WeiBull Analysis

Weibull Reliability Analysis
= http://www.rt.cs.boeing.com/MEA/stat/scholz/
http://www.rt.cs.boeing.com/MEA/stat/reliability.html
http://www.rt.cs.boeing.com/MEA/stat/scholz/weibull.html
Fritz Scholz (425-865-3623, 7L-22)
Boeing Phantom Works
Mathematics & Computing Technology
Applied Statistics
Weibull Reliability AnalysisFWS-5/21-23/20021
Wallodi Weibull
Seminal Paper
The Weibull Distribution
Weibull distribution, useful uncertainty model for
wearout failure time T
when governed by wearout of weakest subpart
material strength T
when governed by embedded aws or weaknesses,
It has often been found useful based on empirical data (e.g. Y2K)
It is also theoretically founded on the weakest link principle
T = min (X
1
, . . . , X
n
) ,
with X
1
, . . . , X
n
statistically independent random strengths or
failure times of the n links comprising the whole.
The X
i
must have a natural nite lower endpoint,
e.g., link strength 0 or subpart time to failure 0.
Theoretical Basis
Under weak conditions Extreme Value Theory shows
1
that for
large n
P(T t) 1 exp
_
_
_
_
_
t
_
_
_
_
for t , > 0, > 0
The above approximation has very much the same spirit as the
Central Limit Theorem which under some weak conditions on
the X
i
asserts that the distribution of T = X
1
+. . . +X
n
is
approximately bell-shaped normal or Gaussian for large n
Assuming a Weibull model for T, material strength or cycle time to
failure, amounts to treating the above approximation as an equality
F(t) = P(T t) = 1 exp
_
_
_
_
_
t
_
_
_
_
for t , > 0, > 0
1
see: E. Castillo, Extreme Value Theory in Engineering, Academic Press, 1988
S. Coles, An Introduction to Statistical Modeling of Extreme Values, Springer Verlag, 2001
Weibull Reproductive Property
If X
1
, . . . , X
n
are statistically independent
with X
i
Weibull(
i
, ) then
P(min(X
1
, . . . , X
n
) > t) = P(X
1
> t) P(X
n
> t)
=
n
i=1
exp
_
_
_
_
_
t
i
_
_
_
_
_
= exp
_
_
_
_
_
t
_
_
_
_
with =
_
_
_
n
i=1
i
_
_
_
1/
Hence T = min(X
1
, . . . , X
n
) Weibull(, )
This is similar to the normal reproductive property
X
1
, . . . , X
n
independent normal =
n
i=1
X
i
is normal
Weibull Parameters
The Weibull distribution may be controlled by 2 or 3 parameters:
the threshold parameter
T with probability 1
= 0 = 2-parameter Weibull model.
the characteristic life or scale parameter > 0
P(T +) = 1 exp
_
_
_
_
_
_
_
_
_
_
= 1 exp(1) = .632
regardless of the value > 0
the shape parameter > 0, usually 1
2-Parameter Weibull Model
We focus on analysis using the 2-parameter Weibull model
Methods and software tools much better developed
Estimation of in the 3-parameter Weibull model
leads to complications
When a 3-parameter Weibull model is assumed,
it will be stated explicitly
Relation of & to Statistical Parameters ( = 0)
The expectation or mean value of T
= E(T) =
_
0
t f(t) dt = (1 + 1/)
with (t) =
_
0
exp(x) x
t1
dx
The variance of T
2
= E(T )
2
=
_
0
(t )
2
f(t) dt =
2
_
(1 + 2/)
2
(1 + 1/)
_
p-quantile t
p
of T, i.e., by denition P(T t
p
) = p
t
p
= [log(1 p)]
1/
, for p = 1 exp(1) = .632 = t
p
=
Weibull Density
The cumulative distribution function F(t) = P(T t) is just
one way to describe the distribution of the random quantity T
The density function f(t) is another representation ( = 0)
f(t) = F
(t) =
dF(t)
dt
=

_
_
_
t
_
_
_
1
exp
_
_
_
_
_
t
_
_
_
_
t 0
P(t T t +dt) f(t) dt
F(t) =
_
t
0
f(x) dx
Weibull Density & Distribution Function
0 5000 10000 15000 20000
cycles
Weibull density = 10000, = 2.5
total area under density = 1
cumulative distribution function
p
p
0
1
Weibull Densities: Eect of
cycles
p
r
o
b
a
b
i
l
i
t
y

d
e
n
s
i
t
y
0 2000 4000
= 0 = 1000 = 2000
= 1000, = 2.5
cycles
p
r
o
b
a
b
i
l
i
t
y

d
e
n
s
i
t
y
0 2000 4000 6000 8000
= 1000
= 2000
= 3000
= 0, = 2.5
cycles
p
r
o
b
a
b
i
l
i
t
y

d
e
n
s
i
t
y
0 1000 2000 3000 4000
= .5
= 1
= 2
= 4
= 7
= 0, = 1000
Failure Rate or Hazard Function
A third representation of the Weibull distribution is through the
hazard or failure rate function
(t) =
f(t)
1 F(t)
=

_
_
_
t
_
_
_
1
=
d
dt
[log(1 F(t))]
(t) is increasing t for > 1 (wearout)
(t) is decreasing t for < 1
(t) is constant for = 1 (exponential distribution)
P(t T t +dt|T t) =
P(t T t +dt)
P(T t)

f(t) dt
1 F(t)
= (t) dt
F(t) = 1exp
_
_
t
0
(x) dx
_
and f(t) = (t) exp
_
_
t
0
(x) dx
_
Exponential Distribution
The exponential distribution is a special case: = 1 & = 0
F(t) = P(T t) = 1 exp
_
_
_
_
_
_
for t 0
This distribution is useful when parts fail due to
random external inuences and not due to wear out
Characterized by the memoryless property,
a part that has not failed by time t is as good as new,
past stresses without failure are water under the bridge
Good for describing lifetimes of electronic components,
failures due to external voltage spikes or overloads
Unknown Parameters
Typically will not know the Weibull distribution: , unknown
Will only have sample data = estimates

,

get estimated Weibull model for failure time distribution

= double uncertainty
uncertainty of failure time & uncertainty of estimated model
Samples of failure times are sometimes very small,
only 7 fuse pins or 8 ball bearings tested until failure,
long lifetimes make destructive testing dicult
Variability issues are often not suciently appreciated
how do small sample sizes aect our condence in
estimates and predictions concerning future failure experiences?
Estimation Uncertainty
0 20000 40000 60000 80000
0
.
0
0
.
0
0
0
0
2
0
.
0
0
0
0
4
cycles
A Weibull Population: Histogram for N = 10,000 & Density

63.2% 36.8%
characteristic life = 30,000
shape = 2.5
true model
estimated model from 9 data points
Weibull Parameters & Sample Estimates
t = t p p-quantile
p=P(T < t )
characteristic life = 30,000
63.2% 36.8%
shape parameter = 4

25390 3.02 33860 4.27 29410 5.01
parameter estimates from three samples of size n = 10
Generation of Weibull Samples
Using the quantile relationship t
p
= [log(1 p)]
1/
one can generate a Weibull random sample of size n by
generating a random sample U
1
, . . . , U
n
from a uniform [0, 1] distribution
and computing T
i
= [log(1 U
i
)]
1/
, i = 1, . . . , n.
Then T
1
, . . . , T
n
can be viewed as a random sample of size n
from a Weibull population or Weibull distribution
with parameters & .
Simulations are useful in gaining insight on estimation procedures
Graphical Methods
Suppose we have a complete Weibull sample of size n: T
1
, . . . , T
n
Sort these values from lowest to highest: T
(1)
T
(2)
. . . T
(n)
Recall that the p-quantile is t
p
= [log(1 p)]
1/
Compute t
p
1
< . . . < t
p
n
for p
i
= (i .5)/n, i = 1, . . . , n
Plot the points (T
(i)
, t
p
i
), i = 1, . . . , n
We expect these points to cluster around main diagonal
Weibull Quantile-Quantile Plot: Known Parameters
T
t
p
0 5000 10000 15000
0
5
0
0
0
1
0
0
0
0
1
5
0
0
0
Weibull QQ-Plot: Unknown Parameters
Previous plot requires knowledge of the unknown parameters &
Note that
log (t
p
) = log() +w
p
/ , where w
p
= log [log(1 p)]
Expect points
_
log[T
(i)
], w
p
i
_
, i = 1, . . . , n, to cluster around line
with slope 1/ and intercept log()
This suggests estimating & from a tted line, by eye or least
squares
Maximum Likelihood Estimation
If t
1
, . . . , t
n
are the observed sample values one can contemplate
the probability of obtaining such a sample or of values nearby, i.e.,
P (T
1
[t
1
dt/2, t
1
+dt/2], . . . , T
n
[t
n
dt/2, t
n
+dt/2])
= P (T
1
[t
1
dt/2, t
1
+dt/2]) P (T
n
[t
n
dt/2, t
n
+dt/2])
f
,
(t
1
)dt f
,
(t
n
)dt
where f(t) = f
,
(t) is the Weibull density with parameters (, )
Maximum likelihood estimation maximizes this probability
over & = maximum likelihood estimates (m.l.e.s)

and

= the most likely Weibull model explaining the data

General Remarks on Estimation
MLEs tend to be optimal in large samples (lots of theory)
Method is very versatile in extending to may other data scenarios
censoring and covariates
Least squares method applied to QQ-plot is not entirely appropriate
tends to be unduly aected by stray observations
not as versatile to extend to other situations
Weibull Plot: n = 20
cycles/hours
p
r
o
b
a
b
i
l
i
t
y

10 20 50 100 200 500 1000
.001
.010
.100
.200
.500
.900
.990
.632
true model, Weibull( 100 , 3 )
m.l.e. model, Weibull( 108 , 3.338 )
least squares model, Weibull( 107.5 , 3.684 )
Weibull Plot: n = 100
cycles/hours
p
r
o
b
a
b
i
l
i
t
y

10 20 50 100 200 500 1000
.001
.010
.100
.200
.500
.900
.990
.632
m.l.e. model, Weibull( 102.8 , 2.981 )
least squares model, Weibull( 102.2 , 3.144 )
Tests of Fit (Graphical)
The Weibull plots provide an informal diagnostic
for checking the Weibull model assumption
The anticipated linearity is based on the Weibull model properties
Strong nonlinearity indicates that the model is not Weibull
Sorting out nonlinearity from normal statistical point scatter
takes a lot of practice and a good sense for the eect
of sample size on the variation in point scatter (simulation helps)
Formal tests of t are available for complete samples
2
and also for some other censored data scenarios
2
R.B. DAgostino and M.A. Stephens, Goodness-of-Fit Techniques, Marcel Dekker 1986
Formal Goodness-of-Fit Tests
Let F
(t) be the tted Weibull distribution function

Let

F
n
(t) =
#{T
i
t; i=1,...,n}
n
be the empirical distribution function
Compute a discrepancy metric D between F
and

F
n
,
D
KS
(F
,

F
n
) = sup
t
(t)

F
n
(t)
Kolmogorov-Smirnov
D
CvM
(F
,

F
n
) =
_
0
_
F
(t)

F
n
(t)
_
2
f
(t) dt Cramer-von Mises

D
AD
(F
,

F
n
) =
_

0
_
F
(t)

F
n
(t)
_
2
F
(t)(1 F
(t))
f
(t) dt Anderson-Darling
The distributions of D, when sampling from a Weibull population,
are known and p-values of observed values d of D can be calculated
p = P(D d) = BCSLIB: HSPFIT
Kolmogorov-Smirnov Distance
0 50 100 150 200 250
0
.
0
0
.
2
0
.
4
0
.
6
0
.
8
1
.
0
K-S distance

n = 10
Weibull Plots: n = 10 (solid line is true model)
3.8 4.2 4.6 5.0

-
3
-
2
-
1
0
1
p(KS) = 0.62
p(CvM) = 0.54
p(AD) = 0.62
n = 10

4.0 4.4 4.8

-
3
-
2
-
1
0
1
p(KS) = 0.28
p(CvM) = 0.27
p(AD) = 0.35
3.8 4.4 5.0

-
3
-
2
-
1
0
1
p(KS) = 0.88
p(CvM) = 0.62
p(AD) = 0.66
3.8 4.2 4.6 5.0

-
3
-
2
-
1
0
1
p(KS) = 0.88
p(CvM) = 0.64
p(AD) = 0.69
4.0 4.4 4.8

-
3
-
2
-
1
0
1
p(KS) = 0.72
p(CvM) = 0.53
p(AD) = 0.6
3.8 4.2 4.6

-
3
-
2
-
1
0
1
p(KS) = 0.16
p(CvM) = 0.21
p(AD) = 0.32
4.0 4.4 4.8

-
3
-
2
-
1
0
1
p(KS) = 0.34
p(CvM) = 0.23
p(AD) = 0.22
3.8 4.2 4.6

-
3
-
2
-
1
0
1
p(KS) = 0.86
p(CvM) = 0.56
p(AD) = 0.57
Weibull Plots: n = 20
3.5 4.0 4.5 5.0

-
3
-
2
-
1
0
1
p(KS) = 0.55
p(CvM) = 0.49
p(AD) = 0.49
n = 20

2.5 3.5 4.5

-
3
-
2
-
1
0
1
p(KS) = 0.13
p(CvM) = 0.027
p(AD) = 0.028
4.2 4.6
-
3
-
2
-
1
0
1
p(KS) = 0.12
p(CvM) = 0.053
p(AD) = 0.048
3.8 4.2 4.6 5.0

-
3
-
2
-
1
0
1
p(KS) = 0.46
p(CvM) = 0.39
p(AD) = 0.38
3.6 4.0 4.4 4.8

-
3
-
2
-
1
0
1
p(KS) = 0.46
p(CvM) = 0.41
p(AD) = 0.41
3.0 4.0 5.0

-
3
-
2
-
1
0
1
p(KS) = 0.47
p(CvM) = 0.63
p(AD) = 0.69
4.0 4.4 4.8

-
3
-
2
-
1
0
1
p(KS) = 0.85
p(CvM) = 0.6
p(AD) = 0.64
3.6 4.0 4.4 4.8

-
3
-
2
-
1
0
1
p(KS) = 0.04
p(CvM) = 0.017
p(AD) = 0.023
2.0 3.0 4.0 5.0

-
4
-
3
-
2
-
1
0
1
p(KS) = 0.7
p(CvM) = 0.41
p(AD) = 0.43
n = 50

3.0 4.0 5.0

-
4
-
3
-
2
-
1
0
1
p(KS) = 0.22
p(CvM) = 0.35
p(AD) = 0.32
3.5 4.0 4.5

-
4
-
3
-
2
-
1
0
1
p(KS) = 0.15
p(CvM) = 0.23
p(AD) = 0.12
3.5 4.5
-
4
-
3
-
2
-
1
0
1
p(KS) = 0.76
p(CvM) = 0.49
p(AD) = 0.52
3.5 4.0 4.5

-
4
-
3
-
2
-
1
0
1
p(KS) = 0.88
p(CvM) = 0.66
p(AD) = 0.72
3.5 4.0 4.5 5.0

-
4
-
3
-
2
-
1
0
1
p(KS) = 0.27
p(CvM) = 0.22
p(AD) = 0.23
3.5 4.0 4.5 5.0

-
4
-
3
-
2
-
1
0
1
p(KS) = 0.85
p(CvM) = 0.56
p(AD) = 0.62
3.5 4.0 4.5 5.0

-
4
-
3
-
2
-
1
0
1
p(KS) = 0.18
p(CvM) = 0.035
p(AD) = 0.039
3.0 4.0 5.0

-
4
-
2
0
p(KS) = 0.58
p(CvM) = 0.26
p(AD) = 0.3
n = 100

3.5 4.5
-
4
-
2
0
p(KS) = 0.86
p(CvM) = 0.68
p(AD) = 0.72
3.0 4.0 5.0

-
4
-
2
0
p(KS) = 0.61
p(CvM) = 0.48
p(AD) = 0.55
3.5 4.5
-
4
-
2
0
p(KS) = 0.21
p(CvM) = 0.26
p(AD) = 0.43

3.0 4.0 5.0
-
4
-
2
0
p(KS) = 0.47
p(CvM) = 0.3
p(AD) = 0.37
3.0 4.0 5.0

-
4
-
2
0
p(KS) = 0.053
p(CvM) = 0.025
p(AD) = 0.032
3.5 4.5
-
4
-
2
0
p(KS) = 0.87
p(CvM) = 0.58
p(AD) = 0.6
2.5 3.5 4.5

-
4
-
2
0
p(KS) = 0.72
p(CvM) = 0.45
p(AD) = 0.43
Estimates and Condence Bounds/Intervals
For a target
= , = , = t
p
= [log(1 p)]
1/
, or = P
,
(T t)
one gets corresponding m.l.e

by replacing (, ) by (
,

)
Such estimates

vary around the target due to sampling variation
Capture the estimation uncertainty via condence bounds
For 0 < < 1 get 100% lower/upper condence bounds

L,
&

U,
P
_
L,

_
= or P
_

U,
_
=
For > .5 get a 100(2 1)% condence interval by [
L,
,

U,
]
P
_
L,

U,
_
= 2 1 =

Condence Bounds & Sampling Variation, n = 10
9
0
%

c
o
n
f
i
d
e
n
c
e

i
n
t
e
r
v
a
l
s
1 2 3 4 5
2
5
0
0
0
3
5
0
0
0
samples
L: Lawless Exact Method (RAP); B: Bain Exact Method (Tables); M: MLE Approximate Method
L B M
true
9
0
%

c
o
n
f
i
d
e
n
c
e

i
n
t
e
r
v
a
l
s
1 2 3 4 5
0
2
4
6
8
L B M
true
Condence Bounds & Sampling Variation, n = 10
9
5
%

l
o
w
e
r

c
o
n
f
i
d
e
n
c
e

b
o
u
n
d
s
1 2 3 4 5
0
5
0
0
0
1
5
0
0
0
2
5
0
0
0
samples
L: Lawless Exact Method (RAP); B: Bain Exact Method (Tables); M: MLE Approximate Method
L B M
true 10-percentile
9
5
%

u
p
p
e
r

c
o
n
f
i
d
e
n
c
e

b
o
u
n
d
s
1 2 3 4 5
0
.
0
0
.
1
0
0
.
2
0
L B M
true P(T < 10,000)
Condence Bounds & Eect of n

9
5
%

c
o
n
f
i
d
e
n
c
e

i
n
t
e
r
v
a
l
s
1
5
0
0
0
2
5
0
0
0
3
5
0
0
0

5 10 20 50 100 200 500 1000
sample size
true

9
5
%

c
o
n
f
i
d
e
n
c
e

i
n
t
e
r
v
a
l
s
1
2
3
4
5
6
7

true
Condence Bounds & Eect of n

9
5
%

u
p
p
e
r

c
o
n
f
i
d
e
n
c
e

b
o
u
n
d
s
0
.
0
0
.
0
5
0
.
1
5
sample size

5 10 20 50 100 200 500 1000
true P(T < 10000)

9
5
%

l
o
w
e
r

c
o
n
f
i
d
e
n
c
e

b
o
u
n
d
s
5
0
0
0
1
0
0
0
0
2
0
0
0
0

true t .10
Incomplete Data or Censored Failure Times
Type I censoring or time censoring:
units are tested until failure or until a prespecied time has elapsed
Type II censoring or failure censoring:
only the r lowest values of the total sample of size n become known
this shows up when we put n units on test at the same time
and terminate the test after the rst r units have failed
Interval censoring, inspection data, grouped data:
i
th
unit is only known to have failed between two known
inspection time points, i.e., failure time T
i
falls in (s
i
, e
i
]
only bracketing intervals (s
i
, e
i
), i = 1, . . . , n, become known
Incomplete Data or Censored Failure Times (continued)
Random right censoring:
units are observed until failed or removed from observation
due to other causes (dierent failure modes, competing risks)
Multiple right censoring:
units are put into service at dierent times
and times to failure or censoring are observed
Data can also combine several of the above censoring phenomena
It is important that the censoring mechanism should not correlate
with the (potential) failure times,
i.e., no censoring of anticipated failures
All data (censored & uncensored) should enter analysis
otherwise bias results
Type I or Time Censored Data: n = 10
cycles
0 2000 4000 6000 8000 10000
1
2
3
4
5
6
7
8
9
10
unit
?
?
X
X
?
X
X
?
?
X
Type II or Failure Censored Data: n = 10
cycles
0 2000 4000 6000 8000 10000
1
2
3
4
5
6
7
8
9
10
unit
X
X
X
X
X
X
?
?
?
?
Interval Censored Data: n = 10
cycles
0 2000 4000 6000 8000 10000
1
2
3
4
5
6
7
8
9
10
unit
( ]
?
( ?
( ]
?
( ]
( ?
?
( ]
?
Multiply Right Censored Data: n = 10
cycles
0 2000 4000 6000 8000 10000
1
2
3
4
5
6
7
8
9
10
unit
X
?
X
X
X
X
?
?
?
?
Nonparametric MLEs of CDF
For complete, type I & II censored data the nonparametric MLE is
F(t) =
#{T
i
t; i = 1, . . . , n}
n
,
< t < for complete sample,
< t t
0
for type I censored (at t
0
) sample,
< t T
(r)
for type II censored (at T
(r)
, r
th
smallest failure time)
For multiply right censored data with failures at t
1
< . . . < t
k
the
nonparametric MLE (Kaplan-Meier or Product Limit Estimator) is
F(t) = 1
_
(1 p
1
)
1
(t)
(1 p
k
)
k
(t)
_
,
i
(t) =
_
_
1 for t
i
t
0 for t
i
> t
where p
i
= d
i
/n
i
, n
i
= # units known to be at risk just prior to t
i
and d
i
= # units that failed at t
i
The nonparametric MLE for interval censored data is complicated
Empirical CDF (complete samples)
0 5000 15000 25000
0
.
0
0
.
4
0
.
8
n = 10
empirical and true
and Weibull mle
distribution functions
0 5000 15000 25000
0
.
0
0
.
4
0
.
8
n = 30
0 5000 15000 25000
0
.
0
0
.
4
0
.
8
n = 50
0 5000 15000 25000
0
.
0
0
.
4
0
.
8
n = 100
Nonparametric MLE for Type I Censored Data
0 5000 15000 25000
0
.
0
0
.
4
0
.
8
...
n = 10
nonparametric mle of CDF
Weibull mle of CDF
and true CDF
0 5000 15000 25000
0
.
0
0
.
4
0
.
8
. . . . . . .
n = 30
0 5000 15000 25000
0
.
0
0
.
4
0
.
8
. . . .. . . ..
n = 50
0 5000 15000 25000
0
.
0
0
.
4
0
.
8
. . . . . . .. . .. . . . . . . .
n = 100
Nonparametric MLE for Type II Censored Data
0 5000 15000 25000
0
.
0
0
.
4
0
.
8
.. .
n = 10
nonparametric mle of CDF
Weibull mle of CDF
and true CDF
0 5000 15000 25000
0
.
0
0
.
4
0
.
8
... . ... .
n = 30
0 5000 15000 25000
0
.
0
0
.
4
0
.
8
..... ... .. . .. ..
n = 50
0 5000 15000 25000
0
.
0
0
.
4
0
.
8
................. ....... . .
n = 100
Kaplan-Meier Estimates (multiply right censored samples)
0 5000 15000 25000
0
.
0
0
.
4
0
.
8
. . . . . . . .
n = 10
Kaplan-Meier CDF
Weibull mle for CDF
true CDF
0 5000 15000 25000
0
.
0
0
.
4
0
.
8
. . . . . . . . . . . . . . . . . .
n = 30
0 5000 15000 25000
0
.
0
0
.
4
0
.
8
. . . . . . . . . . . . . . . . . . . . . . . . . . . .
n = 50
0 5000 15000 25000
0
.
0
0
.
4
0
.
8
. . . . . . . .. . .. . . . . . . . . .. . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
n = 100
Nonparametric MLE for Interval Censored Data
cycles
C
D
F
0 5000 10000 15000
0
.
0
0
.
2
0
.
4
0
.
6
0
.
8
1
.
0
superimposed is the true Weibull(10000,2.5)
that generated the 1,000 interval censored data cases
inspection points roughly 3000 cycles apart
Nonparametric MLE for Interval Censored Data
cycles
C
D
F
0 5000 10000 15000
0
.
0
0
.
2
0
.
4
0
.
6
0
.
8
1
.
0
superimposed is the true Weibull(10000,2.5)
that generated the 10,000 interval censored data cases
inspection intervals randomly generated from same Weibull
Application to Y2K Questionnaire Return Data
days
p
r
o
b
a
b
i
l
i
t
y

o
f

q
u
e
s
t
i
o
n
n
a
i
r
e

r
e
t
u
r
n
0 50 100 150 200 250 300
0
.
0
0
.
2
0
.
4
0
.
6
0
.
8
1
.
0
estimated .99-quantile
8124 cases
Plotting Positions for Weibull Plots (Censored Case)
For complete samples we plotted (log[T
(i)
], log [log(1 p
i
)]) with
p
i
=
i .5
n
=
1
2
_
F(T
(i)
) +

F(T
(i1)
)
_
=
1
2
_
_
i
n
+
i 1
n
_
_
For type I & II & multiply right censored samples
we plot in analogy (log[T
(i)
], log [log(1 p
i
)]), where
T
(1)
< . . . < T
(k)
are the k distinct observed failure times and
p
i
=
1
2
_
F(T
(i)
) +

F(T
(i1)
)
_
and

F(t) is the nonparametric MLE of F(t) for the censored sample
Weibull Plot, Multiply Right Censored Sample n = 50
cycles/hours
p
r
o
b
a
b
i
l
i
t
y
10 20 50 100 200 500 1000

.001
.010
.100
.200
.500
.900
.990
.632
m.l.e. model, Weibull( 96.27 , 2.641 ) n = 50
least squares model, Weibull( 96.3 , 2.541 ) n = 50
Software: MLEs & Condence Bounds for Censored Data
For complete & type II censored data RAP computes exact coverage
condence bounds for , , t
p
, and P(T t).
WEIBREG and commercial software (Weibull++, WeibullSMITH,
common statistical packages) compute approximate condence
bounds for above targets, based on large sample m.l.e. theory
RAP & WEIBREG are Boeing code, runs within DOS mode
of Windows95 (interface clumsy, but job gets done), contact me
Boeing has a site license for Weibull++ Version 4
in the Puget Sound area (with upgrade pressure)
contact David Twigg (425) 266-7919, david.twigg@pss.boeing.com
There is a Weibull++ Version 6 out
Other Software
S-Plus Statistical data analysis environment ( $2000 for PC licence)
R is a GNU version of S-plus, freely available
either from: http://cran.r-project.org/ for lots of operating
systems or inside Boeing for Windows
http://starfly.ca.boeing.com/KIRTSweb/oss_toolkit/
toolkit/optional.html#L048
Both S-Plus and R have Terry Therneaus survival analysis package
built in or as add-on. Using it takes some sophistication.
My web-based Weibull analysis tool is based on R
http://www.rt.cs.boeing.com/MEA/stat/scholz/weibull.html
Also freely available is the very user friendly package SPLIDA by Bill
Meeker which can be added to S-Plus 4.5 or S-Plus 2000
http://www.public.iastate.edu/~splida/
Software in Progress (Needs Pull!)
The current version of web-based Weibull analysis tool and also
most other versions based on large sample approximations for the
MLEs have some defect for extreme sample with very few failures:
Condence bounds for quantiles t
p
are not always monotone
increasing in p, which does not make sense. Also quantile and
probability bounds are not inverses to each other.
I have an updated version of WEIBREG based on a new method
that avoids this problem.
It also includes a bootstrap version of the new monotone method.
However, it is not yet released.
Quantile and probability bounds are true inverses of each other.
The plan is to update the web based tool and to also create web
based versions of RAP and ABVAL to bypass the legacy issue.
Weibull Plot With Condence Bound
1 2 3 5 10 20 50 100 200 500 1000
0.001
0.010
0.100
0.200
0.500
0.900
0.990
0.999
alpha = 102.4 [ 82.52 , 143.8 ] 95 %

beta = 2.524 [ 1.352 , 3.696 ] 95 %
n = 30 , r = 13
Weibull Plot With Condence Bound (Variation)
1 2 3 5 10 20 50 100 200 500 1000
0.001
0.010
0.100
0.200
0.500
0.900
0.990
0.999
alpha = 89.79 [ 77.4 , 109.4 ] 95 %

beta = 3.674 [ 2.15 , 5.198 ] 95 %
n = 30 , r = 13
cycles x 1000
p
r
o
b
a
b
i
l
i
t
y

1 2 5 10 20 50 100
.
0
0
1
.
0
1
0
.
1
0
0
.
5
0
0
.
9
0
0
.
9
9
9
.
6
3
2
m.l.e.
95 % qconfidence bounds
95 % pconfidence bounds
Problem Weibull Plot
^
= 38600 , 95 % conf. interval ( 30690 , 48560 )
^
= 6.088 , 95 % conf. interval ( 1.987 , 18.65 )
n = 5 , r = 2
Data: 15900
+
, 20000
+
, 25500, 36000
+
, 40010
Weibull Regression Model & Analysis
Recall t
p
= [log(1 p)]
1/
or
log(t
p
) = log() +w
p
/ with w
p
= log[log(1 p)]
Often we deal with failure data collected under dierent conditions
dierent part types
dierent environmental conditions
dierent part users
Linear regression model for log() (multiplicative on )
log(
i
) = b
1
Z
i1
+. . . +b
p
Z
ip
typically Z
i1
1
where Z
i1
, . . . , Z
ip
are known covariates for i
th
unit
Now we have p + 1 unknown parameters , b
1
, . . . , b
p
parameter estimates & condence bounds using MLEs
Accelerated Life Testing: Inverse Power Law
Units last too long under normal usage conditions
Increase stress to accelerate failure time
Increase voltage and accelerate life via inverse power law
T(Volt) = T(Volt
U
)
_
_
_
Volt
Volt
U
_
_
_
c
, where usually c < 0
this means
(Volt) = (Volt
U
)
_
_
_
Volt
Volt
U
_
_
_
c
or log [(Volt)] = log [(Volt
U
)] +c log (Volt) c log (Volt
U
)
= b
1
+b
2
Z
with
b
1
= log [(Volt
U
)] c log (Volt
U
) , b
2
= c, and Z = log (Volt)
Accelerated Life Testing: Arrhenius Model
Another way of accelerating failure in processes involving
chemical reaction rates is to increase the temperature
Arrhenius proposed the following acceleration model with
temperature measured in Kelvin (tempK = temp
C + 273.15)
T(temp) = T(temp
U
) exp
_
temp

temp
U
_
_
or
log [(temp)] = log [(temp
U
)] +

temp

temp
U
= b
1
+b
2
Z
with
b
1
= log [(temp
U
)]

temp
U
, b
2
= , and Z = temp
1
Pooling Data With Dierent s
Suppose we have three groups of failure data
T
1
, . . . , T
n
1
W(
1
, ), T
n
1
+1
, . . . , T
n
1
+n
2
W(
2
, ),
T
n
1
+n
2
+1
, . . . , T
n
1
+n
2
+n
3
W(
3
, )
We can analyze the whole data set of N = n
1
+n
2
+n
3
values jointly
using the following model for
j
and dummy covariates Z
1,j
, Z
2,j
& Z
3,j
log(
j
) = b
1
Z
1,j
+b
2
Z
2,j
+b
3
Z
3,j
, j = 1, 2, . . . , N
where b
1
= log(
1
), b
2
= log(
2
)log(
1
) and b
3
= log(
3
)log(
1
)
Z
1,j
= 1 for j = 1, . . . , N, Z
2,j
= 1 for j = n
1
+ 1, . . . , n
1
+n
2
, Z
2,j
= 0 else
and Z
3,j
= 1 for j = n
1
+n
2
+ 1, . . . , N & Z
3,j
= 0 else.
Advantage: Smaller estimation error
all data are used in estimating the (assumed) common
Pooling Data (continued)
log(
1
) = b
1
1 +b
2
0 +b
3
0 = log(
1
)
.
.
.
.
.
.
log(
n
1
) = b
1
1 +b
2
0 +b
3
0 = log(
1
)
.
.
.
.
.
.
log(
n
1
+1
) = b
1
1 +b
2
1 +b
3
0 = log(
1
) + [log(
2
) log(
1
)] = log(
2
)
.
.
.
.
.
.
log(
n
1
+n
2
) = b
1
1 +b
2
1 +b
3
0 = log(
1
) + [log(
2
) log(
1
)] = log(
2
)
.
.
.
.
.
.
log(
n
1
+n
2
+1
) = b
1
1 +b
2
0 +b
3
1 = log(
1
) + [log(
3
) log(
1
)] = log(
3
)
.
.
.
.
.
.
log(
N
) = b
1
1 +b
2
0 +b
3
1 = log(
1
) + [log(
3
) log(
1
)] = log(
3
)
Accelerated Life Testing: Model & Data
Voltage
t
h
o
u
s
a
n
d

c
y
c
l
e
s
50 100 150 200 250 300 350 400 450 500

1
2
5
10
20
50
100
200
Accelerated Life Testing: Model, Data & MLEs
Voltage
t
h
o
u
s
a
n
d

c
y
c
l
e
s
m
le line
true line
mle for .01-quantile
95% lower bound for .01-quantile
50 100 150 200 250 300 350 400 450 500
1
2
5
10
20
50
100
200
Analysis for Data from Exponential Distribution
T E() Exponential with mean , i.e., E() = W( = , = 1)
It suces to get condence bounds for since all other quantities
of interest are explicit and monotone functions of
F
(t
0
) = P
(T t
0
) = 1 exp(t
0
/)
R
(t
0
) = P
(T > t
0
) = exp(t
0
/)
and
t
p
() = [log(1 p)]
Complete & Type II Censored Exponential Samples
T E() Exponential with mean , E() = W( = , = 1)
For complete exponential samples T
1
, . . . , T
n
E()
or for a type II censored sample of this type,
i.e., with observed r lowest values T
(1)
. . . T
(r)
, 1 r n,
one gets 100% lower condence bounds for by computing
L,
=
2 TTT
2
2r,
,
where TTT = T
(1)
+ +T
(r)
+ (n r)T
(r)
= Total Time on Test,
and
2
2r,
= -quantile of the
2
2r
distribution
get this in Excel via = GAMMAINV(, r, 1) or = CHIINV(1 , 2r)/2
These bounds have exact condence coverage properties
Multiply Right Censored Exponential Data
Here we dene TTT = T
1
+. . . +T
n
as the total time on test
but now T
i
is either the observed failure time of the i
th
part
or the observed right censoring time of the i
th
part
The number r of observed failures is random here, 0 r n
Approximate 100% lower condence bounds for are computed as
L,
=
2 TTT
2
2r+2,
This also holds for r = 0, i.e., no failures at all over exposure period
Weibull Shortcut Methods for Known
T W(, ) (Weibull) T
E() Exponential with mean =
As 100% lower condence bound for compute

L,
=
_
_
_
_
_
2 TTT()
2
f,
_
_
_
_
_
1/
where for complete or type II censored samples we take f = 2r and
TTT = T
(1)
+ +T
(r)
+ (n r)T
(r)
and for multiply right censored data we take f = 2r + 2 and
TTT = T
1
+ +T
n
.
Sensitivity should be studied by repeating analyses for several s
A- and B-Allowables, ABVAL Program
My rst Weibull work led to the program ABVAL (1983)
supporting Cecil Parsons and Ron Zabora w.r.t. MIL-HDBK-5
ABVAL computes A- and B-Allowables based on samples
from the 3-parameter Weibull distribution.
The A-Allowable is a 95% lower condence bound
for 99% of the population, i.e., for the .01-quantile.
The B-Allowable is a 95% lower condence bound
for 90% of the population, i.e., for the .10-quantile.
Hybrid approach of using Mann-Fertig threshold estimate
combined with maximum likelihood estimates for and ,
took lot of simulation and tuning and is very specialized in its
capabilities
It is now a method in MIL-HDBK-5, I still maintain software
Selected References
Abernethy, R.B. (1996), The New Weibull Handbook, 2nd edition,
Reliability Analysis Center 800-526-4802
Bain, L.J. (1978), Statistical Analysis of Reliability and Life-Testing
Models, Marcel Dekker, New York
Jeng, S.-L. and Meeker, W.Q. (2000), Comparisons of approximate
condence procedures for type I censored data, J. Amer. Statist.
Assoc. 42, 135-148.
Kececioglu, D. (1993/4), Reliability & Life Testing Handbook Vol. 1
& 2, Prentice Hall PTR, Upper Saddle River, NJ.
Lawless, J.F. (1982), Statistical Models and Methods for Lifetime
Data, John Wiley & Sons, New York
Mann, N.R., Schafer, R.E., and Singpurwalla, N.D. (1974), Methods
for Statistical Analysis of Reliability and Life Data, John Wiley &
Sons, New York
References (Continued)
Meeker, W.Q. and Escobar, L.A. (1995), Teaching about
approximate condence regions based on maximum likelihood
estimation, The American Statistician 49 48-53.
=Meeker, W.Q. and Escobar, L.A. (1998), Statistical Methods for
Reliability Data, John Wiley & Sons, New York
Meeker, W.Q. and Hahn, G.J. (1985), Volume 10: How to Plan an
Accelerated Life TestSome Practical Guidelines, American Society
for Quality Control
Robinson, J.A. (1983), Bootstrap condence intervals in
location-scale models with progressive censoring, Technometrics 25
179-187.
Nelson, W. (1982), Applied Life Data Analysis, John Wiley & Sons,
New York
References (Continued)
Nelson, W. (1985), Weibull analysis of reliability data with few or
no failures, Journal of Quality Technology 17, 140-146
Nelson, W. (1990), Accelerated Testing, Statistical Models, Test
Plans and Data Analyses, John Wiley & Sons, New York
Therneau, T.M. and Grambsch, P.M. (2000), Modeling Survival
Data, Extending the Cox Model, Springer Verlag, New York.
Scholz, F.W. (1994), Weibull and Gumbel distribution exact
condence bounds. BCSTECH-94-019 (RAP)
Scholz, F.W. (1996) Maximum likelihood estimation for type I
censored Weibull data including covariates. ISSTECH-96-022
Scholz, F.W. (1997), Condence bounds for type I censored Weibull
data including covariates. SSGTECH-97-025 (WEIBREG)
The Boeing Company, P.O. Box 3707, MS-7L-22, Seattle WA 98124-2207

WeiBull Analysis

Enviado por

Dados do documento

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

WeiBull Analysis

Enviado por

Direitos autorais:

Formatos disponíveis

Weibull Reliability Analysis

get estimated Weibull model for failure time distribution

= the most likely Weibull model explaining the data

(t) be the tted Weibull distribution function

(t) dt Cramer-von Mises

3.8 4.2 4.6 5.0

4.0 4.4 4.8

3.8 4.4 5.0

3.8 4.2 4.6 5.0

4.0 4.4 4.8

3.8 4.2 4.6

4.0 4.4 4.8

3.8 4.2 4.6

3.5 4.0 4.5 5.0

2.5 3.5 4.5

3.8 4.2 4.6 5.0

3.6 4.0 4.4 4.8

3.0 4.0 5.0

4.0 4.4 4.8

3.6 4.0 4.4 4.8

2.0 3.0 4.0 5.0

3.0 4.0 5.0

3.5 4.0 4.5

3.5 4.0 4.5

3.5 4.0 4.5 5.0

3.5 4.0 4.5 5.0

3.5 4.0 4.5 5.0

3.0 4.0 5.0

3.0 4.0 5.0

3.0 4.0 5.0

2.5 3.5 4.5

Weibull Reliability AnalysisFWS-5/21-23/200235

10 20 50 100 200 500 1000

alpha = 102.4 [ 82.52 , 143.8 ] 95 %

alpha = 89.79 [ 77.4 , 109.4 ] 95 %

50 100 150 200 250 300 350 400 450 500

E() Exponential with mean =

As 100% lower condence bound for compute

Você também pode gostar