You are on page 1of 10

State dependent model: Markov analysis

State-Dependent Systems: Markov Analysis

Markov analysis
Markov analysis looks at a system being in one of several states, for example all components composing the system are operating. Another possible state is that in which one component has failed but the others component continue to operate. The main assumption in a Markov analysis is that the probability that the system will under goes transition from one state to another depends only on the current state of the system and not on states of the system that the system may have experienced. In other words, transition probability is not dependent on the history of the system This, memorylessness, property is similar to exponential distribution

Two components system


State 1 2 3 4 Component 1 Operating Failed Operating Failed Component 2 Operating Operating Failed Failed

1
2

2
3

2
4

If the two components are parallel (redundant), only state 4 results in system failure. Reliability of the system would be: Rs(t) = P1(t)+ P2(t)+ P3(t); Where is Pi is probability of operation If two components are in series, then states 2, 3, and 4 would result is failure. Reliability of the system would be: Rs(t) = P1(t); Where is P1 is probability of operation

State dependent model: Markov analysis

Considering constant failure rate, the system may be represented as shown in above figures. From the above figure, we can drive

dP (t ) 1 = (1 + 2 ) P (t ) 1 dt dP2 (t ) = 1 P1 (t ) 2 P2 (t ) dt dP3 (t ) = 2 P (t ) 1 P3 (t ) 1 dt

1
2

2
3

2
4

Solving these simultaneous differential equations:

P1 (t ) = e ( 1+2 )t P2 (t ) = e 2t e ( 1+2 )t P3 (t ) = e 1t e ( 1+2 )t P4 (t ) = 1 P (t ) P2 (t ) P3 (t ) 1


Reliability of series system

Rs (t ) = P1 (t ) = e ( 1+2 )t
Reliability of a parallel system

R p (t ) = P1 (t ) + P2 (t ) + P3 (t ) = e 1t + e 2t e ( 1+2 )t

State dependent model: Markov analysis

Load sharing system


Two component are in parallel (redundancy) but have a dependency If one component fails, the failure rate of the other component increase as a result of additional load placed on it The defined states remain the same, however, 1+, 2+ are increased failures rates of component 1 and 2.

dP (t ) 1 = (1 + 2 ) P1 (t ) dt dP2 (t ) = 1 P1 (t ) + P2 (t ) 2 dt dP3 (t ) + = 2 P1 (t ) 1 P3 (t ) dt

1
2

2
3

2+
4

1+

Solving these simultaneous differential equations:

P1 (t ) = e ( 1+2 )t + t 1 2 e ( 1+ 2 )t P2 (t ) = e 1 + 2 + 2
+ 1 t e e ( 1+2 )t P4 (t ) = 1 P (t ) P2 (t ) P3 (t ) 1

2 P3 (t ) = + 1 + 2 1

State dependent model: Markov analysis

Reliability of the system R(t) = P1(t)+ P2(t)+ P3(t) If 1+ = 2+ = + and 1 = 2 =

2 +t e 2 t e 2 + 1 2 1 1 + MTTF = R (t )dt = 2 2 + + 2 0 R (t ) = e 2 t +
Standby system
The standby system differs from the active redundant system in a way that standby system will have no failure or reduce failure while in standby mode. Once standby component become active it may experience the same failure rate as the online system (for identical components) or different failure rate. In other words, failure rate of standby unit depends on the state of the primary unit. Depending on the probability of failure during switching to a standby unit, these system are generally much more reliable than online redundant system. Figure below represents the system, where state 3 represents a failure of the standby system with 2- as failure rate.

State dependent model: Markov analysis

dP (t ) 1 = (1 + ) P1 (t ) 2 dt dP2 (t ) = 1 P1 (t ) 2 P2 (t ) dt dP3 (t ) = P1 (t ) 1 P3 (t ) 2 dt

1
2

23

2
4

Solving these simultaneous differential equations:

P1 (t ) = e P2 (t ) =

( 1 + ) t 2

1 e 2t e ( 1+ )t 2 1 + 2 2

P3 (t ) = e 1t e ( 1+2 )t P4 (t ) = 1 P1 (t ) P2 (t ) P3 (t )
Reliability of the system R(t) = P1(t)+ P2(t)+ P3(t)

R (t ) = e 1t +
0

1 e 2t e ( 1+ )t 2 1 + 2 2
1

MTTF = R (t )dt =

1 2 (1 + ) 2

State dependent model: Markov analysis

Identical Standby Units


Considering there are k identical units of which one is on-line and remaining are standby, when online unit fails, the first standby is placed on and so on. The time in which kth failure will be observed is the sum of k identical and independent exponential distribution. The reliability of k identical system will be:

Rk (t ) = e MTTF =

k 1 i =0

(t )i
i!

Standby system with Switching Failures


The standby system may fail during switching from standby mode to online. Let p is the probability of failure of switching device on demand that switches the standby system to operation. Below figure shows the state of the system. (1-p)1 The resulting equations are:
2 1

23

p1
4

dP1 (t ) = (1 p )1 + p1 + P1 (t ) 2 dt = (1 + ) P (t ) 2 1 dP2 (t ) = (1 p )1 P1 (t ) 2 P2 (t ) dt
Solving these simultaneous differential equations

State dependent model: Markov analysis

P1 (t ) = e ( 1+2 )t (1 p)1 P2 (t ) = 1 + 2 2 P3 (t ) = e
1t

e 2t e ( 1+ )t 2

( 1+ ) t 2

P4 (t ) = 1 P (t ) P2 (t ) P3 (t ) 1
Reliability of the system R(t) = P1(t)+ P2(t)+ P3(t)

R (t ) = e 1t +

(1 p)1 1 + 2 2

e 2t e ( 1+ )t 2

Three components Standby system


Considering a system with one active and two standby units. Considering no units fails while in standby and all the three units have same constant failure rate when on-line. State 1 2 3 4 Unit 1 Operating Failed Failed Failed Unit 2 Standby Operating Failed Failed Unit 3 Standby Standby Operating Failed

State dependent model: Markov analysis 1

dP (t ) 1 = P1 (t ) dt dP2 (t ) = P1 (t ) P2 (t ) dt dP3 (t ) = P2 (t ) P3 (t ) dt

With initial conditions P1(t) = 0, P2(t) = 0, and P3(t) = 0, Solving these differential equations:

P (t ) = e t 1 P2 (t ) = te t

(t )2 e t P (t ) =
3

Reliability of the system R(t) = P1(t)+ P2(t)+ P3(t)

(t )2 R (t ) = e 1 + t + 2 3 MTTF = R (t )dt =
1t 0

As may be evident from the results that it is the same case as discussed in the identical standby unit section. System having three identical units with one online and two standby will have MTTF three times higher than single unit system

State dependent model: Markov analysis

Degraded System
In many situation system may continue to operate in a degraded mode (less than required level) following certain types of failures Such situation may be represented by three states: State 1: Fully operational, State 2: degraded state; State 3: Failed state
1

dP (t ) 1 = (1 + 2 ) P1 (t ) dt dP2 (t ) = 2 P (t ) 3 P2 (t ) 1 dt
The solution is

2
2

1 3
3

P (t ) = e ( 1+2 )t 1 P2 (t ) =

2 e 3t e ( 1+2 )t 1 + 2 3

P3 (t ) = 1 P1 (t ) P2 (t )
Reliability of the system R(t) = P1(t)+ P2(t)

R(t ) = e ( 1+2 )t +
0

2 e 3t e ( 1+2 )t 1 + 2 3

MTTF = R(t )dt =

1 2 1 1 + 1 + 2 1 + 2 3 3 1 + 2

State dependent model: Markov analysis

10

Three State Devices


Components having three states (operating stage, failed open, and failed short) may be analyzed using Markov analysis

The differential equations are:

State 1 Operating

dP (t ) 1 = (1 + 2 ) P1 (t ) dt dP2 (t ) = 1 P1 (t ) dt dP3 (t ) = 2 P1 (t ) dt
The solution is

State 2 Failed open

State 3 Failed short

P1 (t ) = e ( 1+2 )t P2 (t ) =

1 + 2 2 P3 (t ) = 1 e ( 1+2 )t 1 + 2

[1 e

( 1+ 2 ) t

] ]

Reliability of the system

R (t ) = P1 (t ) = e ( 1+2 )t MTTF = R (t )dt =


0

1 1 + 2