Anda di halaman 1dari 38

Maintenance & Safety:

ENGG ZC242
BITS Pilani
Pilani Campus

Girish Kant
Department of Mechanical Engineering

BITS Pilani
Pilani Campus

Statistical Distribution in
Preventive Maintenance
CHAPTER 10, Lecture - 15

Learning Objectives

Normal Distribution
Weibull Distribution
Binomial Distribution
Poisson Distribution

ENGG ZC242, Maintenance & Safety,


Lecture-15, 30/10/2014

3
BITS Pilani, Pilani Campus

INTRODUCTION
Statistical distributions, particularly normal, gamma,
Weibull, binomial and Poisson distributions, play a vital
role in estimation during the quality control inspections
and in predictive maintenance.
Weibull distribution can be used to provide a close
estimate of life of parts which fail frequently like bulbs,
and to give information on child mortality rates, etc.
With the assistance of Weibull distribution, cost
optimization of parts in categories of periodic usage /
consumption / failure can be done. This is briefly dealt
with here while providing a case study for the
distribution
ENGG ZC242, Maintenance & Safety,
Lecture-15, 30/10/2014

4
BITS Pilani, Pilani Campus

TYPES OF DISTRIBUTIONS
Different types of statistical distributions are in use.
They are classified as:
(a) discrete, (b) continuous, and (e) miscellaneous.
For the purpose of application in the area of
maintenance engineering, normal and Weibull
distributions are used in continuous distributions.
In the case of discrete distributions binomial and
Poisson's distributions are widely used. The following
Tables 10.1-10.3 provide the list of distributions widely
used in the area of mathematics.
ENGG ZC242, Maintenance & Safety,
Lecture-15, 30/10/2014

5
BITS Pilani, Pilani Campus

Discrete type of statistical


distributions
1 .Bernoulli

11. Parabolic fractal

2. Binomial

12. Rademacher

3. Boltzmann
4. Compound Poisson

13. Poisson
14. Skellam uniform

5. Degenerate
6. Degree
7. Geometric
8. Hypergeometric

15. Yule-Simon
16. Zeta
17. Zipf

9. Logarithmic

10. Negative binomial


6
ENGG ZC242, Maintenance & Safety, Lecture-15, 30/10/2014

BITS Pilani, Pilani Campus

Continuous statistical
distributions
1 .Beta

21. Levy skew alpha-stable

2. Beta prime
3. Cauchy
4. Chi-square
5. Exponential
6. Exponential power
7. Fisher-Tippett
8. Generalized extreme value
9. Generalized hyperbolic
10. Gamma
11. Hotelling's T-square
12. Hyperbolic secant
13. Hyperexponential
14. Hypoexponential
15. Inverse Chi-square
16. Inverse Gamma
17. Kumaraswamy
18. Landau
19. Laplace

22. Logistic
23. Log-normal
24. Maxwell-Boltzmann
25. Normal (Gaussian)
26. Pareto
27. Pearson
28. Raised cosine
29. Rayleigh
30. Relativistic Breit-Wigner
31. Rice
32. Triangular
33. Type-1 Gumbel
34. Type-2 Gumbel
35. Uniform
36. Voigt
37. Von Mises
38. Weibull
39. Wigner semicircle

ENGG ZC242, Maintenance & Safety, Lecture-15, 30/10/2014

7
BITS Pilani, Pilani Campus

NORMAL DISTRIBUTION

Some of the engineering applications such as reduction


in area from the tension test yield strength and tensile
strength, and the bilateral tolerance variations in a
machined component have been found to follow the
normal curve to a suitable degree of approximations.

A typical bell-shaped curve of a normal distribution with


X varying from + infinity to - infinity is shown in Figure
10.1.

ENGG ZC242, Maintenance & Safety,


Lecture-15, 30/10/2014

8
BITS Pilani, Pilani Campus

NORMAL DISTRIBUTION

Normal distribution curve


ENGG ZC242, Maintenance & Safety, Lecture-15, 30/10/2014

9
BITS Pilani, Pilani Campus

NORMAL DISTRIBUTION
If the normal distribution is represented as standard
normal curve, then the plot will represent as follows:
Area under the standard normal curve facilitates
evaluation of probability of occurrence of any event or its
non-occurrence.

ENGG ZC242, Maintenance & Safety,


Lecture-15, 30/10/2014

10
BITS Pilani, Pilani Campus

NORMAL DISTRIBUTION

Standardized normal frequency curve


ENGG ZC242, Maintenance & Safety,
Lecture-15, 30/10/2014

11
BITS Pilani, Pilani Campus

NORMAL DISTRIBUTION

Table gives the area under the standard normal


variable z. This in turn provides information on
probability of occurrence in a particular
standard normal distribution.

ENGG ZC242, Maintenance & Safety,


Lecture-15, 30/10/2014

12
BITS Pilani, Pilani Campus

Areas under standardized


normal frequency curve

ENGG ZC242, Maintenance & Safety,


Lecture-15, 30/10/2014

13
BITS Pilani, Pilani Campus

NORMAL DISTRIBUTION
One specific advantage of normal distribution is that
when the data does not fit in the normal curve, it is
worthwhile to transform the data so that the same is
normalized, such as the following:

ENGG ZC242, Maintenance & Safety,


Lecture-15, 30/10/2014

14
BITS Pilani, Pilani Campus

DEFINITION OF WEIBULL
DISTRIBUTION

The Weibull distribution is one of the most widely used


probability distributions in the reliability engineering
discipline. Developed in 1937 by Dr. Waloddi Weibull, it
was first introduced to greater scale in 1951.

Weibull distribution has since become a standard in


reliability for modeling time-dependent failure data.

ENGG ZC242, Maintenance &


Safety, Lecture-15, 30/10/2014

15
BITS Pilani, Pilani Campus

DEFINITION OF WEIBULL
DISTRIBUTION
Derived from a simple power function, the Weibull
distribution can be used to model a variety of failure
characteristics such as infant mortality, random
failures, wear-out, and failure-free periods.
It is quite, proven when conducting failure rate
predictions and analyses with a small sample base,
and it can handle inadequacies in data.
The Weibull distribution can also be used to
determine the cost-effectiveness and maintenance
periods of reliability-centred maintenance activities.
ENGG ZC242, Maintenance & Safety,
Lecture-15, 30/10/2014

16
BITS Pilani, Pilani Campus

DESCRIPTION OF THE
WEIBULL FUNCTION
The Weibull distribution is often used to describe the
life-times of parts. These can be light bulbs, capacitors,
disk drives, ball bearings, etc.
When a number of parts are put on test, they do not all
fail at the same time and there is usually some spread
in the failure times. If thousands of parts are tested and
a histogram is made for the life-times, it would look like
the one shown below.

ENGG ZC242, Maintenance & Safety,


Lecture-15, 30/10/2014

17
BITS Pilani, Pilani Campus

WEIBULL DISTRIBUTION

Typical Weibull distribution where a = 300 and b = 10


ENGG ZC242, Maintenance & Safety, Lecture-15, 30/10/2014

18

BITS Pilani, Pilani Campus

WEIBULL DISTRIBUTION
If parts fail according to a Weibull distribution, the fitted
curve over the histogram is called the Weibull density
function. The same is given by:

ENGG ZC242, Maintenance & Safety,


Lecture-15, 30/10/2014

19
BITS Pilani, Pilani Campus

WEIBULL DISTRIBUTION
Integral of this function is the formula for probability
function, and the same is given below: It is a function of
a and b and is given by the probability that any single
part will fail at a particular time, t is

where a is the scale parameter, b is the shape


parameter, and F is the cumulative distribution function.
If a and b are known, then we could plug them into the
above formula and calculate F (the probability of
failure) at any time, t.
ENGG ZC242, Maintenance & Safety,
Lecture-15, 30/10/2014

20
BITS Pilani, Pilani Campus

The parameters a and b can


be estimated from the data.

Or

The parameters a and b are estimated from the data and the
straight-line equation explained earlier. The same is shown
as an example for life of ball bearing.

ENGG ZC242, Maintenance & Safety,


Lecture-15, 30/10/2014

21
BITS Pilani, Pilani Campus

WEIBULL DISTRIBUTION
Probability figures can be arrived from the ratio of
rank of failure to the sample size. Typical example in
a table is given below which provides failure
information, ranking and the probability.
The failure times are ranked from lowest to highest.
There are 10 failure times, and each one can be
assigned a rank of 1, 2, 3,... etc. The probability of
failure at a particular time t, F(t), can be roughly
estimated by the rank of the failure time divided by
the sample size. In this case it is chosen as 10.

ENGG ZC242, Maintenance & Safety,


Lecture-15, 30/10/2014

22
BITS Pilani, Pilani Campus

EXAMPLE USING WEIBULL


DISTRIBUTION
A pump has parts (like valves, bearings, wiring, etc.)
that wear out with time. As these parts wear out, they
have to be replaced. One strategy would be to wait for
the part to fail and then replace it. However, what if the
failure causes a considerable damage to the pump then
it is not prudent to wait until it fails.

ENGG ZC242, Maintenance & Safety,


Lecture-15, 30/10/2014

23
BITS Pilani, Pilani Campus

EXAMPLE USING WEIBULL


DISTRIBUTION
It is advisable to replace it before it fails so that no
damage will occur. On the other hand, you don't want
to replace the valve too early because it may have a
lot of life left in it. A compromise is necessary. A
procedure is explained herein under.
First thing is to have information on how the failure
times pile up. If we have a long history on this pump
and the data on failure times of the valves are noted.

ENGG ZC242, Maintenance & Safety,


Lecture-15, 30/10/2014

24
BITS Pilani, Pilani Campus

EXAMPLE USING WEIBULL


DISTRIBUTION
For example, some valves last 3 weeks, others last
10 weeks, etc. Suppose a plot of failure times in the
form of a histogram is made. A histogram is basically
a pile of data. Usually, most of the data will pile up in
the middle; say 8 weeks.

Sometimes, a valve will only last a couple of weeks


and other times it might last up to 20 weeks. From
the failure times, there are statistical methods we
can use to determine a formula describing its shape.
Let's assume that this formula comes from the
Weibull distribution.
ENGG ZC242, Maintenance & Safety,
Lecture-15, 30/10/2014

25
BITS Pilani, Pilani Campus

USING WEIBULL
DISTRIBUTION

Histogram which can be approximated to a Weibull function


ENGG ZC242, Maintenance & Safety,
Lecture-15, 30/10/2014

26
BITS Pilani, Pilani Campus

USING WEIBULL
DISTRIBUTION
The next step is to determine the cost of the valve. The
same has two components, one is the cost of the valve
itself represented by C1 and the other is the cost of not
replacing the valve in time represented by C2. We want
to pick a replacement time, T, which minimizes the longterm average cost per unit time. The formula for the cost
uses an integral and is given by:

ENGG ZC242, Maintenance & Safety,


Lecture-15, 30/10/2014

27
BITS Pilani, Pilani Campus

USING WEIBULL
DISTRIBUTION

Optimization of cost
ENGG ZC242, Maintenance & Safety,
Lecture-15, 30/10/2014

28
BITS Pilani, Pilani Campus

BINOMIAL DISTRIBUTION
Whenever random variables takes on discrete values,
normal distribution cannot be applied as it deals with only
continuous random variables. Such a situation occurs in
quality control wherein a part is either accepted or rejected.

ENGG ZC242, Maintenance & Safety,


Lecture-15, 30/10/2014

29
BITS Pilani, Pilani Campus

BINOMIAL DISTRIBUTION
Random variables follow binomial distribution when the
following characteristics occur:

There are n trials


The trials are independent
Each trial has only two possible outcomes
Probability remains constant from trial to trial, i.e. either
acceptance or rejection probability in the case of quality
control problem remains the same.

ENGG ZC242, Maintenance & Safety,


Lecture-15, 30/10/2014

30
BITS Pilani, Pilani Campus

BINOMIAL DISTRIBUTION
If overall probability of unsuccessful event be p and the
probability of success will be q = 1 - p; then the probability
of occurrence of any value r = 0, 1,2 .....,n (where n is the
number of trials),

Further, when samples are randomly drawn from finite


population without replacement so that the probability of
an event cannot be constant from trial to trial then it is
hypergeometric distribution.
ENGG ZC242, Maintenance & Safety,
Lecture-15, 30/10/2014

31
BITS Pilani, Pilani Campus

POISSON DISTRIBUTION
If there are isolated events, which occur in a continuum of
events, the same is described by Poisson distribution. In
this case the probability of occurrence of r such events is
given by,

ENGG ZC242, Maintenance & Safety,


Lecture-15, 30/10/2014

32
BITS Pilani, Pilani Campus

An example of breakdowns in a
shop will illustrate the Poisson
distribution:

In order to determine the probability of failure of 0,1,2 and 3


failures per day the following procedure is adopted:
Average breakdown per day

ENGG ZC242, Maintenance & Safety,


Lecture-15, 30/10/2014

33
BITS Pilani, Pilani Campus

Applying Poisson's equation

ENGG ZC242, Maintenance & Safety,


Lecture-15, 30/10/2014

34
BITS Pilani, Pilani Campus

Applying Poisson's equation

Plot of the probability mass function (Poisson's


distribution) for given average events occurring over a
period of time.
ENGG ZC242, Maintenance & Safety,
Lecture-15, 30/10/2014

35
BITS Pilani, Pilani Campus

Applying Poisson's equation


The other examples of Poisson's distribution can be stated
as follows:
The number of cars that pass through a certain point on a
road during a given period of time.
The number of spelling mistakes a secretary makes while
typing a single page.
The number of phone calls at a call centre per minute.
The number of times a web server is accessed per minute
The number of road kill found per unit length of road.
Number of mango trees that can be counted in a forest
having all kinds of trees.

ENGG ZC242, Maintenance & Safety,


Lecture-15, 30/10/2014

36
BITS Pilani, Pilani Campus

SUMMARY

Different types of statistical distributions such as


normal, gamma and Weibull are explained.
Enumeration of Weibull distribution in comparison to
normal distribution in a state of failure of machinery is
also explained.

Brief account of discrete distributions like binomial and


Poisson's are enumerated with reference to
maintenance engineering.

ENGG ZC242, Maintenance & Safety,


Lecture-15, 30/10/2014

37
BITS Pilani, Pilani Campus

Thanks

ENGG ZC242, Maintenance &


Safety, Lecture-15, 30/10/2014

38
BITS Pilani, Pilani Campus

Anda mungkin juga menyukai