Reliability Modeling
Decomposition Method
In some situations we encounter systems, such as the one shown in the figure below, which cannot
be solved by the reduction method. The decomposition method also known as the conditional
probability approach or Bayes’ theorem method can be used in such cases.
1 2
3 4
Decomposition Method
1 2
5 A
3 4
Decomposition Method
1 2
5 A
3 4
Decomposition Method
Example: Consider the same bridge system in figure below. Suppose that the reliability of component i is Ri , i = 1,
2, . . . , 5. Calculate the system reliability.
1 2
5 A
3 4
Solution: Component 5 is chosen as the keystone component, denoted A. Assume that it never fails and is
replaced with a line in the system configuration.
1 2
Pr(system good | A) = [1 − (1 − R1)(1 − R3)][1 − (1 − R2)(1 − R4)].
A
3 4
Decomposition Method
The next step is to assume that component 5 has failed and is removed from the system structure. Figure below shows the
new configuration, which is a parallel–series system. The conditional reliability is
1 2
Pr(system good | A) = 1 − (1 − R1.R2) (1 − R3.R4).
3 4
If the reliability and unreliability of component 5 are Pr(A) = R5 and Pr(A) = 1−R5, respectively,
R = [1 − (1 − R1) (1 − R3)] [1 − (1 − R2)(1 − R4)] R5 + [1 − (1 − R1R2) (1 − R3.R4)] (1 − R5)
Decomposition Method
For some complex systems, we may select additional keystone components and apply the earlier approach
successively until each term in the equation is easily obtainable.
For example, if two keystone components, say A and B, are selected, the original system will be decomposed into
four subsystems with conditions
A · B, A · B, A · B, and A · B, respectively,
where,
A · B is the event that both A and B are functioning,
A · B is the event that A is not functioning and B is,
A · B is the event that A is functioning and B is not, and
A · B is the event that both A and B are not functioning.
Decomposition Method
By applying the rule of total probability, the reliability of the original system can be written as
R = Pr(system good |A · B) Pr(A) Pr(B) + Pr(system good |A · B) Pr(A) Pr(B) +
Pr(system good |A · B) Pr(A) Pr(B) + Pr(system good |A · B) Pr(A) Pr(B)
Load Strength Interference
Shock Based Reliability Models
Shock models in system reliability are usually defined by the time between two
consecutive shocks, the damage caused by a shock, the criteria for system failure.
Shock Based Reliability Models
Four principal models are considered:
1. Basic shock model, where the system fails when the number of shocks exceed a given
number.
2. Extreme shock model, where the system breaks down as soon as the magnitude of an
individual shock exceeds some given level.
3. Cumulative shock model, where the system fails when the cumulative shock magnitude
exceeds some given level.
4. Run shock model, where the system works until ‘k’ consecutive shocks with critical
magnitude occur.
Shock Based Reliability Models
Consider a component, which fails if the number of shocks, N(t), exceed 3.
Determine its reliability at 500 hrs. The time between shocks is exponentially distributed
with λ=0.001.
( t ) x t
n
R(t ) Pr( N (t ) n ) e
0 x!
(500 ) x 500
3
R(500) Pr( N (500) 3) e
0 x!
Shock Based + Load-Strength Reliability Models
Consider a component, which fails if the number of shocks exceeds 2 or the magnitude of
a shock exceeds ‘5’.
The time between shocks f(T) is exponentially distributed with λ=0.001.
The magnitude of shock, M, is a continuous random variable with cdf Ф(M).
Determine the component reliability at 500 hrs.
R(t ) Pr[ N (t ) 0] Pr[ N (t ) 1, M m)] Pr[ N (t ) 2, M1 m, M 2 m)
( t ) x t
n
R ( t ) e t e .[ ( m)]x
1 x!
[0.001(500)]x 0.001( 500 )
2
R(500) e 0.001( 500 )
e .[ (5)]x
1 x!
Assume shock magnitude to follow exponential distribution with a mean of 3
Shock Based + Load-Strength Reliability Models
Consider a component, which fails if the number of shocks exceeds 2 or the cumulative
magnitude of a shock exceeds ‘5’.
The time between shocks f(T) is exponentially distributed with λ=0.001.
The cumulative magnitude of shock, M, is a continuous random variable with cdf Ф(M),
where, Mi is the magnitude of the ith shock and is an i.i.d variable. Determine the
component reliability at 500 hrs.
R(t ) Pr[ N (t ) 0] Pr[ N (t ) 1, M m)] Pr[ N (t ) 2, M m)
( t ) x t
n n
R(t ) e t
e .[ (m)] where , M M i
1 x! 1
[0.001(500)]x 0.001( 500 )
2
R(500) e 0.001( 500 )
e .[(5)]
1 x!
Assume shock magnitude to follow normal distribution with a mean and std dev of of 2 and
0.3 respectively
Other Shock Based Reliability Models
Two major types are distinguished depending on whether the effect of the shock on the system
is independent of its arrival time or not.
When there exists a dependence between the effect of the shock and its arrival time, the
damage caused by a shock is modelled by a random variable (representing the shock’s
magnitude), which is a function of its arrival time
Tool life
V Tn Dx Fy = C
QA