Limiting State Probabilities

Matrix Multiplication Method:

λ
1 2
Up Down
µ
1 2
stochastic transitional probability matrix, P = 1 1 - λ.∆
∆t λ.∆
∆t
2 µ.∆ ∆t 1 - µ.∆
∆t

1 2
For limiting state probabilities evaluation, P = 1 1 - λ λ
∆t terms disappear during computation 2 µ 1-µ

limiting state probability vector, α = [P1 P2]

Using the equation αP = α ,

[P1 P2] 1-λ λ = [P1 P2]

µ 1-µ

The two matrix equations are:

P1.(1 - λ) + P2.(µ
µ) = P1 or - λP1 + µP2 = 0
λ) + P2.(1 - µ) = P2
P1.(λ or λP1 - µP2 = 0
If ∆t was used in the P matrix, then the two equations would be:
- λ∆t P1 + µ∆t P2 = 0
λ∆t P1 - µ∆t P2 = 0
since ∆t is finite & non-zero, it is cancelled out resulting:
- λ P1 + µ P2 = 0 The same equations as the case where
λ P1 - µ P2 = 0 ∆t was not used in the P matrix

These are identical equations that can be written as:

µ
P1 = P2 .. Eq. 1
λ

The other equation is P1 + P2 = 1 .. Eq. 2

Solving Equations 1 and 2,

µ λ
P1 = A = P2 = U =
λ +µ λ +µ
 Limiting State Probabilities

System with two identical repairable components

λ
2λ λ
1 2 3
Both 1 Up Both
Up µ 1 Down µ
2µ Down

State Space Diagram

1 2 3
λ 2λ
1 1-2λ λ 0
stochastic transitional P = 2 µ 1-λ λ-µµ λ
probability matrix, 3 0 µ 1-2µ
2µ µ
2 µ.∆∆t 1 - µ.∆
∆t
limiting state probability vector, α = [P1 P2 P3]

Using the equation αP = α ,

λ 2λ
1-2λ λ 0
[P1 P2 P3] µ 1-λλ-µµ λ = [P1 P2 P3]
0 µ 1-2µ
2µ µ
2 µ.∆
∆t 1 - µ.∆
∆t
λ) + P2.(µ
P1.(1 - 2λ µ) = P1 λP1 + µP2 = 0 … Eq. 1
or - 2λ
λ + P3.(1 - 2µ
P2.λ µ) = P3 or λP2 - 2µ µP3 = 0 … Eq. 2
P1 + P2 + P3 = 1 … Eq. 3

Solving Equations 1, 2 and 3,

µ2 2λµ λ2
P1 = , P2 = and P3 =
(λ + µ)2 (λ + µ)2 (λ + µ)2
 Limiting State Probabilities

Alternate Method:

λ
2λ λ
1 2 3
Both 1 Up Both
Up µ 1 Down µ
2µ Down

State Space Diagram

µ
Availability of each component, A=
λ +µ

λ
Unavailability of each component, U=
λ +µ

Using Binomial Expansion,

2 µ2
P1 = P(Both Up) =A =
(λ + µ)2

2λµ
P2 = P(1 Up, 1 Down) = 2AU =
(λ + µ)2

2 λ2
P3 = P(Both Down) =U =
(λ + µ)2
 Limiting State Probabilities

Frequency & Duration

λ
1 1 2
Failure rate, λ =
m Up Down
µ
1
Repair rate, µ = Single Repairable Component
r

µ m m 1/λ f
Pup = = = = =
λ +µ m+r T 1/f λ

where, T = mean time between failures

f = freq of encountering the Up state = Pup . λ
Similarly, f = freq of encountering the Dn state = PDn . µ

In a 2-state system, fup = fDn i.e. Pup . λ = PDn . µ

Frequency of encountering State i

= P(being in State i) x (rate of departure from State i)
= P(not being in State i) x (rate of entry from State i)

This concept can be used to deduce limiting state probabilities

when the transition rates are known
→ Frequency Balance Approach
 Limiting State Probabilities

Frequency Balance Approach

λ
1 2
Up Down
µ

Single Repairable Component

Frequency of encountering State 1

= P(being in State 1) x (rate of departure from State 1)
= P(not being in State 1) x (rate of entry from State 1)
λ = P2.µ
= P1.λ µ .. Eq. 1

Similarly, Frequency of encountering State 2

µ = P1.λ
= P2.µ λ .. Eq. 2

Eq. 1 and 2 are the same equations. The other equation is

P1 + P2 = 1 .. Eq. 3

µ λ
Solving Equations 1 and 3, P1 = and P2 =
λ +µ λ +µ

1
mean duration in State i, mi =
rate of departure from State i

1 1
m1 = MTTF = m2 = MTTR =
λ µ
 Limiting State Probabilities

System with two identical repairable components

λ
2λ λ
1 2 3
Both 1 Up Both
Up µ 1 Down µ
2µ Down

State Space Diagram

Using Frequency Balance Approach:

Frequency of encountering State 1,

λ = P2.µ
P1.2λ µ … Eq. 1
Frequency of encountering State 2,
P2.( µ +λ
λ) = P1.2λ
λ + P3.2µ
µ … Eq. 2
Frequency of encountering State 3,
µ = P2.λ
P3.2µ λ … Eq. 3

Only two of the above 3 equations are independent.

The other equation is, P1 + P2 + P3 = 1 ... Eq. 4

Solving Equations 1, 3 and 4,

µ2 2λµ λ2
P1 = , P2 = and P3 =
(λ + µ)2 (λ + µ)2 (λ + µ)2
 Reliability/Availability Evaluation
in Repairable Systems

λ
2λ λ
1 2 3
Both 1 Up Both
Up µ 1 Down µ
2µ Down

parallel system with 2 identical repairable components

Availability: probability of finding the system in the up state

A = P1 + P2 = 1 – U
λ2
Unavailability, U = P3 =
(λ + µ)2

Reliability:
- probability of staying out of the failed state
- failure states → absorbing states
- time dependent probability
- limiting state reliability = 0
- MTTF (average time before the system enters the
absorbing state) can also be calculated
o differential equations method
o truncated matrix method
 Parallel System with 2 Non-repairable Components

1 λ
2λ 2 λ 3
Both 1 Up Both
Up 1 Down Down

The differential equations are:

λ P1(t)
P’1(t) = - 2λ --- Eq. 1
P’2(t) = λ P1(t) – λ P2(t) --- Eq. 2
2λ
P’3(t) = λ P2(t) --- Eq. 3

Solving the equations using Laplace Transform, we get

P1(t), P2(t) and P3(t)
Rs(t) = P1(t) + P2(t)

For non-repairable components, we can directly apply Product

Rule of Unreliability for a parallel system:
Qs(t) = Q1(t) x Q2(t) = (1 - e − λt ) (1 - e − λt )

Rs(t) = 1- Qs(t) = 2 e − λt - e − 2 λt

∞
3
MTTF = ∫ R s ( t )dt =
0
2λ
 Parallel System with 2 Repairable Components

λ
2λ
1 2 λ 3
Both 1 Up Both
Up µ 1 Down Down

The differential equations are:

λ P1(t) + µ P2(t)
P’1(t) = - 2λ --- Eq. 1
P’2(t) = λ P1(t) – (λ
2λ λ + µ) P2(t) --- Eq. 2
P’3(t) = λ P2(t) --- Eq. 3

Solving the equations using Laplace Transform, we get

P1(t), P2(t) and P3(t)

Rs(t) = P1(t) + P2(t)

∞
3λ + µ
MTTF = ∫ R s ( t )dt = 2λ 2
0
 MTTF using Truncated Matrix Method

In discrete Markov chain:

average # of time intervals before entering the absorbing state is
given by
N = [ I – Q ] -1
where,
I = identity matrix
Q = truncated matrix created by deleting row(s) and column(s)
associated with the absorbing states

In continuous Markov process:

Average time before the system enters the absorbing state (i.e.
MTTF) is given by
M = [ I – Q ] -1
where,
M = termed as the fundamental matrix
mij = average time spent in State j before being absorbed,
given that the process starts in State i
 λ
2λ
1 2 λ 3
Both 1 Up Both
Up µ 1 Down Down

1 2 3
stochastic transitional λ 2λ
1 1-2λ λ 0
probability matrix, P = 2 µ 1-λ λ-µµ λ
3 0 0 1
2 µ.∆∆t 1 - µ.∆
∆t

1 2
λ 2λ
1 1-2λ λ
State 3 is the absorbing state Q = 2 µ 1-λ λ-µ
µ

M = [ I – Q ] –1

1 0 λ 2λ
1-2λ λ 1 λ+µ µ λ
2λ
={ 0 1 – µ 1-λλ-µ
µ
} -1 = 2λ 2 µ λ
2λ

3λ + µ
If the system starts in State 1, MTTF = m11 + m12 = 2
2λ
2λ + µ
If the system starts in State 2, MTTF = m21 + m22 = 2
2λ

If Series System:

λ]
States 2 and 3 are the absorbing states, Q = [ 1-2λ
1
M=[I–Q] –1
λ]
= [ 1 – 1 + 2λ –1
= 2λ