Beruflich Dokumente
Kultur Dokumente
for
Life Data
Jang JuSu
Historical Background
1. Waloddi Weibull (1887-1979) invented the Weibull distribution in
1937.
2. His 1951 paper represents the culmination of his work in reliability
analysis.
3. The U.S.Air Force recognized the merit of Weibulls methods and
funded his research to 1975.
4. Leonard Johnson at General Motors, improved Weibulls methods.
(Weibull used mean rank values for plotting but Johnson suggested the use of
median rank values)
Application to
Schedule Maintenance
Project Management for a System or Plant
Warranty and Cost Analysis
Weibull Distribution
1. Probability Density Function
f (t )
,t 0
F (t ) 1 e
,t 0
: Shape parameter
: Scale parameter
: Location parameter
Weibull Distribution
Shape parameter variations with 10 .
0.4
0.4
0.3
0.3
0.5
0.2
0.2
0.1
0.1
10
15
20
25
30
0.4
0.4
0.3
0.3
0.2
10
20
25
30
25
30
2.5
0.2
0.1
15
0.1
10
15
20
25
30
10
15
20
Weibull Distribution
Scale parameter variations with
0.8
0.08
0.6
0.4
0.04
0.2
0.02
0.5
1.5
10
0.06
2.5
0.008
10
15
20
25
30
0.0008
100
0.006
0.004
0.0004
0.002
0.0002
50
100
150
200
250
1000
0.0006
300
500
1000
1500
2000
2500
3000
Weibull Distribution
Location parameter shift the curve along x-axis.
Weibull Distribution
Weibull Distribution
F (t ) 1 e
F ( ) 1 e
1 e 1 0.632 .
10
Interpretation
Assume that we have 2-types data, failure and suspended,
1. For scale parameter ,
In general the more we have suspended data the shape
change but the scale parameter ,
parameter ,
, hardly
,
will be
increased.
2. For shape parameter ,
11
Interpretation
Inadequate burn-in test or screening,
Production problems, misassembled, quality control,
Overhaul problems,
Solid state electronic failure.
12
Interpretation
Failures due to nature,
Mixtures of data from 3 or more failure modes or different
13
s.
Types of Data
0
1
2
Time
Complete
Data
0
1
Unit
Unit
Unit
Time
Type II
0
1
2
Multiply Failure
Censored
2
3
4
Time
0
1
Random
Censored
2
Unit
Unit
Unit
Time
3
4
14
Time
Type I
Time
Interval
(Grouped) Data
Failure time
Status
Median Rank
1500
0.0946
7.3132
-2.3088
1750
2250
S
F
0.2523
7.7187
-1.2353
4000
4300
F
F
0.4099
0.5676
8.2940
8.3664
-0.6397
-0.1763
5000
7000
0.8041
8.8537
0.4887
15
reverse rank Rn 1 ( N 1)
Rn
reverse rank 1
R0 0 , n 1 , 2 , 3 , . . .
2. Benards Approximation formula
( Rn 0.3)
BRn
N 0.4
n 1, 2 , 3,...
MOASOFT Training Course for
16
F (t ) 1 e
1 /(1 F (t )) e
1 F (t ) e
t
t
ln[1 /(1 F (t ))]
17
B
MOASOFT Training Course for
eA
18
e
XC
CX t (Y XC ) 0
X Y X XC
t
C ( X t X ) 1 X tY
MOASOFT Training Course for
y1
y2
x1 ,1
x2 ,1
..
y
n
19
.. ,1
x ,1
n
B
C
A
20
f (t ) t
,t 0
ti
ti e
i 1
k
21
ln L k ln ln ti
i 1
k
1 k
ti
i 1
(ln L) k
1
ln ti ti ln ti 0
t
i
(ln L)
k
2 0
MOASOFT Training Course for
22
G( )
t
i ln ti
i 1
t
i
1 r
1
ln ti 0
r i 1
i 1
t
i
i 1
23
G( n )
n 1 n
G`( n )
We so as to use Newton method have to find the derivative of
24
G( )
G`( )
t
i 1
(ln ti )
t
i 1
(t
ln t )
i 1
k
( t )
i 1
2
i
25
1
2
(ln L)
1 ti
(ln L)
n
2
t
i
(ln
(ln L)
1 ti
n
2
t
i
t
ti 2
) i
r
ti
2 ( 1)
1
ti
ln
1
1
ti
ln
1
26
ti 2
(ln )
2 ( 1)
Z / 2 var( )
exp
Z / 2 var( )
exp
(ln t ln )
Z / 2 var( )
Z / 2 var( )
exp
exp
27
28
Conclusion
In case of small sample -> Monte Carlo Simulation!
this will make an analysis enhanced with small sample.
Topics
Graphical
Small sample
Forecast
Risk Analysis
Confidence
Convergence
Mixed mode
MLE
None
Bad( more bias)
Bad
?
Rigorous
?
Only one
Rank Regression
OK
more accurate
more accurate
?
None
Always
Available
29
Arrhenius Model
Arrhenius Law
where
A e
E
kT
30
Arrhenius Model
Linearized relationship
log 0 ( 1 / T )
where
1 log(e) ( E / k ) 0.4343 E / k
E 2.303 k 1
31
Arrhenius Model
Arrhenius acceleration factor K
1 1
K / ' exp{( E / k )[ ]}
T T'
Example : E = 0.65 eV, T = 453 K(180 C), T = 533 K(260 C).
K = exp{(0.65/ 8.6171x10-5 )[1/453 1/533]} = 12
32
Arrhenius Model
33