Chapter 8

Reliability and Maintainability Prediction

November, 2023
By : Behailu Mamo

1
 a) Reliability prediction

It is the process of calculating the anticipated system

reliability from assumed component failure rates.

Reliability calculation is an imprecise calculation, but

provides a quantitative measure of how close a

design comes to meeting design objectives and
permits comparisons to be made between different
design proposals.

2
 Reliability prediction is important for the
following reasons:
 It provides an early indication of a system’s potential
to meet the design reliability requirements.
 It enables an assessment of life cycle costs to be
carried out.
 It enables establishment of areas which contribute
major part system unreliability.
 It enables the achievement of a given availability.
 It provides prior knowledge as to the magnitude of
expected maintenance work load.
3
 For the purpose of analysis, a large and complex plant
is regarded as a hierarchy of parts ranked according to
their function and replaceability.
 The figure below shows how a plant can be divided to
its basic components.

4
Hierarchy of parts of a plant

Performing the overall

PLANT function

-Performing major
UNITS plant functions

-Non-replaceable
ITEMS -Replaceable complex
-Replaceable simple

-Individual parts
COMPONENTS -Simple replaceable

5
Block Diagram Analysis
 Block diagrams are logic diagrams which show the functional
relationship among the system elements.

 To be able to predict reliability of a system, the system is

represented as a number of functional blocks that are
interconnected according to effect.

 The interconnection of blocks can be one of the possible three

types: Series, Parallel, Composite

 However, there are cases when these interconnections may not

represent a system.
6
 The approach to block diagram analysis requires
establishment of failure criteria. Moreover, in
defining the blocks, the following rules should be
kept in mind among others.
 Each block should represent the maximum number of
components in order to simplify the diagram.
 The function of each block should be easily identified.
 Blocks should be mutually independent in that failure in
one block should not affect the probability of failure in
other blocks.
 There should be only one environment within a block.

7
Block interconnections
a) Series-connected components
For such a system to work all components must work. For a
system of two blocks failure of either block prevents operation
of the system.

A B

Reliability of a series-connected system is

Rsy t   R1 A
t  R2 t  Rn t 

8
Reliability of a component is

    
Ri t  exp   i t dt 
t

 0   
System reliability is
 n t   t n 
Rsy t   exp    i t dt   exp    i t dt 
 i 1 0   0 i 1 

For constant failure rate λ, system reliability is

Rsy t   exp 1  2  n t 

9
 For identical-unit series system

1.00

R sy
n increasing

0.0
1.00 0.0
R unit

1.00
10
For constant failure rates, system MTBF is

 n 
MTBF   Rt dt   exp   i t  dt 
t t 1
n
 i 1 

0 0
i
i 1

11
Example

12
Solution

13
Example

14
Solution

15
b) Parallel-connected components
For such a system,
A system failure
B constitutes failure of all
components. The system of two blocks fails if both
blocks fail.
A

Parallel connections can be in:

 full active redundancy,
B
 partial active redundancy, or
A
 standby.
C
16
 I.All systems activated (full active redundancy)
 In this type of parallel connection all components
are functioning simultaneously.

 For the system of two subsystems A and B, there are

three possible ways in which the system is able to
operate:
 both subsystems A and B are working;
 subsystem A is working, B has failed; or
 subsystem A has failed, B is working

 The system fails if only both subsystems fail.

17
 Assuming the failure probabilities of components A and B to
be FA t  and FB t  , respectively, the probability that the
system fails is given by

Fsy t   FA t  FB t 
 The survival probability is then

Rsy t   1  Fsy t 
If survival probabilities of the components are RA t  and
RB t  , the system reliability is given by

Rsy t   RA t   RB t   RA t  RB t 

18
For n redundant units

Rsy  1  11.00
 R1 t  1  R2 t0.0
1  Rn t 
0.0

R unit

1.00

R sy n increasing

0.0
1.00 0.0
R unit
19
For full active redundancy and constant failure rate λ for
each component, system MTBF is

0
t

0

MTBF   Rsy t dt   1  1  e   dt
 t n

n
1 1
 
 i 1 i

20
Example

21
Solution

22
ii) Partial Active Redundancy
Partial active redundancy is also known as r-out-of-m
unit network. At least r units out of m active units must
function successfully for the system success. The system
reliability is given by

 m   t i
  
m
Rsy     e  t m i
1 e
i r  r 

m m!
  
 r  i!m  i !

23
The MTBF of partially active redundant systems, with
constant failure rate λ, is given by

 m  m   t i 
MTBF      e
t
 
1 e  t

m i
 dt
 ir  r  
0

1 m 1
 
 i r i

24
Example

25
Solution

26
iii) Parallel Redundant System in Standby
 In the standby configuration one or more subsystems are ready to

take over operation upon failure of the basic unit.

 The standby units are not operative until a failure-sensing device

senses the failure of the basic unit and turns on and connects the
standby unit.

B
27
 In parallel systems in standby, the following assumptions are

initially made.
 The means of sensing that failure has occurred and for
switching from the defective to the standby is assumed to
be failure free.
 The standby subsystems are assumed to have identical,
constant failure rates to the main subsystem.
 The standby subsystems are assumed not to fail while in the
idle state.
 Defective subsystems are assumed to remain so, No repair is

28
effected until the system has failed.
 The Poisson distribution is used to determine the reliability of

stand by systems since they display the constant λt characteristic

of the Poisson process.
 t
 probability of no failure is e
 t
 probability of one failure is given by t  e
 probability of two failures is
t  2
e  t

2!
For a system with n standby units, system reliability is


Rsy  1  t  
 t
2

 t    t
n 1
e
 2 !  n  1 ! 
 29
 Assuming that the switching process does not affect the basic
unit, for one standby unit, the system reliability is given by

Rsy  e  t
 Rsw  t   e  t

 where Rsw is reliability of the fault sensing and switching device.

 For a system with n-1 standby units, system reliability is

Rsy  e  t

 Rsw   t  
 t
2

 t    t
n 1
e
 2 !  n  1 ! 


30
 MTBF of a system with n-1 standby units is

MTBF   Rt dt

t


t
  1  t 
t 

t  
2 n

 dt
0
 2! n  1!
 For a constant failure rate λ of all units, the reliability is

n
MTBF 


31
Example

32
Solution

33
c) Composite connection
A
 This is a combination of series and parallel connections. For the

system shown, the system fails if block A fails or if both blocks B

B is not sufficient to cause system
and C fail. Failure of B or C alone
failure.

B
A

34
 d) Neither Series Nor Parallel Connections
 These are cases in which the system configuration cannot be

reduced to simple series and parallel models, nor any

combination of these.

A C B

A' B'

35
 For such situations where uncertainty exists to the

configuration of the system, Baye’s theorem can be

helpful to determine the system reliability. The use of
Baye’s theorem is demonstrated as follows.
 Consider the block diagram shown above. The connection of
the blocks in this case is neither parallel nor series.
 To ensure the reliability of the system, i.e. for the system to
operate successfully, at least one of the paths AA’ or BB’
should function.
 However, since neither A nor B is fully reliable, a third element
C is added to supply either A’ or B’. Hence four paths are
possible:
36 AA’, BB’, CA’ or CB’
System reliability is determined as follows:

 System success requires that at least one of the paths AA’, BB’,

CA’ or CB’ be successful.

With C working successfully,

Rsy Cs  1  1  RA' 1  RB' RC

With C unsuccessful,

Rsy Cus  1  1  RA RA' 1  RB RB' 1  RC 

37
 Hence, system reliability is given by the sum of the above

reliability terms

Rsy  Rsy Cs  Rsy Cus

 Substituting and simplifying yields

Rsy  1  RC 1  RA' 1  RB '  

1  RC 1  RA RA' 1  RB RB ' 

38
 Thank you
If you have any questions ???

39
b) Maintainability prediction
 In assigning any maintenance workload, it is important to
specify the time during which the job has to be accomplished.

 Doing this will help optimize the resource allocation and will
help eliminating idle man-hour, idle machine time and so on.

 Hence, maintenance engineers should engage themselves in

determining the time which a specific maintenance job is
going to take.

 This is done by determining the maximum repair time, the

median repair time and the maintainability probability.

 To this end determining the maintainability function plays an

40
important role.
I. Maximum Repair Time (case of constant M)

 Maximum repair time is the maintenance (restoration,

replacement) time which should not be exceeded or
exceeded only with a small probability.

 Determination of maximum repair time depends on

the probability of repair wanted.

Example: For a maintainability probability of 0.9,

t max  2.3026 MTTR

41
 The maintainability probability function for maximum
time is
 t M max 
 
 
M t M max  1  e  MTTR 


In general,
 t M max 
 
 
M t M max  1    1  e  MTTR 


This yields

t M max
ln   
MTTR
42
Given 
M t M max  and knowing the MTTR of an
equipment, the maximum time required for a
maintenance task is

tM max  MTTR  ln 

where

  1  M t M max

43
Case of  t  not constant

For such cases, the time-to-restore Probability density

function(pdf) is
t
   t dt
g t    t  e 0
And the maintainability probability is
t
  t dt

M t     t  e 0
t
dt
0

44
Let

   t dt  v    t dt  dv
t

The maintainability probability function is

M t    e dv  e |
t
v v t
0
0

Upon simplification,
t
   t dt
M t   1  e 0
45
Given M t  and g t  , the repair rate can be

determined from

g t 
 t  
1  M t 

46
 For times-to-restore probability distribution is given by
Weibull pdf, 
 t  
 1 
M t1   1  e   

 The maximum repair time can be obtained from

logarithmic transformation
1
 t1   
 ln1  M t1    
  
 From which t1 can be obtained as

t1     ln1  M t1 

1

47
 Thank you
If you have any questions ???

48