MAINTENANCE MANAGEMENT

IM – 503

Lecture – 6: Probability
Distributions
Dr Muhammad Fahad
Associate Professor/Director Product Development Centre
Dept of Industrial & Manufacturing
NED University of Engineering & Technology
 Slide
1-2

Random Variable

 A numerical characteristic whose value depends on the

outcome of a random experiment is called a random
variable.
 RV can either be discrete or continuous

Countable Uncountable
Sequence of Real Numbers Interval of Real Numbers
 Slide
1-3

Random Variable

 A continuous random variable is a variable that can

assume any value on a continuum (can assume an
uncountable number of values)
 thicknessof an item
 time required to complete a task

 temperature of a solution

 height

 These can potentially take on any value, depending

only on the ability to measure accurately.
 Slide
1-4

Probability
 Probability Density Function (pdf)
 For a continuous random variable X, we define pdf f(x) of X
by using following conditions:

 f(x) ≥ 0, x  R
P

  f(x)  1 f(x)
 b
 P(a<x<b) =
 f(x)dx t

a
 Slide
1-5

Probability
 Cumulative Distribution Function (cdf)

P
t is time.
 f(y) is the probability density function.
f(y)
 F(t) is the cumulative distribution function

 As t tends to ∞, F(t) tends to unity.

a b t
 In maintenance studies, we are interested in the
probability of a failure occurring before some specified
time (or between a time interval (a to b).
 Slide
1-6

Probability
 Reliability Function

 R(t) is the reliability function.

 F(t) is the cumulative distribution function
 A function complimentary to the cdf
 Also referred to as survival function
 Indicates the probability the system will survive at least to
some specified time t
 Slide
1-7

Probability
 Hazard Function (Failure Rate)
f(t)
h(t) 
 f(t) is the pdf
R(t)
 R(t) is the reliability function.

 Probability that the system will fail in the next interval of

time given that it is good at the start of the interval.

 Expected Value

 E(t) is the expected or mean value (MTBF)

 Slide
1-8

Probability
 A maintenance department is keeping history record about
the failure pattern of a electronic components in common
use by the electrical section. The data is summarized in table
1 where time is in years .
Time 1 2 3 4 5 6 7 8 9 10 >10
No. of
22 16 12 10 8 7 5 4 4 3 9
Failures

 Estimate the failure pattern associated with this component.

 Slide
1-9

Probability
Time 1 2 3 4 5 6 7 8 9 10 >10
No. of
22 16 12 10 8 7 5 4 4 3 9
Failures
Cumulative
22 38 50 60 68 75 80 84 88 91 100
Number
Frequency
0.22 0.38 0.50 0.60 0.68 0.75 0.80 0.84 0.88 0.91 1.00
of failure

25
Frequency Histograms
20
No. of Failures

15

10

0
1 2 3 4 5 6 7 8 9 10 >10

Time (yrs)
 Slide
1-10

Probability
Time 1 2 3 4 5 6 7 8 9 10 >10
No. of
22 16 12 10 8 7 5 4 4 3 9
Failures
Cumulative
22 38 50 60 68 75 80 84 88 91 100
Number
Frequency
0.22 0.16 0.12 0.1 0.08 0.07 0.05 0.04 0.04 0.03 0.09
of failure

μ   x.p(x)
Probability Density Function

μ  4.44
Probability. of Failures

1 2 3 4 5 6 7 8 9 10 >10

Time (yrs)
 Slide
1-11

Probability Distributions

Probability
Distributions

Discrete Continuous
Probability Probability
Distributions Distributions

Binomial Normal

Poisson Exponential

Hypergeometric Weibull
 Slide
1-12

Normal Distribution
 Normal Distribution

1
f(t)  e  (1/2)[(t  μ )/σ ] 2

σ 2π
Where e = the mathematical constant approximated by 2.71828
π = the mathematical constant approximated by 3.14159
μ = the population mean
σ = the population standard deviation
t = any value of the continuous variable
 Slide
1-13

Normal Distribution

1
Probability Density
f(t)  e  (1/2)[(t  μ )/σ ] 2

σ 2π
Function

x
1
Cumulative F(t)   e  (1/2)[(t  μ )/σ ] 2
dt
Distribution
Function
 σ 2π

 t - μ
 P z 
 σ 
 Slide
1-14

Normal Distribution
Reliability Function R(t)  1  F(t)
 t - μ
 1  P z  
 σ 
f(t)  t -μ 
Hazard Function h(t)   P 
R(t)  σR(t) 

Expected Value = μ
 Slide
1-15

Normal Distribution
 Slide
1-16

Normal Distribution
 Slide
1-17

Normal Distribution
 Normal failure laws often apply to
 Relatively straightforward maintenance tasks and
repair actions (e.g., simple removal and replacement
tasks) which consistently require a fixed amount of
time to complete.
 Situations where lifetimes are the result of a sum of
many other random variables.
 Wear out processes, stress – strength related
problems
 Slide
1-18

Normal Distribution
 A component has normally distributed failure time with μ
= 20 and σ = 2. Find the reliability of the component at
18 time units.

Normal
Table
 Slide
1-19

Normal Distribution
 An average light bulb manufactured by the Acme
Corporation lasts 300 days with a standard deviation of
50 days. Assuming that bulb life is normally distributed,
what is the probability that an Acme light bulb will last at
most 365 days.
 Slide
1-20

Normal Distribution
 An average light bulb manufactured by the Acme
Corporation lasts 300 days with a standard deviation of
50 days. Assuming that bulb life is normally distributed,
what is the probability that an Acme light bulb will last at
most 365 days.
μ  300 Normal
σ  50 Table

 t - μ  365 - 300 
F(365)  P  z   P z 
 σ   50 

F(365)  0.9032  90%

 Slide
1-21

Normal Distribution
 The refractory lining of an oven wears out (must be
replaced) periodically. The wear-out distribution is
believed to be normal with a mean of 2700 hours and a
standard deviation of 250 hours. After how many hours
of operation should the oven be relined so the wear out
is virtually impossible?

 Schedule relining so that it occurs at a point in the

extreme left tail of the distribution, say at μ – 3σ

 Reline every 2700 – 3(250) = 1950 hours.

 Slide
1-22

Lognormal Distribution
 Lognormal Distribution

1
f(t)  e  (1/2)[(lnt  μ )/σ ] 2

σ.t. 2 π
Where e = the mathematical constant approximated by 2.71828
π = the mathematical constant approximated by 3.14159
μ = the population mean
σ = the population standard deviation
t = any value of the continuous variable
 Slide
1-23

Lognormal Distribution
1
f(t) 
Probability Density  (1/2)[(lnt  μ )/σ ] 2
Function e
σt 2 π

x
1
Cumulative
Distribution F(t)   e  (1/2)[(lnt  μ )/σ ] 2
dt
Function  σt 2 π
 lnt - μ 
 P z  
 σ 
 Slide
1-24

Lognormal Distribution

 lnt - μ 
R(t)  1  F(t)  1  P  z 
σ 
Reliability
Function

(lnt - μ)/σ 
Hazard Function h(t)  P  
 σtR(t) 

σ2
(μ 
e
)
2
Expected Value
 Slide
1-25

Lognormal Distribution
 Slide
1-26

Lognormal Distribution
 Slide
1-27

Lognormal Distribution
 Lognormal failure laws apply
 When the logarithm of a random variable follows
normal distribution.
 When the degradation in life time is proportional to
the previous amount of degradation.
 In situations such as corrosion, crack growth,
mechanical wear.
 Slide
1-28

Lognormal Distribution
 The failure time of a device follows a log normal
distribution with μ = 4 and σ = 1. Find the reliability of
the device at t = 100.
 Slide
1-29

Exponential Distribution
 Exponential Distribution

f(t)  λe  λt

Where e = the mathematical constant approximated by 2.71828

λ = failure rate = 1/mean time to failure
t = any value of the continuous variable
 Slide
1-30

Exponential Distribution

Probability Density
Function
f(t)  λe  λt

Cumulative
Distribution F(t)  1 e  λt

Function

Reliability Function R(t)  e  λt

Hazard Function h(t)  λ

Expected Value = 1/ λ
 Slide
1-31

Exponential Distribution
 Slide
1-32

Exponential Distribution
 Slide
1-33

Exponential Distribution
 Exponential failure laws apply
 To maintenance tasks and maintenance actions
whose arrival rates are constant and completion
times are independent of previous maintenance
experience.
 To model occasions when there is no wear-out or
cumulative damage.
 Electronic components such as capacitors or
integrated circuits
 Random shocks (e.g. failure of a tire due to puncture
from a nail)
 Inappropriate to model mechanical components.
 Slide
1-34

Exponential Distribution

 Suppose that a component has useful life that is

satisfactorily modeled by an exponential distribution with
mean failure time = 1000 hr. What is the probability that
this component would fail before 2000 hours?