies on
components and controlling the usage of the redundancies. In fact, most memory integrated circuits and VLSI,
which includes internal memory blocks, currently use a hierarchical redundancy scheme to increase the yield
and reliability of the chip.

25
Different types of models in reliability optimization
Allocation of component reliabilities
(a) Continuous component reliabilities
System reliability can be improved by selecting the component reliabilities together with resource
constraints. Therefore the problem of maximizing the system reliability and the selection of
component reliabilities subject to the resource constraints is of the form as follows:
Maximize RS  f ( R1, R2 ,..., Rn )

subject to gi ( R1 , R2 ,..., Rn )  bi for i  1, 2,..., m .

where r jl  r j  r ju for j  1, 2,..., n .

The above problem is nonlinear type problem and also called reliability allocation problem.

(b) Discrete and continuous component reliabilities

Let, there are k numbers discrete choice of component reliability and and n  k numbers continuous
choice of component reliability.
Therefore the problem of maximizing the system reliability and the selection of component
reliabilities subject to the resource constraints is of the form as follows:
Maximize RS  f ( R1 ( x1 ), R2 ( x2 ),..., Rk ( xk ), Rk 1 ,..., Rn )

subject to gi ( R1 ( x1 ), R2 ( x2 ),..., Rk ( xk ), Rk 1,..., Rn )  bi for i  1, 2,..., m .

where r jl  r j  r ju for j  1, 2,..., n

and x j (j  1, 2,..., k ) being an integer.

This type of problem can be treated as nonlinear mixed integer programming problem.

Redundancy allocation
The system reliability can be improved through the selection of redundancy levels at stages with
respect to resource constraints.
Therefore the problem of maximizing the system reliability and the selection of component
reliabilities subject to the resource constraints is of the form as follows:
Maximize RS  f ( x1, x2 ,..., xn )

subject to gi ( x1, x2 ,..., xn )  bi for i  1, 2,..., m .

where r jl  r j  r ju for j  1, 2,..., n .

and x j being an integer.

The above problem is called redundancy allocation problem.

26
Reliability-redundancy allocation
The system reliability can be improved by selecting of component reliabilities as well as redundancy
level at stages subject to the resource constraints. Therefore the problem of maximizing the system
reliability and the selection of component reliabilities subject to the resource constraints is of the form
as follows:
Maximize RS  f ( x1, x2 ,..., xn ; R1, R2 ,..., Rn )

subject to gi ( x1, x2 ,..., xn ; R1, R2 ,..., Rn )  bi for i  1, 2,..., m .

where r jl  r j  r ju for j  1, 2,..., n .

and  j  x j   j for j  1, 2,..., n .

This problem is called the reliability-redundancy allocation problem.

Redundancy allocation for cost minimization

If we take cost function as a objective function of some redundancy allocation problems in reliability
system then the problem is called redundancy allocation for cost minimization.
Therefore the problem of minimizing the system cost and selection of component reliabilities subject
to the resource constraints is of the form as follows:
n
Minimize CS   c j (x j )
j 1

subject to gi ( x1, x2 ,..., xn )  bi for i  1, 2,..., m .

and  j  x j   j for j  1, 2,..., n .

Problems

1. What is MTBF? The failure rate of an electronic sub-system is 0.0005 failures/hour. If an

operational period of 500 hours with probability of success p=0.95 is desired, what level of
parallel redundancy is needed?
t
2. Show that R(t) = exp [   (t )dt ] where R(t) is the reliability function and (t) represents the
0

failure rate.
3. In a system, there are n number of components connected in parallel with reliability Ri(t),
i=1,2, …,n. Find the reliability of the system. If R1(t) = R2(t) = ….= Rn(t) = e-t then what will
be the expression of system reliability?

27
4. How many identical components each of which is 90% reliable over a period of 50 hours be
used to obtain a 99.99% parallel redundancy system over 100 hours.
5. Show that the reliability of an item can be expressed as an exponential function i.e.,
t
exp [   (t )dt ] .
0

6. An industrial process is controlled by a computer and two similar components are operated in
stand-by redundancy, such that, if a computer fails, another is instantaneously brought into use
in its place. The failure rate of each computer is given by =0.01 failure/hour. Compare the
improvement in reliability over a single computer when one and then two computers are stand-
by. The operating period is 100hours. and the switch system is 100% perfect.

7. A system consists of 5 identical and independent units with one unit operating and 4 units
stand-by. One of the stand-by units takes over when any operating unit fails. Assuming that the
switching device is perfect. Obtain the system reliability for a period of 100 hours if each of
the 5 units has a failure rate of 1000 failures/ 106 hours.
8. Obtain reliability over 100 hours period of a system consisting of two subsystem A and B
connected in parallel where A consist of 4 identical components in series and B consist of
identical components in parallel. Each component has a reliability 0.90 over a period of 100
hours.
9. A system is connected in series of 500 transistors, 10500 resistors and 500 capacitors. Failure
rate of these components are as follows:
transistor : t = 0.7 10-7 per hour
resistor: r = 0.1 10-6 per hour
capacity : c =2  0.2 10-6 per hour
What is the failure rate of the system? What is the reliability system of 100 hours.

28