Académique Documents
Professionnel Documents
Culture Documents
Definitions
Distribution function:
If FX (x) is a continuous function of x, then X is a
continuous random variable.
FX (x) : grows only by jumps Discrete rv
F (x ) : both jumps and continuous growth Mixed rv
X
Note
Definitions
(Contd.)
Equivalence:
Distribution function
pdf properties:
Definitions
(Continued)
Equivalence: pdf
density function
density
dF
f(t) =
dt
F (t ) =
f ( x ) dx
= f ( x ) dx
0
, for a non-negative
random variable
Example 3.1
Exponential Distribution
Exponential Distribution
Exponential Distribution
Also
10
12500
25000
37500
50000
11
12
Memoryless property
GY ( y | t ) = P(Y y | X > t )
= P( X y + t | X > t )
P (t < X y + t )
=
= 1 e y
P( X > t )
Copyright 2006 by K.S. Trivedi
13
Memoryless property
14
Memoryless property
15
16
and denoting
by constant
17
Example 3.2
Therefore,
18
Example 3.3
distributed
19
20
Definitions
(Contd.)
Equivalence:
Reliability
Survivor function
R(t) = 1 -F(t)
Copyright 2006 by K.S. Trivedi
21
22
(t, t + t].
24
Failure-Time Distributions
Relationships
f(t)
R(t)
h(t)
R(t)
h(t)
t
f(t)
F(t)
F(t)
f (u ) du
F ' (t )
R ' (t )
1 R (t )
h (t ) e 0
t
h ( u ) du
1 e 0
h ( u ) du
f (u ) du
f (t )
1 F (t )
d
F ' (t )
log e R (t )
dt
f (u ) du (1 F (t ))
Copyright 2006 by K.S. Trivedi
e 0
h ( u ) du
1
25
Bathtub curve
DFR phase: Initial design, constant bug fixes
CFR phase: Normal operational phase
IFR phase: Aging behavior
h(t)
(burn-in-period)
(wear-out-phase)
CFR
(useful life)
DFR
IFR
t
Decreasing failure rate
26
Weibull Distribution
F (t ) = 1 e
t 0
Reliability: R(t ) = e
Weibull distribution is capable of modeling DFR ( < 1),
CFR ( = 1) and IFR ( >1) behavior.
is called the shape parameter and is the scale
parameter.
t
27
F (t ) = 1 e
( t )
R(t ) = e
t0
Reliability:
In this text we will use the definition on the previous slide
( t )
28
Example 3.4
29
5.0
1.0
2.0
3.0
4.0
30
31
Bathtub curves
Early-life period
Steady-state period
32
Early-life Period
Also called infant mortality phase or reliability
growth phase or decreasing failure rate (DFR
phase).
Caused by undetected hardware/software
defects that are being fixed resulting in
reliability growth.
Can cause significant prediction errors if
steady-state failure rates are used.
Availability models can be constructed and
solved to include this effect.
DFR Weibull Model can be used.
33
Steady-state Period
Failure rate much lower than in early-life
period.
Either constant (CFR) (age independent)
or slowly varying failure rate.
Failures caused by environmental shocks.
Arrival process of environmental shocks
can be assumed to be a Poisson process.
Hence time between two shocks has
exponential distribution.
34
35
Failure-Rate Multiplier
7
6
5
4
3
2
1
0
2,190
4,380
36
1 t 8,760
hW (t ) = C1 t
t > 8,760
= h SS
where:
C1 = hW (1), h SS =
37
Failure-Rate Multiplier
7
6
h1
5
4
3
h2
h SS
1
0
2,190
4,380
38
39
HypoExponential (HYPO)
40
Hypo(1,5)
Hypo(1,5) pdf
Hypo(1,5) CDF
41
Failure
probable
state
Failed state
aging
42
Erlang Distribution
43
Erlang Distribution
44
45
46
47
48
49
50
51
52
Log-logistic Distribution
53
54
55
56
57
58
59
60
f(x)=f(-x)
F(-z) = 1-F(z).
Copyright 2006 by K.S. Trivedi
61
Example 3.5
62
63
Example 3.6
64
Similarly,
65
66
67
Uniform distribution
x a
F ( x) =
,
b a
1,
x<a
a x<b
bx
68
Uniform CDF
69
70
Comparison of
probability density functions (pdf)
1.8
1.6
1.4
1.2
1
pdf
0.8
0.6
0.4
0.2
1.
02
0.
96
0.
9
0.
84
0.
78
0.
72
0.
66
0.
6
0.
54
0.
48
0.
42
0.
36
0.
3
0.
24
0.
18
0.
12
0.
06
time
71
Comparison of cumulative
distribution functions (cdf)
1.2
0.6
U(0,1) cdf
0.4
0.2
1.
02
0.
9
0.
96
0.
84
0.
78
0.
72
0.
6
0.
66
0.
48
0.
54
0.
42
0.
3
0.
36
0.
24
0.
18
0.
12
0
0
0.
06
cdf
0.8
time
72
Pareto Distribution
73
Pareto Distribution
f ( x) = k x
(Contd.)
, x k, , k > 0
74
75
76
Log-normal
77
Defective Distribution
Recall that
Example:
If p c < 1, then , FX ( x ) = p c (1 e
) is a
78
79
80
Example 3.8
81
Example 3.9
82
Example 3.10
83
84
85
86
Inverse Transform
Method:
87
Example 3.11
88
Inverse Transform
89
Some Examples
Exponential Distribution
CDF
F ( x) = 1 e x , x 0
Random Variate x
ln(1 u )
x=
x=
ln(u )
90
Exponential variate
For 10 Observations
1.2
1.2
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0.2
2 3
4 5 6
7 8
9 10 11 12 13 14 15 16
10 11 12 13 14 15 16
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0.2
11
13
15
2 3
4 5 6
7 8
9 10 11 12 13 14 15 16
92
Weibull Distribution
CDF
FX ( x ) = 1 e
ln(1 u )
x=
93
Write down the random deviate formual for the alternate for
of the Weibull Distribution
CDF
FX ( x) = 1 e
( x )
CDF
94
Pareto Distribution
CDF
FX ( x ) = 1
xk
x
=0
xk
where k>0 is location parameter and is shape
parameter.
Random Variate x
k
x=
(1 u )
95
Rayleigh Distribution
x 2 / 2 2
x0
CDF FX ( x ) = 1 e
=0
otherwise
where 2 is the variance.
Random Variate x
x = 2 2 ln(1 u )
96
Log-Logistic Distribution
CDF
FX ( x) = 1
1
1 + ( x )
Random Variate x
1 u
x=
u
1
97
Density
Expo
e x x > 0
f(x)
1 e x x > 0
1 e x
Weibull
e x x > 0
Pareto
k x 1 x > k ,
,k > 0
Rayleigh
1 x 2
x 2 / 2 2
exp
e
2
2
x
X=F-1(u) Simpl.
F(x)
k
1
x
x=
form
ln(1 u )
x=
ln(1 u )
x=
ln(u )
1
1
ln(u )
x =
1
k
x= 1
u
2
x 0 x = 2 ln(1 u )
x = 2 2 ln(u )
x=
k
(1 u )
98
Name
Density
LogLogistic
Cauchy
1
(t )
1
t0
1 + ( t )
[1 + (t ) ]2
Triangul
ar
f(x)
(x + )
2
2 x
1
a a
1 1
x
+ arctan
2
2
x2
x
a
2a
X=F-1(u) Simplifie
d form
1 u
x=
1 u
x=
1 u
1 u
tan (u )
tan (u )
a (1 1 u )
a (1 u )
1
2
99
and distribution F ( x j ) = pi
i =0
This is as
x = x j , if F ( x j 1 ) < u F ( x j )
j 1
i =1
i =1
P( X = x j ) = P ( pi < U pi ) = p j
Copyright 2006 by K.S. Trivedi
100
Bernoulli distribution
CDF with parameter (1-q)
FX ( x ) = 0 x < 0
= q 0 x <1
=1 x 1
101
p (1 p ) j 1 j 1
p) j
1 (1 p ) j 1 < u 1 (1 p ) j
(1 p ) j 1 < 1 u (1 p ) j
= min{ i : i >
log( 1 u )
}
log( 1 p )
log( 1 u )
=
log(
1
p
)
log(u )
log(1 p )
102
Example 3.13
103
Example 3.14
104
105
106
107
108
Order Statistics: k of n
To be specific:
109
Order Statistics: k of n
(Continued)
110
Order Statistics: k of n
(Continued)
n j
F ( y ) [1 F ( y )]n j
j
< y <
hence
n j
FYk ( y ) = F ( y ) [1 F ( y )]n j , < y <
j =k j
n
111
Overview:
General iid Random Variables
Let Y1 ,,..., Yn denote the order statistics of the random variables X1 ,..., Xn,,
which are iid with common distribution function F.
Y1
Yn k +1
FYn k +1 ( y )
n
n j
= F ( y )[1 F ( y )]n j
j = n k +1 j
Rk |n (t ) = 1 FYn k +1 (t )
n
n j
= R (t )[1 R(t )]n j
j =k j
Yn
Copyright 2006 by K.S. Trivedi
112
Overview:
General iid Random Variables
Let Y1 ,,..., Yn denote the order statistics of the random variables X1 ,..., Xn,,
which are iid with common distribution function F.
Y1
Yn k +1
Yn
FYn k +1 ( y )
n
n j
= F ( y )[1 F ( y )]n j
j = n k +1 j
FYn ( y ) = [ F ( y )]
Rk |n (t ) = 1 FYn k +1 (t )
n
n j
= R (t )[1 R(t )]n j
j =k j
R parallel (t ) = 1 FYn (t )
= 1 [ F (t )]n = 1 [1 R(t )]n
113
Overview:
General iid Random Variables
Let Y1 ,,..., Yn denote the order statistics of the random variables X1 ,..., Xn,,
which are iid with common distribution function F.
Y1
Yn k +1
Yn
FY1 ( y ) = 1 [1 F ( y )]
FYn k +1 ( y )
n
n j
= F ( y )[1 F ( y )]n j
j = n k +1 j
FYn ( y ) = [ F ( y )]
Rseries (t ) = 1 FY1 (t )
= [1 F (t )]n = [ R(t )]n
Rk |n (t ) = 1 FYn k +1 (t )
n
n j
= R (t )[1 R(t )]n j
j =k j
R parallel (t ) = 1 FYn (t )
= 1 [ F (t )]n = 1 [1 R(t )]n
114
Series system:
Rseries (t ) = [ R(t )]
or Ri (t )
i =1
Parallel system:
i =1
115
Example 3.16
116
117
Example 3.18
118
Example 3.20
119
Voter
R(t)
RTMR (t ) = 3R (t ) 2 R (t )
2
120
TMR
(Continued)
R (t ) = e
we get:
RTMR (t ) = 3e
2 t
2e
3t
121
TMR
(Continued)
122
TMR
(Continued)
123
124
125
126
127
128
129
130
131
132
133
134
135
A file-server
Two workstations
Computing network connecting them
Copyright 2006 by K.S. Trivedi
136
Workstation 2
137
138
139
bind
lambdaW 0.0001
lambdaF 0.0003
end
block wfs1
* each component is non-restorable and has exp time to fail dist
comp Workstation exp(lambdaW)
comp FileServer exp(lambdaF)
parallel work Workstation Workstation
series sys work FileServer
end
* define function R(t) for reliability at time t
func R(t) 1-tvalue(t;wfs1)
* vary the time t from t=0 to 10000 in steps of 1000 hours and print R(t)
loop t,0,10000,1000
expr R(t)
end
Copyright 2006 by K.S. Trivedi
140
end
1.2
1
Reliability
0.8
0.6
exp
0.4
0.2
0
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000
tim e
141
Example 3.21
142
143
Description
Reliability:
R (t ) = [1 (1 Rc (t )) ][1 (1 Rv (t )) ]
2
144
145
146
Output selection
147
148
Plot definition
149
150
151
152
and
c1
c2
v1
= c1 c2 + v1 v2 v3
v2 v3
153
Assume Rc (t ) = e
c t
and Rv (t ) = e
v t
ct
= (2e
2ct
vt
)(3e
2vt
3e
3vt
+e
154
155
Example 3.22
156
157
Example 3.23
158
159
Y1
= 1 exp( ny )
= exp( nt )
Yn k +1
FYn k +1 ( y )
n
n
= [1 exp( y )] j
j = n k +1 j
exp( ( n j )y )
Rk |n (t ) = 1 FYn k +1 (t )
Yn
FYn ( y ) = [1 exp( y )]
Y1 EXP( n )
R parallel (t ) = 1 FYn (t )
= 1 [1 exp( t )]n
160
Y1
= 1 exp( ny )
= exp( nt )
Y1 EXP( n )
Yn k +1
FYn k +1 ( y )
n
n
= [1 exp( y )] j
j = n k +1 j
exp( ( n j )y )
Rk |n (t ) = 1 FYn k +1 (t )
Yn
FYn ( y ) = [1 exp( y )]
Yn k +1