Classic Reliability Theory

Part II: Failure Rate Models

SE 411
Dr. Pingfeng Wang
Office: TB-310
Email: [email protected]
Tel: 217-300-7078

1
 Constant Failure Rate

The Exponential Distribution

2
 Questions

If l(t) = l, please find:

• f(t), R(t) f(t)

• MTTF
• MTTF(T0) 5 6 2 1

• R(t|T0)
• tr 3
l(t) 4
R(t)

3
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 The Exponential Distribution
Let
l (t) = l , t  0
Then -  0t l d t '
R (t ) = e = e -lt
, t0

4
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 The CDF and PDF
-lt d F (t)
F (t ) = 1 - e f(t) = = l e -lt
dt

5
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 The MTTF
- lt
 e  1
- lt
MTTF =  e dt = |0 =
0
-l l
or  1
MTTF =  l te − lt
dt =
0 l
-M T T F
-1
note that R (M T T F ) = e M TTF = e = .3 6 8

6
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 The Design Life

For a given reliability R, let

-l tR
R( tR )= e = R
1
then tR = - ln R
l
Question : What is median time to failure, t med
What is standard deviation, 
7
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 Example #1 - CFR model

An electrical transmission line has been tested and

found to have a CFR of .001 defects per foot. Find:
a. R(100)
b. R(1000)
c. MTTF
d. 90 percent design life

8
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 Memoryless Property

- l ( t+ T 0 )
R(t + T 0 ) e
R(t | T 0 ) = = -l T
R( T 0 ) e 0

-lt -lT 0
e e -lt
= -l T 0
=e = R(t)
e

9
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 Failure Modes
Ri(t) is the reliability function for the ith failure mode,
then, assuming independence among the failure modes,
the system reliability, R(t) is found from
n
R(t ) =  R i(t ) , where R i(t ) = e -  0t l i (t )d t

i =1

n
 t n 
R (t ) =  e -  0t l i (t )dt
= exp  −   li (t ') dt '
i =1  0
i =1 
n
-  0t l (t )d t where l(t) =  l i(t)
=e
i=1
10
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 Failure Modes and CFR
If a system consists of n independent,
serially related components each with
CFR, then
n
l (t) = l =  l i
i=1
and
t
- 0 l d t - lt
R(t) = e =e

11
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 Example #2 - Failure Modes

An engine tune-up kit consists of 3 parts each

having CFRs (in failures per mile) of .000034,
.000017, and .0000086.
Find the MTTF, median time to failure, standard
deviation, and reliability of the tune-up kit at
10,000 miles.

12
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 Parts Count Approach
An integrated circuit board consists of the following components
each having a CFR.
Component a-Failure Rate(10-5) b- Quantity (a) x (b)
Diodes, silicon .00041 10 .0041
Resistors .014 25 .3500
Capacitors .0015 12 .0180
Transformer .0020 2 .0040
Relays .0065 6 .0390
Inductive devices .0004 12 .0048
total .4199 x 10-5

Therefore Rsys(t) = e- .000004199 t and MTTF = 1/.4199 x 105

13
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 All Failure Modes are CFR
1 1 1
MTTF = = =
l n n
1
l
i=1
i  MTTFi
i=1

1
where MTTFi =
li
If all components have identical failure rates, then:
1
l = n l1 and MTTF =
n l1
14
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 Poisson Process
If a component having a constant failure rate l is
immediately repaired or replaced upon failing, the
number of failures observed over a time period t has a
Poisson distribution. The probability of observing n
failures in time t is given by the Poisson probability
mass function pn(t):

( lt )
− lt n
e
pn ( t ) = n = 0, 1, 2, K
n!
E[ n ] =  = l t
2

15
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 The Gamma Distribution
LetTi = time between failure i – 1 and failure i having an
exponential distribution with parameter l.
k
Yk =  Ti
i =1
Yk = the time of the kth failure. The sum of k independent
exponential random variables has a gamma distribution with
parameters l and k.
If k is an integer (i.e. the Erlang distribution),

l k t k −1e − lt
fY (t ) = for k , l , t  0
 (k )

16
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 The Gamma Distribution
The CDF is the probability that the kth failure will occur by
time t.
The mean value for Yk is k/l, and the variance is k/l2.
The mode is (k – 1)/l.

Pr Yk  t  = FYk ( t ) = 1 − e
k −1
( lt ) i
− lt

i =0 i!
Pn ( t ) = Pr Yn  t − Pr Yn +1  t = FYn ( t ) − FYn+1 ( t )
( lt )
− lt n
e
= ; the Poisson
n!

17
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 Related Failure Distributions

Random Variable Distribution Parameter(s)

T, time to failure Exponential l
Yk, time of the kth failure Gamma (Erlang) l, k
N, number of failures in time t Poisson l

18
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 EXAMPLE #3
A specially designed welding machine has a non-repairable motor
with a constant failure rate of 0.05 failures per year. The company
has purchased two spare motors. If the design life of the welding
machine is 10 yr, what is the probability that the two spares will be
adequate?

19
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 Redundancy
2 identical components with CFR
- lt 2
R(t) = 1- (1- e )
- lt -2 lt
= 1-(1- 2 e + e )
= 2 e- lt - e-2 lt
f(t) 2l e- lt - 2l e-2 lt
l (t) = =
R(t) 2 e- lt - e-2 lt
l (1 - e- lt )
= - lt Not a CFR process!
(1-.5 e )
20
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 21
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 Redundancy, CFR & MTTF

  - lt -2 l t
MTTF =  R(t)dt =  (2e - e
0 0
)dt

2 1 3 1.5
= - = =
l 2l 2l l

22
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
Time-Dependent Failure Rate

Weibull’s wobble but they don’t fall down!

-old English saying

The Weibull Distribution

23
 Time Dependent Failure Mode

l (t) = a t b , with a  0
rewrite as:

l (t) =
 FG t IJ  -1
;  > 0,  > 0 ; t  0
 H K
where
 is the shape parameter and  is the characteristic life

24
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 Reliability Function


− 1 t
t   t  − 
 
R(t) = e -0

 
 
dt '
= e

note that:
FI
-G J


R(  ) = e H  K = e-1 = .368

25
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 Graph of the Reliability Function
Beta
1.2

R(t)
0.5
1
1.5

2.0
0.8
4.0

0.6 Theta = 2

0.4

0.2

0
0.0 1.0 2.0 3.0 4.0 5.0 6.0

t
-0.2

26
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 The Probability Density Function (PDF)
1.0
f(t)
0.9

0.8
Beta
0.5
0.7 1.5
2.0
0.6
4.0

0.5

0.4 Theta = 2
0.3

0.2

0.1

0.0 t
0.0 1.0 2.0 3.0 4.0 5.0 6.0
-0.1

 -1
dR(t) 

t t
- 
f(t) = - =   e  
dt   

27
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 The Hazard Rate Function

6
Theta = 2 Beta
0.5
5
f(t) / R(t) 1.5

2.0
4
4.0
Hazard Rate

0 t
0.0 1.0 2.0 3.0 4.0 5.0 6.0

28
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 Weibull Shape Parameter
Value Property
0<<1 Decreasing Failure Rate (DFR)
=1 Exponential Distribution (CFR)
1<<2 IFR-concave
=2 Rayleigh Distribution (LFR)
>2 IFR - Convex
3 <=  <= 4 IFR - Approaches Normal
Distribution - Symmetrical

 = 3.43927 Most closely approximates the normal

 = 3.43938 Mean = median

29
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 The Characteristic Life
1.6
f(t)
1.4
Theta
1.2 0.5

1.0
1

2.0

0.8

Beta = 1.5
0.6

0.4

0.2

0 t
0.0 1.0 2.0 3.0 4.0 5.0 6.0

30
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 The Mean Time to Failure (MTTF)

F 1I
MTTF =   G 1+ J
H K
 (x) = the gamma function = z 
0
y
x-1 -y
e dy
 (x) = (x - 1) (x - 1)

1
lim MTTF = lim   (1 + ) =  (1) = 
 →  → 

31
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 Gamma Function - selected values
x Gamma(x) x Gamma(x)
1.53 .88757 2.23 1.12023
1.54 .88818 2.24 1.12657
1.55 .88887 2.25 1.133
1.56 .88964 2.26 1.13954
1.57 .89049
2.27 1.14618
1.58 .89142
2.28 1.15292
1.59 .89243
1.6 .89352 2.29 1.15976
1.61 .89468 2.3 1.16671
1.62 .89592 2.31 1.17377
1.63 .89724 2.32 1.18093
1.64 .89864 2.33 1.18819
1.67 .9033
2.34 1.19557
1.68 .905
2.35 1.20305
1.69 .90678

32
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 Example Problem
Let T = a random variable, Let T = a random variable,
the time to failure of a the time to failure of a fuse.
circuit card T has a Weibull T has a Weibull distribution
distribution with beta equal with beta equal to 1.5 and
to .5 and theta equal to 500 theta equal to 500
(thousands of hours). (thousands of hours).
Find R(50,000) and the
Find R(50,000) and the MTTF.
MTTF.

33
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 The Variance and Standard Deviation

ìï æ 2ö é æ 1öù
2
üï
s = q í G ç 1+ ÷ - ê G ç 1+ ÷ ú
2 2
ý
ïî è b ø ë è b ø û ïþ

Note: lim  2
=0

→

34
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 Design Life and Median

( )
b
R(t) = e = R
- t
set q

1
then t R =  (- ln R )
1
and t .50 = t med =  (- ln.5 )

35
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 The Mode

f(tmod e ) = Max f(t)

t 0

 1

  1  
  1-  for  > 1
t mod e =   


0 for   1

36
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 Conditional Reliability


 t+T 0 
-   

R(t | T 0 )= e   
=e

-
t+T 0   T 0 
 + 
     
-  T 0 
e  

37
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 Failure Modes
 −1
 t 
n  n  1  
l(t) =    =  t  -1     
i=1  i   i   i=1   i  

which is a Weibull hazard rate function with a

shape parameter of and a characteristic life of
1
 −
 n
1  
   
i=1   i 

 

38
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 Identical Weibull Components

n components have identical hazard rate functions:

 -1
t 
l(t) =  
  
then
 −1
n
t  n
l (t ) =    (t )
 −1
=
i=1    




and t
R (t ) = e − n 
 

39
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 MTTF - Redundant Weibull Components

 - t  -2  t  
 
 
MTTF =  R p (t )dt =   2e    - e     dt
0 0

 
 t   t 
= 2 e − 
  dt −  e −2  
  dt
0 0

 1   1
= 2    1+  − 1/    1+ 
  2  

 1
=    1+   2 - 2 −1/  
 

40
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 Hazard Rate Function
Redundant Weibull Components

 

t
 −1  2 − 2e
− 
t   

 

l s(t) =  
    −  
 t 


 2 − e    

41
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 Derivation of Hazard Rate Function
2 component redundant system

dR(t) t

   t   -1 −2  t    2   t   −1
 −    − e     −
− 
= 2e     
dt            

 −1
t  t

 t


=−   e
− 
 
 2 − 2e
− 
 

  

 −  

t
 −1  2 − 2e    
dR (t ) 1 t  
l p (t ) = −  =  
dt R (t )      −  

 t 
 2 − e   

42
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 Minimum Extreme Value Distribution

 t -t 
- i 0 
Let Ti = failure time of ith component, then R i( t i ) = e   

with T = min{t1, t2, … , tn}

F(t) = P {T  t} = P { t 1 < t or t 2 < t,..or ., t n < t}
= 1 - P { t 1 > t , t 2 > t , ... t n > t}
n
= 1 - P { t 1 > t} P { t 2 > t} ... P { t n > t} = 1- é e ù
b
æ t-t ö
-ç 0 ÷
è q ø
êë úû
 
 t-t   t-t 0 
-n  0 
R(t) = 1- F(t) = e    =e - 
  / n1/  

43
SE-411 (Spring 2020) All Rights Reserved @ P. Wang
 Exercise - Weibull
A certain brand lightning arrester has a Weibull failure
distribution with a shape parameter of 2.4 and a
characteristic life of 10 years. Find:
a. R(5 yrs)
b. MTTF
c. Standard deviation
d. Median and Mode
e. 99% and 95% design life
f. R(5|5)

44
SE-411 (Spring 2020) All Rights Reserved @ P. Wang