Section 8b Peussdistributions 2 Slides Compatibility Mode

An Introduction to Reliability and
Life Distributions
Dr Jane Marshall
Product Excellence using 6 Sigma
Module
PEUSS 2011/2012
Reliability and Life distributions
Page 1
Objectives of the session

Probability distribution functions
Life time distributions
Fitting Reliability distributions using Hazard
Plotting
Interpretation
PEUSS 2011/2012
Page 2
Data types of interest

Sample data from a population of items
For example:
100 ipods put on test, 12 fail, analyse the times to
failure
1000 aircraft engine controllers operating in-service,
collect all the times to failure data and analyse
Not only times but distance or cycles etc.
PEUSS 2011/2012
Page 3
Histogram
Histogram of hours to failure
35
120.00%
30
100.00%
25
Frequency
80.00%
20
60.00%
15
40.00%
10
20.00%
5
0
.00%
9
5
37
6.
23
75
3.
46
5
12
1.
69
5
8.
91
75
.8
45
11
5
25
.2
.6
73
00
13
16
e
or
M
Hours to failure
Frequency
PEUSS 2011/2012
Cumulative %
Page 4
Probability distribution
Hours to failure
35
Frequency
30
25
20
15
10
5
-500
0
-5 0
500
1000
1500
2000
2500
Hours to failure
The area under the curve is equal to 1

The area under the curve between two values is the
probability
PEUSS 2011/2012
Page 5
Failure Time distributions

PDF (Probability density function)
The CDF (Cumulative Distribution Function)
The CDF gives the probability that a unit will fail before time t
or alternatively the proportion of units in the population that
will fail before time t.
The Survival Function (sometimes known as

reliability function)
Complement of the CDF.
The Hazard Function

Conditional probability of failing in the next small interval
given survival up to time t.
PEUSS 2011/2012
Page 6
Probability density Function:

PDF - Probability of falling between two values
Frequency
PDF, f(t) (%)
1.2
1
0.8
P(t1<t<t2)=
0.6
t2
t1
f(t) dt
0.4
0.2
0
1
Value
PEUSS 2011/2012
Page 7
Probability distributions
Hours to failure
Relative frequency
0.5
0.4
0.3
0.2
0.1
0
-500
-0.1 0
500
1000
1500
2000
2500
hours to failure
Probability of failure between 500 and 1000 hours is given by the area
PEUSS 2011/2012
Page 8
Standard Normal distribution
-1s
-2s
+1s
68.27%
-3s
+2s
+3s
95.45%
99.73%
PEUSS 2011/2012
Page 9
Cumulative distribution function

The CDF known as F(t)
Frequency
CDF, F(t) (%)
1
1.2
F( t) =
-f ( t ) dt
0.8
0.6
0.4
0.2
0
1
5t 5
Value
F(t) gives the

probability that a
measured value will fall
between - and t
Failure Function, F(t)

PEUSS 2011/2012
Page 10
Cumulative distribution
Cumulative probabilty
cumulative probability
1.2
1
0.8
0.6
0.4
0.2
0
-500
-0.2 0
500
1000
1500
2000
2500
hours to failure
The probability of failure before 500 hours is 0.8

or 80% will have failed by 500hrs
PEUSS 2011/2012
Page 11
Survival function
The survival function or reliability function R(t)
Frequency
(%)
R(t)
1.2
1
0.8
R(t) = 1 - F(t) and

F(t) = 1 - R(t)
0.6
0.4
0.2
0
1
5t 5
Value
PEUSS 2011/2012
Page 12
Survival Function
Survival Function
Probability of survival
1
0.8
0.6
0.4
0.2
0
-500
500
1000
1500
2000
2500
Hours to failure
The probability of surviving up to 500 hrs is 0.2

Or 20% have survived up to 500 hrs
PEUSS 2011/2012
Page 13
Hazard function
1 - F (t )
R (t )
Figure shows
increasing hazard
function
Frequency
(%)
h(t)
The Hazard function is defined as probability of

failure in next time interval given survival to time
Reliability Function R(t)
1
t
1.2
1
h(t) = f ( t ) = f ( t )
0.8
0.6
0.4
0.2
0
1
2Hazard
3 3 Function
4 5 5h(t)6
Value
PEUSS 2011/2012
Page 14
Hazard function
Bath-tub curve
Useful Life
Time
Infant
Mortality
PEUSS 2011/2012
Wear Out
Page 15
Probability distributions
Exponential distribution
Weibull distribution
Normal distribution
Lognormal distribution
PEUSS 2011/2012
Page 16
Simplest of all life models

One parameter,
PDF, f(t) = e- t
CDF, F(t) = 1- e- t and R(t) = e- t
Hazard function, h(t) = i.e. constant
MTBF = 1/ and failure rate =
1/ is the 63rd percentile i.e. time at which 63%
of population will have failed
PEUSS 2011/2012
Page 17
Probability
1.0
Hazard Function
0.160
0.0
0.155
10
20
30
40
Survival Function
50
Rate
0.5
0.150
0
PEUSS 2011/2012
10
20
30
40
50
Page 18
Failure rate - example
10 components of a particular type in each PCB

5 PCBS in each unit
200 units in the field
Total operating time to date for all units is 10,000 hours
There have been 30 confirmed failures of this component
The failure rate is given by:
30/5*200*10*10,000 = 0.000003 = 3 fpmh (failures per million hours)
The MTTF is 1/0.000003 = 333,333
PEUSS 2011/2012
Page 19
Example
100 units in the field

Total operating hours is 30,000
Number of confirmed failures is 60
MTBF = 30,000*100/60 = 50,000
Removal rate includes all units removed
regardless of whether they have failed
Use 200 removals
MTBR = 15000
PEUSS 2011/2012
Page 20
10
Most useful lifetime in reliability analysis
2 parameter Weibull
Shape parameter -
Scale parameter -
When < 1 decreasing hazard function

When > 1 increasing hazard function
When =1 constant hazard function
is the characteristic life, 63rd percentile
PEUSS 2011/2012
Page 21
PDF : f (t )
CDF : F (t ) 1 e
t

1
Re liability : R (t ) e
PEUSS 2011/2012
Page 22
11
t-1
When =1, h(t)= 1/ = therefore =1/

When >3.5 the distribution approximates to a
normal distribution
h(t) =
PEUSS 2011/2012
Page 23
Three parameter Weibull

A three parameter distribution can be used if
failures do not start at t=0, but after a finite
time . The parameter, is called the failurefree time or location parameter
F (t ) 1 e
PEUSS 2011/2012
(t )
24
12
Hazard function
Bath-tub curve and the Weibull

Useful Life
<1
=1
>1
Time
Infant
Mortality
PEUSS 2011/2012
Wear Out
Page 25
Normal Distribution
Not used as often in reliability work
Can represent severe wear-out mechanism
Rapidly Increasing hazard function
e.g.s, filament bulbs, IC wire bonds
Location parameter, m , is the mean

Scale parameter, , is the standard deviation
Lognormal more versatile, always positive
PEUSS 2011/2012
Page 26
13
Fitting parametric distributions
PEUSS 2011/2012
Page 27
Censoring
Repaired and non repaired
Probability plotting
Hazard plotting
PEUSS 2011/2012
Page 28
14
Censoring structures
Complete data
Single censored
Units started together and data analysed before all units have
failed
Right, interval and left
Time censored
Censoring time is fixed
Failure censored
Number of failures is fixed
Multiply censored
Different running times intermixed with failure times field data
PEUSS 2011/2012
Page 29
Complete Data
PEUSS 2011/2012
Page 30
15
Right Censored data
PEUSS 2011/2012
Page 31
Interval Censored
PEUSS 2011/2012
Page 32
16
Left censored
PEUSS 2011/2012
Page 33
Repaired and non-repaired data

Non-repaired data when only one failure can
occur and interested in time to failure
Repaired data when interested in the pattern of

times between failures
PEUSS 2011/2012
Page 34
17
Probability plotting
PEUSS 2011/2012
Page 35
Areas to be covered
Introduction to probability plotting

Assumptions
How to do a Weibull plot
Estimating the parameters
Testing assumptions
Examples
PEUSS 2011/2012
Page 36
18
What is probability plotting?

Graphical estimation method
Based on cumulative distribution function CDF or F(t)
Probability papers for parametric distributions, e.g.
Weibull
Axis is transformed so that the true CDF plots as a
straight line
If plotted data fits a straight line then the data fits the
appropriate distribution
Parameter estimation
PEUSS 2011/2012
Page 37
Assumptions
Data must be independently identically
distributed (iid)
No causal relationship between data items
No trend in the time between failures
All having the same distribution
Non-repaired items
Repaired items with no trend in the time
between failures
Time to first failure of repaired items
PEUSS 2011/2012
Page 38
19
Example of test for trend

Machine H fails at the following running times
(hours):
15, 42, 74, 117, 168, 233, and 410
Machine S fails at the following running times

(hours):
177, 242, 293, 336, 368, 395, and 410
PEUSS 2011/2012
Page 39
Trend Analysis
machine S running times to failure
8
7
6
5
4
3
2
1
0
order number
order number
machine H running times to failure
100
200
300
400
500
8
7
6
5
4
3
2
1
0
0
machine time to failure
100
200
300
400
500
machine time to failure
This system is getting better with

time, the failure times are getting
further and further apart
This system is getting worse with time,

the failure times are getting closer and
closer together.
In neither case can Weibull analysis be used as there is

trend in the data.
PEUSS 2011/2012
Page 40
20
Making a Weibull plot
PEUSS 2011/2012
Page 41
Rank the data

Probability graph papers are based on plots of
the variable against cumulative probability
For n< 50 the cumulative percentage probability
is estimated using median ranks tables
For n< 100 use benards approximation for the
median rank ri
ri = i - 0.3
n+0.4
Where i is the ith order value and n is the sample size
PEUSS 2011/2012
Page 42
21
Example
Failure number (i)
Ranked hrs at failure (ti)
Median Rank from tables

Cumulative % Failed at ti - F(t)
300
6.7
410
16.2
500
25.9
600
35.5
660
45.2
750
54.8
825
64.5
900
74.1
1050
83.8
10
1200
93.3
PEUSS 2011/2012
Page 43
Plot times on x-axis

Plot CDF on y-axis
Fit line through the data
Draw perpendicular line from
estimation point to the
fitted line.
Read off the estimate of
is the value given on
from the intersection line
PEUSS 2011/2012
Page 44
22
Interpreting the plot

If the data produced a straight line then:
The data can be modelled by the Weibull distribution.
If <1 then data shows a decreasing hazard function

e.g. Infant mortality, weak components, low quality
If =1 then data shows a constant hazard function

e.g. useful life of product
If >1 then data shows a increasing hazard function

e.g. wear-out, product reaching end of life
is the value in time by which 63.2% of all failures will

have occurred and is termed the characteristic life
PEUSS 2011/2012
Page 45
Hazard function
Bath-tub curve and the Weibull

Useful Life
<1
=1
>1
Time
Infant
Mortality
PEUSS 2011/2012
Wear Out
Page 46
23
Interpreting the plot

If the data did not produce a straight light then:
There may be an amount of failure-free time
This may appear concave when viewed from the bottom
right hand corner of the sheet
There may be more than one failure mode present

This may appear convex shape or cranked shape (also
known as dog-leg shape)
In this case the data needs to separated into failures
associated with each failure mode using expert judgement
and analysed separately
PEUSS 2011/2012
Page 47
Example: Poor fit due to 3 months

offset
Weibull Plot of Time-in-Service
Weibull Plot of Time in service (Months)
2-Parameter Weibull - 95% CI

Censoring Column in Censoring - ML Estimates
3-Parameter Weibull - 95% CI

Censoring Column in Censoring - ML Estimates
99
Table of Statistics
Shape
1.87010
Scale
57.6561
Mean
51.1897
10
10
5
3
2
5
3
2
0.01
0.1
1.0
10.0
Time in service (Months)
100.0
Table of Statistics
Shape
0.873095
Scale
297.337
Thres
2.9997
90
80
70
60
50
40
30
20
Pe r ce nt
Pe r c e n t
99
90
80
70
60
50
40
30
20
0.01
0.0001
0.0010
0.0100
0.1000
1.0000 10.0000 100.00001000.000010000.0000
Time in service (Months) - Threshold
The same data plotted with a three-Parameter Weibull distribution shows a good
fit with 3 months offset (location 2.99 months)
PEUSS 2011/2012
Page 48
24
Example of two failure modes
Weibull CDF
Mode 2 Beta
= 11.9
Mode 1 Beta
= 0.75
Time to failure ( hours)

PEUSS 2011/2012
Page 49
Adjusted rank for censored data
Rank
1
2
3
4
5
6
7
8
Time
10
30
45
49
82
90
96
100
PEUSS 2011/2012
Status
Suspension
Failure
Suspension
Failure
Failure
Failure
Failure
Suspension
Reverse
rank
8
7
6
5
4
3
2
1
Adjusted rank
Suspended...
[7 X 0
+(8+1)]/ (7+1) = 1,125
Suspended
[5 X 1,125 +(8+1)]/ (5+1) = 2,438
[4 X 2,438 +(8+1)]/ (4+1) = 3,750
[3 X 3,750 +(8+1)]/ (3+1) = 5,063
[2 X 5,063 +(8+1)]/ (2+1) = 6,375
Suspended...
Median
rank
9,8 %
25,5
41,1
56,7
72,3
%
%
%
%
Page 50
25
Weibull Analysis using

software tools
Number of software packages that can do Weibull
plotting (and other distributions), these include:
Minitab
Relex
WinSMITH
Reliasoft
Concentrate on getting good quality data, correct

assumptions and correct interpretation from the
software
PEUSS 2011/2012
Page 51
Hazard Plotting
PEUSS 2011/2012
Page 52
26
Contents
Assumptions
Estimating parameters
Using results for decision making
PEUSS 2011/2012
Page 53
Assumptions
Non-repaired items
Repaired items with no trend in the time
between failures
Time to first failure of repaired items
Individual failure modes from non-repaired items
Can deal with censored data
in particular multiply censored data
PEUSS 2011/2012
Page 54
27
Hazard Plotting
Cumulative hazard function
H(t)=
H(t)=
t
0
h(t) dt
t
0
f(t) /1-F(t) dt
H(t)= -ln[1-F(t)]
Relationship allows derivation of cumulative

hazard plotting paper
PEUSS 2011/2012
Page 55
Weibull Hazard Plotting

h(t) =
t-1 and H(t) = t
()
If H is the cumulative hazard value then

Log t = 1 log H + log
Weibull hazard paper is log-log paper

The slope is 1/ and when H=1, t=
PEUSS 2011/2012
Page 56
28
Hazard plotting procedure

Tabulate times in order and rank
Reverse rank
For each failure, calculate the hazard interval
hi = 1/ no of items remaining after previous
failure/censoring (i.e. 1/reverse rank)
For each failure, calculate the cumulative hazard

function H
n
H = h1 + h2 + . +
Plot the cumulative hazard against life value

PEUSS 2011/2012
Page 57
Example 1 vehicle shock

absorbers
Distance to failure for
Shock absorbers
F denotes failure
PEUSS 2011/2012
Distance (km)
6700 F
17520 F
6950
17540
7820
17890
8790
18450
9120 F
18960
9660
18980
9820
19410
11310
20100 F
11690
20100
11850
20150
11880
20320
12140
20900 F
12200 F 22700 F
12870
23490
13150 F 26510 F
13330
27410
13470
27490 F
14040
27890
14300 F 28100
Page 58
29
Example 1 vehicle shock

absorbers
Rank
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
Reverse
rank
38
37
36
35
34
33
32
31
30
29
28
27
26
25
24
23
22
21
20
Distance
(km)
6700 F
6950
7820
8790
9120 F
9660
9820
11310
11690
11850
11880
12140
12200 F
12870
13150 F
13330
13470
14040
14300 F
Hazard
(1/rank)
1/38
Cumulative
hazard
0.0263
1/34
0.0557
1/26
0.0942
1/24
0.1359
1/20
PEUSS 2011/2012
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
19
18
17
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
17520 F
17540
17890
18450
18960
18980
19410
20100 F
20100
20150
20320
20900 F
22700 F
23490
26510 F
27410
27490 F
27890
28100
1/19
0.2385
1/12
0.3218
1/8
1/7
0.4468
0.5896
1/5
0.7896
1/3
1.1229
0.1859
Page 59
Example 1
Plot the data on log 2 cycle paper x log 2 cycle
paper
Estimate Weibull shape parameter
Estimate Weibull scale parameter
Interpret results
PEUSS 2011/2012
Page 60
30
Example 1
distance
Cumulative hazard plot for shock absorbers on linear

paper
30000
25000
20000
15000
10000
5000
0
0
0.2
0.4
0.6
0.8
1.2
Cumulative hazard
PEUSS 2011/2012
Page 61
Example 1
Log cumulative hazard for shock absorbers
= 2.6
= 28500km
R2 = 0.98
log distance
100000
10000
1000
0.01
0.1
10
log hazard
PEUSS 2011/2012
Page 62
31
Example 1
Since looking at one known failure mode use the estimated
parameters to fit to the distribution
Survival plot for vehicle shock absorbers with

Beta =2.6 and Eta=29000km
1.2
1
0.8
0.6
0.4
0.2
0
0
5000
10000
15000
20000 25000 30000 35000 40000
kilometers
PEUSS 2011/2012
Page 63
Example 2:Hazard plot on

linear paper
Cumulative hazard plot for O ring failures
2000
h o u rs
1500
1000
500
0
0
cumulative hazard
PEUSS 2011/2012
Page 64
32
Example 2: Hazard plot on log

paper
log cumulative hazard for O ring failures
= 1.01
= 360hrs
R2 = 0.98
10000
log hours
1000
100
10
1
0.01
0.1
10
log cumulative hazard
PEUSS 2011/2012
Page 65
Example 2: Interpretation of
results
= 1.01 is approximately an exponential
distribution and constant failure rate
= 360 hrs = 1/ = Mean Time to Failure
Calculating the MTTF from the data gives:
Total hours/number of failures
26839/73 = 367 hrs
PEUSS 2011/2012
Page 66
33
Example 2: Survival function

R(t) - Survival Function for O ring failures
1.2
1
0.8
0.6
0.4
0.2
0
0
500
1000
1500
2000
Hours to failure
PEUSS 2011/2012
Page 67
Example 2: Failure Distribution

F(t) for O ring failures
Probability of failure
1.2
1
0.8
0.6
0.4
0.2
0
0
500
1000
1500
2000
Hours to failure
PEUSS 2011/2012
Page 68
34
Example 3 : pumps
pump no
Two dominant failure

modes
Impeller failure (I)
Motor failure (m)
age at failure
1
1180 m
6320 m
1030 i
120 m
2800 i
970 i
2150 i
700 m
640 i
10
1600 i
11
520 m
12
PEUSS 2011/2012
failure mode
1090 i
Page 69
Example 3: ignoring failure

modes
log cumulative hazard for all failures
10000
age
1000
100
10
1
0.01
0.1
10
cum hazard
PEUSS 2011/2012
Page 70
35
Example 3: impeller failure

log cumulative hazard plot for impeller failures
= 1.95
= 1900hrs
R2 = 0.93
10000
a ge a t fa ilur e
1000
100
10
1
0.1
10
cumulative hazard
PEUSS 2011/2012
Page 71
Example 3: motor failure

log cumulative hazard plot for motor failures
= 0.76
= 3647hrs
R2 = 0.978
a ge a t fa ilure
10000
1000
100
10
1
0.01
0.1
10
cumulative hazard
PEUSS 2011/2012
Page 72
36
Advantages of Cum Hazard

Plotting
It is much easier to calculate plotting positions
for multiply censored data using cum hazard
plotting techniques.
Linear graph paper can be used for exponential
data and log-log paper can be used for Weibull
data.
PEUSS 2011/2012
Page 73
Disadvantages of Cum
Hazard Plotting
It is less intuitively clear just what is being plotted.
Cum percent failed (i.e., probability plots) is meaningful and
the resulting straight-line fit can be used to read off times
when desired percents of the population will have failed.
Percent cumulative hazard increases beyond 100% and is
harder to interpret.
Normal cum hazard plotting techniques require exact

times of failure and running times.
With computer software for probability plotting, the
main advantage of cum hazard plotting goes away
PEUSS 2011/2012
Page 74
37
Summary
Important lifetime distributions
Failure distribution (CDF), Survival function R(t) and
the hazard function h(t)
Some parametric distributions

Exponential, Weibull and Normal
Weibull probability plotting

Distribution fitting using hazard plotting
techniques
PEUSS 2011/2012
Page 75
38

Section 8b Peussdistributions 2 Slides Compatibility Mode

Transféré par

Informations du document

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Section 8b Peussdistributions 2 Slides Compatibility Mode

Transféré par

Droits d'auteur :

Formats disponibles

An Introduction to Reliability and

Reliability and Life distributions

Objectives of the session

Reliability and Life distributions

Data types of interest

Not only times but distance or cycles etc.

Reliability and Life distributions

Reliability and Life distributions

The area under the curve is equal to 1

Reliability and Life distributions

Failure Time distributions

The Survival Function (sometimes known as

The Hazard Function

Reliability and Life distributions

Probability density Function:

Reliability and Life distributions

Reliability and Life distributions

Standard Normal distribution

Reliability and Life distributions

Cumulative distribution function

F(t) gives the

Failure Function, F(t)

Reliability and Life distributions

The probability of failure before 500 hours is 0.8

Reliability and Life distributions

R(t) = 1 - F(t) and

Reliability and Life distributions

The probability of surviving up to 500 hrs is 0.2

Reliability and Life distributions

The Hazard function is defined as probability of

Reliability and Life distributions

Reliability and Life distributions

Simplest of all life models

Reliability and Life distributions

Reliability and Life distributions

Failure rate - example

10 components of a particular type in each PCB

The MTTF is 1/0.000003 = 333,333

Reliability and Life distributions

100 units in the field

Reliability and Life distributions

When < 1 decreasing hazard function

Reliability and Life distributions

Reliability and Life distributions

When =1, h(t)= 1/ = therefore =1/

Reliability and Life distributions

Three parameter Weibull

Reliability and Life distributions

Bath-tub curve and the Weibull

Location parameter, m , is the mean

Reliability and Life distributions

Fitting parametric distributions

Reliability and Life distributions

Fitting parametric distributions

Reliability and Life distributions

Reliability and Life distributions

Reliability and Life distributions

Right Censored data

Reliability and Life distributions

Reliability and Life distributions

Reliability and Life distributions

Repaired and non-repaired data

Repaired data when interested in the pattern of

Reliability and Life distributions