Michael E.

Odigie, PhD, CMQ/OE

Chapter 6
State Dependent Systems

Markov Analysis
Load Sharing Systems
Standby Systems
Degraded Systems

Chapter 6 1
Michael E. Odigie, PhD, CMQ/OE

State Dependent Systems

 We have been assuming independence between

components
 Failure of one component may affect the time to failure of
another component
- Failure of one aircraft engine increases load on the others
 Markov Analysis
- System exists in one of several states
- Probability of transitioning from one state to the other depends only on
current state, not past states
- Memoryless property
- Assume process is stationary (transition probabilities do not change over
time)
- Exponential times to failure (exponential)
Chapter 6 2
Michael E. Odigie, PhD, CMQ/OE

Markov Analysis
Independent Components

System Component 1 Component 2 Series Parallel

State System System
1 Operating Operating Operating Operating
2 Failed Operating Failed Operating
3 Operating Failed Failed Operating
4 Failed Failed Failed Failed

Chapter 6 3
Michael E. Odigie, PhD, CMQ/OE

Markov Analysis

Let Pi(t) = probability of being in state i at time t

P1 (t)+P2 (t)+P3 (t)+P4 (t) = 1

For two components in series:

R S (t) = P1 (t)

For two components in parallel:

R P (t) = P1 (t)+P2 (t)+P3 (t)
Chapter 6 4
Michael E. Odigie, PhD, CMQ/OE

Markov Analysis

 Assume components have constant failure rates

 Generate a Markov transition rate diagram

 States are numbered blocks, arrows are transitions

 Rates are given by constant values of λ

Chapter 6 5
Michael E. Odigie, PhD, CMQ/OE

Markov Analysis

rate out of state 1 rate out of state 1

into state 2 into state 3

Probability of being in state 1 at time t+∆t is equal to the

probability of being in state 1 at time t and not transitioning
to states 2 or 3 in time ∆t.
Chapter 6 6
Michael E. Odigie, PhD, CMQ/OE

Markov Analysis

P1(t+∆t) probability of being in state 1 at time t+∆t

λ1∆t conditional probability of transition to state 2

during ∆t given that the system in state 1

λ1∆t P1(t) joint probability of being in state 1 at time t and

then making a transition to state 2 during ∆t

λ2∆t P1(t) joint probability of being in state 1 at time t and

then making a transition to state 3 during ∆t

Chapter 6 7
Michael E. Odigie, PhD, CMQ/OE

State Equations
State 1:
P1 (t + ∆t) = P1 (t) − λ1 ∆tP1 (t) − λ2 ∆tP1 (t)
State 2:

P2 (t + ∆t) = P2 (t) + 1 ∆tP1 (t) − 2 ∆tP2 (t)

State 3:

P3 (t + ∆t) = P3 (t) + λ2 ∆tP1 (t) − λ1 ∆tP3 (t)

State 4:
P4 (t + ∆t) = P4 (t) + λ2 ∆tP2 (t) + λ1 ∆tP3 (t)
Chapter 6 8
Michael E. Odigie, PhD, CMQ/OE

Rewriting the Equation for State 1:

P1 (t + ∆t) = P1 (t) − λ1∆tP1 (t) − λ2 ∆tP1 (t)

P1 (t + ∆t) − P1 (t)
= −(λ1 + λ2 )P1(t)
∆t

 P1 (t + ∆t) − P1 (t)  dP1 (t)

lim  =
∆t →0 ∆t  dt

dP1 (t)
= −(λ + λ )P (t) Differential equation
1 2 1
Chapter 6
dt 9
Michael E. Odigie, PhD, CMQ/OE

Differential Equations for States 2 & 3

dP1 (t)
= −(λ + λ )P (t)
1 2 1
dt
dP2 (t)
= λ P (t) − λ P (t)
1 1 2 2
dt
dP3 (t)
= λ P (t) − λ P (t)
2 1 1 3
dt
P1 (t) + P2 (t) + P3 (t) + P4 (t) = 1

A fourth differential equation is not needed. Why?

Chapter 6 10
Michael E. Odigie, PhD, CMQ/OE

Solution

− ( λ1 + λ2 )t
P1 (t) = e
− λ2 t
P2 (t) = e −e −( λ1 + λ2 )t

− λ1t
P3 (t) = e −e −( λ1 + λ2 )t

P4 (t) = 1 − e [ − λ1t
+e − λ2 t
−e −( λ1 + λ2 )t
]
Chapter 6 11
Michael E. Odigie, PhD, CMQ/OE

Series and Parallel Systems

Series System Parallel System

R S (t) = P1 (t) R P (t) = P1 (t) + P2 (t) + P3 (t)
R S (t) = e −( λ1 + λ2 )t R P (t) = e −λ1t + e −λ2t − e −( λ1 + λ2 )t
R S (t) = (e − λ1t )(e − λ2 t ) R P (t) = R1 (t) + R2 (t) − R1 (t)R2 (t)
R S (t) = R1 (t)R 2 (t ) same as previous result!
same as previous result!

That was a long way to

go for a simple result
Chapter 6
we already knew! 12
Michael E. Odigie, PhD, CMQ/OE

Load-Sharing System

 Consider a 2-component load-sharing system

 True dependency
 If one component fails, the failure rate of the
other component increases as a result of the
additional load placed on it.

 Block diagram approaches will not work!

Chapter 6 13
Michael E. Odigie, PhD, CMQ/OE

Load-Sharing System
Assume Constant Failure Rates
State Operating
1 Both components
2 Component 2 (Comp 1 failed)
3 Component 1 (Comp 2 failed)
4 Neither component

λ1 failure rate of component 1

λ2 failure rate of component 2
λ1+ increased failure rate of component 1
λ2+ increased failure rate of component 2

Chapter 6 14
Michael E. Odigie, PhD, CMQ/OE

Load-Sharing System
dP1 (t)
= −(λ + λ )P (t)
1 2 1
dt
dP2 (t)
= λ P (t) − λ+ P (t)
1 1 2 2
dt
dP3 (t)
= λ P (t) − λ+ P (t)
2 1 1 3
dt
dP4 (t)
= λ+ P (t) + λ+ P (t)
2 2 1 3
dt

Chapter 6
P1 (t) + P2 (t) + P3 (t) + P4 (t) = 1 15
Michael E. Odigie, PhD, CMQ/OE

Load-Sharing System

−( λ1 + λ2 )t
P (t) = e
1

P2 (t) =
λ1
λ1 + λ2 − λ+2 e
−λ t
−[e
+
−( λ + λ )t
2
1 2
]
P3 (t) =
λ2
λ1 + λ 2 − λ1
+ e
−λ t
− [
e
+
−( λ + λ )t
1
1 2
]
RS (t) = P1 (t) + P2 (t) + P3 (t)
Chapter 6 16
Michael E. Odigie, PhD, CMQ/OE

Load-Sharing System vs. Parallel System

λ1=0.0034 Assume components λ1=0.0034 Assume a component’s
are completely 1
rate of failure increases
1
independent by 50% if the other fails

2 2 λ1+ =0.0051

λ2=0.0105 λ2=0.0105 λ2+ =0.01575

R(100) = ? R(100 ) = ?
use standard equation from before P1 (100 ) = e −( 0.0034 + 0.0105 )(100 ) = 0.2491
other equations yield :
= 1− (1− e −0.0034(100) )(1− e −0.0105(100) )
P2 (100 ) = 0.077
= 0.8126
P3 (100 ) = 0.4192
R(100 ) = P1 (100 ) + P2 (100 ) + P3 (100 )
Chapter 6
R(100 ) = 0.745 17
Michael E. Odigie, PhD, CMQ/OE

Load-Sharing System
(with Identical Components)

If we let λ1 = λ2 = λ and λ1 + = λ + = λ+ , then

2

2λ
− λ [e ]
−2 λt
R(t) = e +
2
+ − λ+ t
− e −2 λt

∞
1 2λ 1 − 1 
MTTF = R(t)dt = +
∫
0 2λ 2λ − λ+  λ+ 2λ 

Chapter 6 18
Michael E. Odigie, PhD, CMQ/OE

Load-Sharing Systems - Example 6.1

Two generators provide electrical power. If either fails, the

other will continue to provide power. However, the increased
load results in a higher failure rate for the remaining
generator. If λ=0.01 failure per day (the original failure rate)
and λ+=0.10 failure per day (the increased failure rate),
determine the system reliability for a 10-day contingency
operation and determine the system MTTF.

Chapter 6 19
Michael E. Odigie, PhD, CMQ/OE

Load-Sharing Systems - Example 6.1

2(0.01)
R(t) = e −2(0.01)t +
2(0.01) − 0.10 e [
−0.10t
− e −2(0.01)t
]
R(10) e −0.2 + [
0.02 −1
e e −0.2] = 0.9314
− 0.08

1 0.02  1 1  = 60 days
MTTF = + − 0.08  0.01 − 0.02 
0.02

Chapter 6 20
Michael E. Odigie, PhD, CMQ/OE

Load-Sharing Systems - Exercise 6.1

Two nickel-cadmium batteries provide electrical power to

operate a satellite transceiver. If both batteries are operating
in parallel, they have an individual failure rate of 0.1 per
year. If one fails, the other can operate the transceiver (at a
reduced power output). However, the increased electrical
demand will triple the failure rate of the remaining battery.
Determine the system reliability at 1, 2, 3, 4, and 5 years.
What is the system MTTF?
Chapter 6 21
Michael E. Odigie, PhD, CMQ/OE

Exercise 6.1 - Solution

2λ
R(t) = e −2λt +
2λ − λ [
+ e − λ t − e −2 λt
+
]
0.2
=e −0.2t +
0.2 − 0.3 e [
−0.3t
− e −0.2t
]= e −0.2t
− 2 e[−0.3t
− e −0.2t
]
R(1) = 0.9745
R(2) = 0.9133
R(3) = 0.8333
R(4) = 0.7456
R(5) = 0.6574
1 0.2  1 − 1  = 8.33 years
MTTF = 0.2 + 0.2 − 0.3  0.3 0.2 
 
Chapter 6 22
Michael E. Odigie, PhD, CMQ/OE

Standby Systems

 Many models for primary/backup components

 Backup may have no or reduced failure rate
while in standby mode
 Once activated, backup may have different
failure rate than primary
 Can also model switching failure from primary to
backup
 Later will consider repairs
Chapter 6 23
Michael E. Odigie, PhD, CMQ/OE

Standby Systems

State Operating
1 Component 1 with component 2 in standby
2 Component 2
3 Component 1 with component 2 failed in standby
4 Neither component

λ1 failure rate of primary

λ2- failure rate of backup in standby
λ2 failure rate of backup while activated

Chapter 6 24
Michael E. Odigie, PhD, CMQ/OE

Standby Systems

dP1 (t)
= −( λ1 + λ −
)P (t)
2 1
dt
dP2 (t)
= λ1 P1 (t) − λ 2 P2 (t)
dt
dP3 (t)
= λ −2 P2 (t) − λ1 P3 (t)
dt
dP4 (t)
= λ 2 P2 (t) + λ1 P3 (t)
dt
P1 (t) + P2 (t) + P3 (t) + P4 (t)= 1
Chapter 6 25
Michael E. Odigie, PhD, CMQ/OE

Standby Systems
− ( λ1 + λ −2 )t
P1 (t ) = e
λ1
P2 (t) =
λ1 + λ 2 − λ 2
− e −λ t
−e
−(λ +λ
[ 2 1
−
2 )t
]
− λ1 t −( λ1 + λ −2 )t
P3 (t ) = e −e
R S (t ) = P1 (t) + P2 (t ) + P3 (t)

R S (t) = e +
λ1
− λ1t e −λ t
− e −( λ + λ
[ 2 1
−
2 )t
]
λ1 + λ −2 − λ 2
1 λ1
MTTF = +
λ1 λ 2 ( λ1 + λ 2 )
−
Chapter 6 26
Michael E. Odigie, PhD, CMQ/OE

Standby Systems
(with Identical Components)

If we let λ1 = λ2 = λ and λ2− = λ− , then

RS (t) = e − λt

λ
λ [− λt
+ − e − e −( λ +λ
− )t
]

MTTF = + λ 1
λ λ + λ−
Chapter 6 27
Michael E. Odigie, PhD, CMQ/OE

Standby Systems - Example 6.2

An active generator has a failure rate (failures per day) of

0.01. An older standby generator has a failure rate of 0.001
while in standby and a failure rate of 0.10 when on-line.
Determine the system reliability for a planned 30 day usage
and compute the system MTTF.

λ1 = 0.01 λ2 = 0.10 λ−2 = 0.001

0.01
RS (30) = e −(0.01)30
+
(0.01+ 0.001− 0.10) e [
−0.10(30)
− e ]
−(0.011)30
= 0.8162
1 0.01
MTTF = + = 109.09 days
0.01 0.1(0.10 + 0.001)
Chapter 6 28
Michael E. Odigie, PhD, CMQ/OE

Standby Systems - Example 6.3

Both units of a two-component standby system are identical

with λ = 0.002 failures per hour and λ− = 0.0001 failures
per hour. Determine the design life on a basis of a 95
percent reliability.

0.95 = R(30) = e −0.002t

+
0.0001
[
0.002 −0.002t −0.0021t
e −e ]
by trial and error R(100) = 0.982
or using Goal Seek in Excel R(200) = 0.935
R(150) = 0.961
R(175) = 0.949
Chapter 6 R(173) = 0.950 29
Michael E. Odigie, PhD, CMQ/OE

Standby Systems - Identical Units

 Assume k identical units with CFR - a primary and k-1

units in standby with no failures in standby.
 The system fails when the kth failure occurs. The time to
kth failure is the sum of k exponential failure times:
Tfailure = T1 + T2 + . . . + Tk
 The sum of k independent and identically distributed
exponential random variables with parameter λ is
gamma with parameters λ and k and MTTF = k/λ.
k −1
(λt)
i

Chapter 6
R k (t) = e
− λt
∑ i! 30
i=0
Michael E. Odigie, PhD, CMQ/OE

Identical Units - Example 6.4

The Rey Lie Able Printing company has four presses: one
operating and three in standby. Each press has an identical
failure rate where MTTF is 50 operating hours. The
company has received a rush order requiring 75 hours of
continuous time on press. If a standby is utilized whenever
the on-line press fails, determine the probability of there
being continuous printing support while the order is being
processed. (similar to spare part analysis!)
1
λ= and k = 4
50
3
(75 / 50) i  3 9 27 
R 4 (75) = e −75 / 50
∑ i!
=e −1.5
1 + + +  = 0.9344
 2 8 48 
i =0

MTTF = 4 = 200 hours

31
(1/ 50)
Michael E. Odigie, PhD, CMQ/OE

Standby Systems with Switching Failures

1
(1 − p ) λ1 λ −
2

2 p λ1 3

λ1
λ2 4

Allows for a probability (p) that switch from primary

to backup does not work!
Chapter 6 32
Michael E. Odigie, PhD, CMQ/OE

Standby Systems with Switching Failures

dP1 (t)
= −[(1 − p)λ1 + pλ1 + λ−2 ]P1 (t) = −(λ1 + λ− )P (t)
2 1
dt
dP2 (t)
= (1 − p)λ1 P1 (t) − λ 2 P2 (t)
dt
dP3 (t)
= λ−2 P1 (t) − λ1 P3 (t)
dt
dP4 (t)
= pλ1 P1 (t) + λ 2 P2 (t) + λ1 P3 (t)
dt
P1 (t) + P2 (t) + P3 (t) + P4 (t) = 1
Chapter 6 33
Michael E. Odigie, PhD, CMQ/OE

Standby Systems with Switching Failures

−( λ1 + λ−2 )t
P1 (t) = e
(1 − p)λ1
P2 (t) = λ + λ− − λ −λ2t
1 2 2
e −e
−( λ1 + λ−2 )t
[ ]
− λ2t −( λ1 + λ−2 )t
P3 (t) = e −e
RS (t) = P1 (t) + P2 (t) + P3 (t)

RS (t) = e
− λ1t (1 − p)λ1
+ λ + λ− − λ e
1
− λ2t

2
−e
2
[
−( λ1 + λ−2 )t
]
Chapter 6 Only Change 34
Michael E. Odigie, PhD, CMQ/OE

Switching Failures - Example 6.5

Consider the standby system described in example 6.2. If

there is a 10 percent probability of a switching failure, the
system reliability becomes:

(0.90)0.01
R(30) = e −(0.01)30
+
(0.01+ 0.001− 0.10) e [
−0.10(30)
− e −(0.011)30
]

= 0.741+
(0.90)(0.01)
[0.04978 − 0.7189]= 0.8087
0.01+ 0.001− 0.1
Chapter 6 Slight decrease from the perfect switching case! 35
Michael E. Odigie, PhD, CMQ/OE

Three-Component System

dP1 (t)
State Unit 1 Unit 2 Unit 3 = −λP1(t)
1 on-line standby standby dt
2 failed on-line standby
3 failed failed on-line
dP2 (t)
4 failed failed failed = λP1(t) − λP2 (t)
dt

dP3 (t)
= λP2 (t) − λP3(t)
dt
Chapter 6 36
Michael E. Odigie, PhD, CMQ/OE

Three-Component System

P1 (t ) = e − λt
P2 (t ) = λ te − λt
P (t ) =
λ 2t 2

3 e − λt
2
R S (t ) = P1 (t ) + P2 (t) + P3 (t )
− λt  λ2 t 2 
R S (t ) = e 1 + λ t + 
 2 
3
MTTF =
Chapter 6
λ 37
Michael E. Odigie, PhD, CMQ/OE

Three-Component System - Example 6.6

Three identical transmitters are available each having a

constant failure rate of 0.0035 per operating hour. A

mission requires 500 hours of continuous transmission.
Determine the reliability of the system.

−(0.0035)500  (0.0035× 500) 2


R(500) = e + 1+ (0.0035× 500) + = 0.744
 2 

3
MTTF = = 857.14 hours
0.0035

Chapter 6 38
Michael E. Odigie, PhD, CMQ/OE

Degraded Systems
The system may continue to perform its function but not at a
specified operating level.
State 1: fully operational 1
State 2: degraded λ2 λ1
State 3: failed
dP1 (t )
= − ( λ1 + λ 2 ) P1 (t ) 2 3
dt
dP2 (t ) λ3
= λ 2 P1 (t ) − λ 3 P2 (t )
dt
dP3 (t )
= λ1 P1 (t ) + λ 3 P2 (t )
dt
Chapter 6 39
Michael E. Odigie, PhD, CMQ/OE

Degraded Systems

P1 (t ) = e − ( λ1 + λ 2 )t
λ2
P2 (t ) =
λ1 + λ2 − λ3 [
e −λ t
− e −(λ
3 1 + λ 2 )t
]
P3 (t ) = 1 − P1 (t ) − P2 (t )

1 λ2 1 1 
MTTF = +  λ3 − λ + λ 
λ1 + λ2 λ1 + λ2 − λ3 1 2 

Chapter 6 40
Michael E. Odigie, PhD, CMQ/OE

Degraded Systems - Example 6.7

A machine used in a manufacturing process experiences

complete failures at a constant rate of 0.01 per day.
However, the machine may degrade randomly, producing
substandard parts (out of tolerances) at a constant rate of
0.05 per day. Once it has degraded, it will fail completely at
a constant rate of 0.07 per day. Determine its reliability
while it is fully operational and while it is degraded.

Chapter 6 41
Michael E. Odigie, PhD, CMQ/OE

Example 6.7 - Solution

P1 (t) = e −(0.01+0.05)t = e −0.06t
0.05
P2 (t) =
0.01 + 0.05 − 0.07 e[−0.07 t
− e −(0.01+0.05)t
]
P1 (1) = e −0.06 = 0.942
0.05
P2 (1) =
0.01 + 0.05 − 0.07 e
−0.07
[
− e −(0.01+0.05) = 0.047 ]
P3 (1) = 1 − 0.942 − 0.047 = 0.011
1 0.05  1 1  = 28.6 days
MTTF = + 0.01 + 0.05 − 0.07  0.07 − 0.01 + 0.05 
0.01 + 0.05  
1
MTTF degraded = = 14.3 days
0.07
1
MTTF state1(fully operationa l) = = 16.67 days
0.01 + 0.05 42