1 s2.0 S0951832022001016 Main

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

Reliability Engineering and System Safety 222 (2022) 108435

Contents lists available at ScienceDirect

Reliability Engineering and System Safety


journal homepage: www.elsevier.com/locate/ress

Maintenance cost-based importance analysis under different


maintenance strategies
Liwei Chen a, Chunchun Cheng a, Hongyan Dui b, *, Liudong Xing c
a
School of Electrical Engineering, Zhengzhou University, Zhengzhou 450001, China
b
School of Management Engineering, Zhengzhou University, Zhengzhou 450001, China
c
Department of Electrical & Computer Engineering, University of Massachusetts Dartmouth, MA 02747, USA

A R T I C L E I N F O A B S T R A C T

Keywords: As system components and their interactions become more and more complex, system reliability problems
Reliability become increasingly prominent. In the reliability research, component importance measures have been widely
Maintenance strategy used as an important decision-making aid index. Particularly, importance measures have been used in designing
Importance
or selecting the optimal maintenance strategy. However, the practical factor of maintenance cost is not
Maintenance cost-based importance measure
comprehensively considered in the existing importance analysis. This paper proposes new importance measures
based on two types of maintenance costs under different maintenance strategies, including the failure-triggered
replacement strategy that replaces damaged components, the age-based replacement strategy where a compo­
nent is preventively replaced and when it has been used for a certain period of time, and the hybrid strategy that
integrates the former two strategies. Methods for computing the proposed maintenance cost-based importance
measures for series systems and parallel systems are given. A case study of a propeller plane system is provided to
illustrate and validate the proposed method.

1. Introduction that have greater impacts on the system. Considering the component
states, Si et al. [3] suggested a comprehensive importance measure,
Modern industrial and technological systems are growing more and which can estimate the impact of a component on the performance of the
more powerful and complex. At the same time, the damage and eco­ entire system in a certain state. Xing and Dugan [4] put forward a
nomic losses caused by component failures in the system are worse than weight-based performance importance measure for phased-mission
ever. Therefore, it is an important issue to study the influence of com­ systems with combinatorial phase requirements and multiple perfor­
ponents on the system function, maintain the system components mance levels. In maritime transportation system, Dui et al. [5] used
reasonably, reduce the system risk caused by component failures, and Copeland method to comprehensively rank the importance of ports and
improve the reliability of system operation. One potential solution is routes and studied the restoration priority of interrupted ports and
through the component importance analysis. routes of different importance measures.
Since 1969 when Birnbaum proposed the first importance measure In terms of the importance analysis for different types of systems, a
[1], considerable research efforts have been dedicated to introducing lot of research and contributions [6–8] have been made to binary-state
and studying of component importance measures, reflecting the systems. However, some systems may undergo performance degrada­
different importance levels of different components of a tions before the complete failure, i.e., the systems are multi-state sys­
multi-component system. The following reviews some representative tems [9]. Some research efforts have been devoted to the multi-state
works in defining new importance measures, importance analysis in importance analysis [10–15]. For example, in [16] the composite
different types of systems (binary and multi-state), and importance importance measures of multi-state systems were defined and evaluated.
analysis considering maintenance cost and strategy. Some of these composite importance measures were extended by
In terms of importance measures, Dui et al. [2] proposed a new considering the epistemic uncertainty associated with component state
importance measure to analyze the influence of external factors on assignment in [17]. Xu et al. [18] proposed a new elasticity-based
system performance, which can be used by engineers to identify factors ordering metric for the importance of multi-state network

* Corresponding author at: Zhengzhou University, Kexue Road, Zhengzhou 450001, China.
E-mail address: duihongyan@zzu.edu.cn (H. Dui).

https://doi.org/10.1016/j.ress.2022.108435
Received 4 December 2021; Received in revised form 25 January 2022; Accepted 2 March 2022
Available online 8 March 2022
0951-8320/© 2022 Elsevier Ltd. All rights reserved.
L. Chen et al. Reliability Engineering and System Safety 222 (2022) 108435

components, which not only considers components’ multi-state char­ 3. Maintenance cost
acteristics, but also quantifies the impact of component capacity
improvement and recovery time on network resilience. There are three major types of system maintenance costs:
In terms of importance analysis considering maintenance strategies
and cost, to the best of our knowledge, only a limited body of research 1) Maintenance cost of improving the system reliability
efforts has been conducted. Different maintenance strategies (e.g., pre­ The maintenance cost for improving the reliability of different
ventive maintenance [19–23], predictive maintenance [24], components is often different. For example, in a water supply system,
condition-based maintenance [25]) may have different significant ef­ improving the reliability of a water pump incurs a different cost from
fects on the system reliability and availability [26,27]. In [28] and [29], upgrading the reliability of a switchboard.
it was shown that the system reliability is not only related to mainte­ 2) Maintenance cost associated with the component failure
nance strategy, but also related to maintenance cost. Thus, in [30] Dui The failure of a system component incurs certain maintenance cost
et al. considered the impact of maintenance cost and time on the system and different component failures may incur different maintenance
reliability, and proposed a method of selecting components for main­ costs.
tenance in transportation networks. 3) Cost associated with the system failure
To the best of our knowledge, the existing importance analysis A system is usually designed and built to accomplish a specific
studies only consider the maintenance cost of upgrading the reliability task. When the system fails, it may cause some losses, such as damage
of system components and a single maintenance strategy. This paper to health and the environment, release of dangerous substances,
makes advancements by proposing a new importance measure that direct or indirect economic losses, and so on. For example, the failure
consider multiple types of maintenance costs and different maintenance of the water supply system can cause the system user’s production
strategy conditions. Specifically, the following contributions are made in line to stop running, resulting in economic losses.
this work.
Because the goal of the entire system design is to extend the working
• Proposing a new maintenance cost-based importance measure time of the system with as little maintenance cost as possible, it is
(MCIM) and its evaluation method under different maintenance necessary to consider the cost factors in the importance analysis. To the
strategy conditions, including a failure-triggered replacement strat­ best of our knowledge, only the first type of maintenance cost has been
egy, an age-based replacement strategy, and a hybrid maintenance considered in the existing importance analysis [32]. In this work, we
strategy. develop a new importance measure, which can reflect the impact of the
• Deriving formulas of computing the MCIM for series systems and second and third types of maintenance costs related to component and
parallel systems under the failure-triggered replacement strategy and system failures.
the age-based replacement strategy. Generally, the system failures caused by malfunctions of different
• Conducting a case study of a propeller aircraft system that contains a components require different system maintenance time, and thus the
mix of series and parallel structures. Effectiveness of the proposed corresponding maintenance costs are different. Therefore, we define the
important measures is investigated in comparison to that of the maintenance cost Ck (t) of the system within (0, t) time due to the failure
traditional importance measures without considering maintenance of component kas in (3).
cost under the failure-triggered replacement strategy, and a hybrid ( )
Ck (t) = Cs,k (t) + Ck (t) = cs,k Λk (t) + ck Λk (t) = ck + cs,k Λk (t), (3)
maintenance strategy that integrates the failure-triggered and age-
based replacement strategies.
where Cs,k denotes the total repair cost for the system failure caused by
the failure of component k, Ck denotes the total repair cost of component
The rest of this paper is as follows Section 2. gives the traditional
k, Λk (t) represents the expected number of failures of component kin (0,
importance measurement equation Section 3. gives the calculation
t), ck denotes the repair cost of component k per failure, cs,k denotes the
method of maintenance cost Section 4. Proposed MCIM Section 5. pro­
repair cost for the system failure caused by each failure of component k
poses the MCIM under the failure-triggered replacement strategy and
and Minus the maintenance cost of component k. It should be noted that
give the relevant formula Section 6. proposes the MCIM under the age-
the failure of a component does not necessarily lead to the failure of the
based replacement strategy and also give the relevant formula. In sec­
system. If the failure of component k does not cause the failure of the
tion 7, through the analysis of an application, some results are got to
system, then cs,k = 0.
demonstrate the proposed methods Section 8. concludes the paper.

2. Traditional importance measures 4. Proposed maintenance cost-based importance measure

Birnbaum importance measure was first proposed by Birnbaum in As reviewed in Section 2, the Birnbaum importance measure reflects
1969. The Birnbaum importance [1] for component i in a binary-state the change in the overall system reliability when the reliability of a
system is defined as in (1). component changes by a certain amount. The greater the value of
Birnbaum importance for a system component, the higher the impor­
IiBM (t) =
∂RS (t)
(1) tance of this component.
∂Ri (t) When selecting a component for maintenance, our goal is to find the
Its physical meaning is the amount of change in system reliability component that incurs the smallest maintenance cost. If a component is
Rs (t) when the reliability of component i Ri (t) changes from 1 to 0. a single-point-of-failure, i.e., a critical component, upon being damaged,
In [31], Si et al. proposed the integrated importance measure (IIM), it is repaired immediately or the system will not function properly. For
whose definition for component iis shown in (2). non-critical components, the importance analysis is used to select the
component that needs to be repaired when the system is maintained. The
IiIIM (t) = Ri (t)λi (t)
∂RS (t)
=−
dRi (t) ∂RS (t)
(2) Birnbaum importance of componentibased on maintenance cost for
∂Ri (t) dt ∂Ri (t) binary-state systems is

dRi (t)/dt Ci (t)


where λi (t) = − Ri (t) is the failure intensity of component i. IiBM,c (t) = (4)
IiBM (t)

2
L. Chen et al. Reliability Engineering and System Safety 222 (2022) 108435

Fig. 1. Block diagram of the series system

The physical meaning of this formula is that when IiBM,c is large, a


small change in the importance of component i will cause a relatively
large change in the maintenance cost of the entire system in the time
period of (0, t).
We can extend (4) using the IIM, leading to the maintenance cost-
based importance measure (MCIM) of component i as

Ci (t)
IiMCIM (t) = (5)
IiIIM (t)
Ci (t)
At any moment t, the ratio IiIIM (t)
reflects the relative change trend of
maintenance cost with respect to the IIM in the time period (0, t). Ac­
cording to (2), we have

Ci (t) Ci (t) Ci (t)


IiMCIM (t) = IIM
= dRi (t) ∂RS (t) = − ′ (6)
Ii (t) − dt ⋅ ∂R (t) Ri (t)IiBM (t)
i

Therefore, when Ri (t)IiBM (t) > Rj (t)IjBM (t), under the same mainte­
′ ′

nance cost, the increase in importance of maintaining component jis less Fig. 2. Block diagram of the parallel system
than that of maintaining component i. Therefore, a higher priority
should be given to the maintenance of component i. (0, t) is
( )
5. MCIM under the failure-triggered replacement strategy Ci (t) = ci + cs,i λi t (11)

In the failure-triggered replacement strategy, the component is In a series system, the Birnbaum importance of component i is
repaired as new after the failure. Without loss of generality, under the
replacement strategy the age of a component is assumed to follow the ∏
n
Rk (t) ( )
exponential distribution with reliability functionR(t) = e− λi t , where λi ∂RS (t) k=1 ∏
n ∑n
IiBM (t) = = = Rk (t) = exp − t λk (12)
denotes the failure intensity of component i. The maintenance cost of ∂Ri (t) Ri (t) k=1,k∕
=i
k=1,k∕
=i

component i per unit time is the sum of the maintenance cost of the
component itself and the cost of system failure caused by the compo­ Therefore, according to (9), for any componenti, its MCIM is
nent. According to (3), it is evaluated as ( ) ( )
Ci (t) ci + cs,i λi t ci + cs,i λi t
( ) ( ) IiMCIM (t) = IIM = = ( ∑ )
Ci (t) = ci + cs,i Λi (t) = ci + cs,i λi t (7) BM
Ii (t) λi (t)Ri (t)Ii (t) λi e− λi t exp − t n
=i λk
(13)
k=1,k∕
( )
Thus, according to (2), the IIM of the component is = (
ci + cs,i t
∑n )
exp − t k=1 λk
dRi (t) ∂RS (t)
IiIIM (t) = − ⋅ = λi e− λi t
⋅IiBM (t) (8)
dt ∂Ri (t)
5.2. Properties in parallel systems under the replacement strategy
Therefore, according to (6), the MCIM of component i is
The system structure block diagram of the parallel system is shown in
( ) ( )
Ci (t) ci + cs,i λi t ci + cs,i teλi t Fig. 2.
IiMCIM (t) = IIM = − λ t BM = (9)
Ii (t) λi e i Ii (t) IiBM (t)
Theorem 2. In a parallel system, if each component is independent of
each other, then the IIM of component i based on maintenance cost is
5.1. Properties in series systems under the replacement strategy
ci t
IiMCIM (t) = ∏
n (14)
The system structure block diagram of the series system is shown in e− λi t (1 − exp( − λk t))
Fig. 1. k=1,k∕
=i

Theorem 1. In a series system, if the components are independent of Proof. In a parallel system, if and only if all components fail
each other, and the age of each component i obeys the exponential ∑
( ni=1 Ri = 0), the entire system fails (Rs = 0). Assuming that it is
distribution with parameterλi , then the IIM of component i based on impossible for more than one component to fail at the same time in any
maintenance cost is short time. So, the probability of the system failure caused by any
( )
ci + cs,i t component i is zero. Therefore, according to (7), the maintenance cost of
IiMCIM (t) = ( ∑n ) (10) component i in the time period (0, t) in the parallel system is
exp − t k=1 λk
( )
Ci (t) = ci + cs,i Λi (t) = ci λi t (15)
Proof. In a series system, when any component fails(Ri = 0), it
causes the entire system to fail(Rs = 0). Assuming that it is impossible In a parallel system, the Birnbaum importance of component i is
for more than one component to fail at the same time in any short time.
So only the system failure caused by a single component is considered.
According to (7), the maintenance cost of component i in the time period

3
L. Chen et al. Reliability Engineering and System Safety 222 (2022) 108435

( )

n
∂ 1− (1 − Rk ) ECC(T) = cf F(T) + cp R(T) (21)
∂RS (t) k=1

n
IiBM (t) = = = (1 − Rk )
∂Ri (t) ∂Ri (t) (16) where cf denotes the repair cost after the component failure, and
k=1,k∕
=i

n cp denotes the preventive maintenance cost.
= (1 − exp( − λk t)) Because the IIM needs to consider the failure intensity, which is the
k=1,k∕
=i inverse of the ECL, as in (22).
For any component i, according to (9), its MCIM is 1 1
λi (T) = = ∫T (22)
C (t) i
ci λi t ECL(T) R(t)dt + τf F(T) + τp R(T)
0
IiMCIM (t) = IIM = ∏
n
Ii (t)
λi e− λi t (1 − exp( − λk t)) Reliability is the expected ratio of the expected working time to the
k=1,k∕
=i length of the cycle, as in (23).
ci t ∫T
= (17) E[U] R(x)dx
∏ (23)
n 0
λi t Ri (T) = = ∫T
e− (1 − exp( − λk t)) ECL(T) R(x)dx + τF(T) + τR(T)
0
k=1,k∕
=i

Thus, according to (2), the IIM for component i is


6. MCIM under the age-based replacement strategy
1 E[U] BM E[U]IiBM (T)
IiIIM (T) = λi (T)Ri (T)IiBM (T) = I (T) = (24)
ECL(T) ECL(T) i ECL2 (T)
In the age-based replacement strategy, when the component is still
functioning but have reached a specified age T, preventive maintenance The maintenance cost per unit time of component i in the steady state
is performed on the component. If the component fails before T, the of the system is
component will be replaced after failure. This kind of repair is referred
ECC(T) cf F(T) + cp R(T)
to as post-maintenance. The established time interval T is called the J∞ (T) = = ∫T (25)
ECL(T)
scheduled replacement time interval. R(t)dt + τf F(T) + τp R(T)
0

The maintenance cycle is defined as the time required for a compo­ Thus, according to (6), the MCIM of component i is evaluated as
nent to reach its preventive maintenance or subsequent maintenance
after preventive maintenance or post-maintenance. Thus, the expected Ci (T) tJ∞ (T) ECC(T) ECL(T)2 ECC(T)ECL(T)
IiMCIM (T) = = IIM =t =t
operational/up time (denoted as U) of a component in a maintenance IIM
Ii (T) Ii (T) ECL(T) E[U]IiBM (T) E[U]IiBM (T)
cycle is (26)
∫T
E[U] = R(x)dx (18) 6.1. Properties in series systems under the age-based replacement strategy
0

As the component is repaired in two ways: repair after the failure,


Theorem 3. In a series system, if the components are independent of
and the probability in this case is F(t). Preventive maintenance when
each other and obey the exponential distribution, the MCIM of compo­
reaching the certain age T, and the probability in this case isR(t) = 1 −
nent i is
F(t). Therefore, the expected maintenance time of the component in a
maintenance cycle is tECC(T)ECL(T)
IiMCIM (t) = ( ∑ ) (27)
̃ = τf F(T) + τp R(T),
E[V] (19) E[U]exp − t nk=1,k∕
=i λk

where τf denotes the repair time following the component failure, and τp Proof. According to (25), the maintenance cost of component i in
denotes the preventive maintenance time. Therefore, the expected cycle the time period (0, t) is
length (ECL) is the expected working/up time of the system plus the ECC(T)
expected repair time of the component. Ci (t) = tJ∞ (T) = t (28)
ECL(T)
∫T
̃ =
ECL(T) = E[U] + E[V] R(x)dx + τf F(T) + τp R(T) (20) Therefore, according to (12) and (26), for any component i, its MCIM
is
0

The expected cycle cost (ECC) can also be calculated in a similar way
as in (21).

Fig. 3. A propeller plane system

4
L. Chen et al. Reliability Engineering and System Safety 222 (2022) 108435

Ci (t) t ECC(T) tECC(T)ECL(T) one of them includes the maintenance costs for repairing the component
(29)
ECL(T)
IiMCIM (t) = = = ( ∑ ) and for system failure.
IiIIM (t) E[U]IiBM
2
(T)
E[U]exp − t n λk
ECL (T) k=1,k∕
=i

7.1. Single failure-triggered replacement strategy


6.2. Properties in parallel systems under the age-based replacement
strategy Assume that all the system components obey the exponential distri­
bution Ri (t) = e(− λi t) with parameters λi , where i = 1, 2, 3, 4, 5, 6, 7, 8, 9.
The reliability function of the system is
Theorem 4. In a parallel system, if the components are independent of ( )
each other and obey the exponential distribution, the MCIM of compo­ Rs (t) = R7 (t)R8 (t)R9 (t)
1 − (R5 (t)(R1 (t) − 1)(R2 (t) − 1) − R5 (t) + 1)
nent i is ∗(R6 (t)(R3 (t) − 1)(R4 (t) − 1) − R6 (t) + 1)

tECC(T)ECL(T) According to (1), the Birnbaum importance measures of the nine


IiMCIM (t) = ∏n (30) components are
E[U] (1 − exp( − λk t)) ( )
k=1,k∕
=i
BM ∂RS (t) R5 (t)R7 (t)R8 (t)R9 (t)(R2 (t)− 1)
I1 (t)= =−
∂R1 (t) ∗(R6 (t)(R3 (t)− 1)(R4 (t)− 1)− R6 (t)+1)
Proof. According to (25), the maintenance cost of component i in ( )
the parallel system during the time period (0, t) is ∂RS (t) R5 (t)R7 (t)R8 (t)R9 (t)(R1 (t)− 1)
I2BM (t)= =−
ECC(T) ∂R2 (t) ∗(R6 (t)(R3 (t)− 1)(R4 (t)− 1)− R6 (t)+1)
Ci (t) = tJ∞ (T) = t (31) ( )
ECL(T) ∂RS (t) R 6 (t)R7 (t)R8 (t)R9 (t)(R4 (t)− 1)
I3BM (t)= =−
∂R3 (t)
For any component i, according to (1) and (26), its MCIM is ∗(R5 (t)(R1 (t)− 1)(R2 (t)− 1)− R5 (t)+1)
( )
BM ∂RS (t) R6 (t)R7 (t)R8 (t)R9 (t)(R3 (t)− 1)
t ECC(T) I4 (t)= =−
Ci (t) tECC(T)ECL(T) ∂R4 (t)
(32) ∗(R5 (t)(R1 (t)− 1)(R2 (t)− 1)− R5 (t)+1)
ECL(T)
IiMCIM (t) = = = ∏n
IiIIM (t) E[U]IiBM (T) ( )
2
ECL (T) E[U] (1 − exp( − λk t)) ∂RS (t) (R1 (t)− 1)(R2 (t)− 1)− 1)
BM
k=1,k∕
=i I5 (t)= =− R7 (t)R8 (t)R9 (t)
∂R5 (t) ∗(R6 (t)(R3 (t)− 1)(R4 (t)− 1)− R6 (t)+1
( )
7. An application ∂RS (t) (R 3 (t)− 1)(R4 (t)− 1)− 1)
I6BM (t)= =− R7 (t)R8 (t)R9 (t)
∂R6 (t) ∗(R5 (t)(R1 (t)− 1)(R2 (t)− 1)− R5 (t)+1
The structure diagram of a propeller plane system is shown in Fig. 3. ( )
Among the system components, engines 1, 2, 3, 4 are represented by ∂RS (t) (R5 (t)(R1 (t)− 1)(R2 (t)− 1)− R5 (t)+1)
I7BM (t)= =− R8 (t)R9 (t)
components 1, 2, 3, 4, portside and starboard are represented by com­ ∂R7 (t) ∗(R6 (t)(R3 (t)− 1)(R4 (t)− 1)− R6 (t)+1)− 1
ponents 5, 6, and control panels, wing 1 and wing 2 are represented by
components 7, 8, and 9, respectively.
From Fig. 3, as component 1 and component 2 are connected in ( )
parallel, the failure of component 1 or component 2 will not cause the ∂RS (t) (R5 (t)(R1 (t)− 1)(R2 (t)− 1)− R5 (t)+1)
I8BM (t)= =− R7 (t)R9 (t)
system failure. Therefore, the maintenance cost caused by the failure of ∂R8 (t) ∗(R6 (t)(R3 (t)− 1)(R4 (t)− 1)− R6 (t)+1)− 1
component 1 or component 2 only includes the cost of component ( )
maintenance and does not include the maintenance cost caused by the I9BM (t)=
∂RS (t)
=− R7 (t)R8 (t)
(R5 (t)(R1 (t)− 1)(R2 (t)− 1)− R5 (t)+1)
system failure. So cs,1 = 0, cs,2 = 0. Similarly,cs,3 = 0, cs,4 = 0, cs,5 = 0, ∂R9 (t) ∗(R6 (t)(R3 (t)− 1)(R4 (t)− 1)− R6 (t)+1)− 1
cs,6 = 0. Component 7, component 8 and component 9 are connected in
series with the rest of the system, so the failure of any of them leads to According to (2), the IIMs of the nine components are
the failure of the entire system. Therefore, the maintenance cost for any

dR1 (t) ∂Rs (t) dR2 (t) ∂Rs (t)


I1IIM (t) = − ⋅ = λ1 e(− λ1 t) BM
I1 (t), I2IIM (t) =− ⋅ = λ2 e(− λ2 t) BM
I2 (t)
dt ∂R1 (t) dt ∂R2 (t)
dR3 (t) ∂Rs (t) dR4 (t) ∂Rs (t)
I3IIM (t) = − ⋅ = λ3 e(− λ3 t) BM
I3 (t), I4IIM (t) =− ⋅ = λ4 e(− λ4 t) BM
I4 (t)
dt ∂R3 (t) dt ∂R4 (t)
dR5 (t) ∂Rs (t) dR6 (t) ∂Rs (t)
I5IIM (t) = − ⋅ = λ5 e(− λ5 t) BM
I5 (t), I6IIM (t) =− ⋅ = λ6 e(− λ6 t) BM
I6 (t)
dt ∂R5 (t) dt ∂R6 (t)
dR7 (t) ∂Rs (t) dR8 (t) ∂Rs (t)
I7IIM (t) = − ⋅ = λ7 e(− λ7 t) BM
I7 (t), I8IIM (t) =− ⋅ = λ8 e(− λ8 t) BM
I8 (t)
dt ∂R7 (t) dt ∂R8 (t)
dR9 (t) ∂Rs (t)
I9IIM (t) = − ⋅ = λ9 e(− λ9 t) BM
I9 (t)
dt ∂R9 (t)

5
L. Chen et al. Reliability Engineering and System Safety 222 (2022) 108435

Fig. 4. BM of components

Fig. 5. IIM of components

6
L. Chen et al. Reliability Engineering and System Safety 222 (2022) 108435

Fig. 6. BM based on maintenance cost

Fig. 7. IIM based on maintenance cost

7
L. Chen et al. Reliability Engineering and System Safety 222 (2022) 108435

According to (4), the maintenance cost-based Birnbaum importance


measures of the nine system components are According to (5), the MCIM of the nine components are

C1 (t) c1 λ1 t BM,c C2 (t) c2 λ2 t BM,c C3 (t) c3 λ3 t


I1BM,c (t) = = ,I (t) = BM = ,I (t) = BM =
I1BM (t) A1 2 I2 (t) A2 3 I3 (t) A3
C4 (t) c4 λ4 t BM,c C5 (t) c5 λ5 t BM,c C6 (t) c6 λ6 t
I4BM,c (t) = BM = , I5 (t) = BM = , I6 (t) = BM =
I4 (t) A4 I5 (t) A5 I6 (t) A6
7
( ) 8
( ) ( )
C (t) cs,7 + c7 λ7 t C (t) c s,8 + c 8 λ 8 t C9 (t) cs,9 + c9 λ9 t
I7BM,c (t) = BM = , I8BM,c (t) = BM = , I9BM,c (t) = BM =
I7 (t) A7 I8 (t) A8 I9 (t) A9

where

( λ5 t) (λ2 +λ5 )t)


)
e(− − e(− − e(− (λ2 +λ3 +λ4 +λ5 +λ6 )t) + e(− (λ2 +λ3 +λ5 +λ6 )t) +
A1 = − (λ2 +λ4 +λ5 +λ6 )t) (− (λ3 +λ4 +λ5 +λ6 )t) (− (λ3 +λ5 +λ6 )t) (− (λ4 +λ5 +λ6 )t) e(− (λ7 +λ8 +λ9 )t)
e(− +e − e − e

( λ5 t) (λ1 +λ5 )t)


)
e(− − e(− − e(− (λ1 +λ3 +λ4 +λ5 +λ6 )t) + e(− (λ1 +λ3 +λ5 +λ6 )t) + (λ7 +λ8 +λ9 )t)
A2 = − e(−
e(− (λ1 +λ4 +λ5 +λ6 )t)
+ e(− (λ3 +λ4 +λ5 +λ6 )t) − e(− (λ3 +λ5 +λ6 )t) − e(− (λ4 +λ5 +λ6 )t)

( λ6 t) (λ4 +λ6 )t)


)
e(− − e(− − e(− (λ1 +λ2 +λ4 +λ5 +λ6 )t) + e(− (λ1 +λ2 +λ5 +λ6 )t) + (λ7 +λ8 +λ9 )t)
A3 = − e(−
e(− (λ1 +λ4 +λ5 +λ6 )t)
+e (− (λ2 +λ4 +λ5 +λ6 )t)
− e(− (λ1 +λ5 +λ6 )t)
− e(− (λ2 +λ5 +λ6 )t)

( λ6 t) (λ3 +λ6 )t)


)
e(− − e(− − e(− (λ1 +λ2 +λ3 +λ5 +λ6 )t) + e(− (λ1 +λ2 +λ5 +λ6 )t) + (λ7 +λ8 +λ9 )t)
A4 = − e(−
e(− (λ1 +λ3 +λ5 +λ6 )t)
+e (− (λ2 +λ3 +λ5 +λ6 )t)
− e(− (λ1 +λ5 +λ6 )t)
− e(− (λ2 +λ5 +λ6 )t)

⎛ ⎞
e(− λ1 t) + e(− λ2 t) − e(− (λ1 +λ2 )t) − e(− (λ1 +λ2 +λ3 +λ4 +λ6 )t) + e(− (λ1 +λ2 +λ3 +λ6 )t)
A5 = ⎝ +e(− (λ1 +λ2 +λ4 +λ6 )t) + e(− (λ1 +λ3 +λ4 +λ6 )t) − e(− (λ1 +λ3 +λ6 )t) − e(− (λ1 +λ4 +λ6 )t) ⎠e(− (λ7 +λ8 +λ9 )t)

+e(− (λ2 +λ3 +λ4 +λ6 )t) − e(− (λ2 +λ3 +λ6 )t) − e(− (λ2 +λ4 +λ6 )t)

⎛ ⎞
e(− λ3 t) + e(− λ4 t) − e(− (λ3 +λ4 )t) − e(− (λ1 +λ2 +λ3 +λ4 +λ5 )t) + e(− (λ1 +λ2 +λ3 +λ5 )t)

A6 = +e (− (λ1 +λ2 +λ4 +λ5 )t)
+e (− (λ1 +λ3 +λ4 +λ5 )t)
− e(− (λ1 +λ3 +λ5 )t)
− e(− (λ1 +λ4 +λ5 )t) ⎠ (−
e (λ7 +λ8 +λ9 )t)

+e(− (λ2 +λ3 +λ4 +λ5 )t) − e(− (λ2 +λ3 +λ5 )t) − e(− (λ2 +λ4 +λ5 )t)

8
L. Chen et al. Reliability Engineering and System Safety 222 (2022) 108435

Fig. 8. IIM of components

⎛ ⎞
e(− (λ1 +λ5 )t) + e(− (λ2 +λ5 )t) − e(− (λ1 +λ2 +λ5 )t) + e(− (λ3 +λ6 )t) + e(− (λ4 +λ6 )t)
⎜− e 3 4 6 − e 1 2 3 4 5 6 + e 1 2 3 5 6
(− (λ +λ +λ )t) (− (λ +λ +λ +λ +λ +λ )t) (− (λ +λ +λ +λ +λ )t) ⎟ (−
A7 = ⎜
⎝ +e(− (λ1 +λ2 +λ4 +λ5 +λ6 )t) + e(− (λ1 +λ3 +λ4 +λ5 +λ6 )t) − e(− (λ1 +λ3 +λ5 +λ6 )t)
⎟e

(λ8 +λ9 )t)

(− (λ1 +λ4 +λ5 +λ6 )t) (− (λ2 +λ3 +λ4 +λ5 +λ6 )t) (− (λ2 +λ3 +λ5 +λ6 )t) (− (λ2 +λ4 +λ5 +λ6 )t)
− e +e − e − e

⎛ ⎞
e(− (λ1 +λ5 )t) + e(− (λ2 +λ5 )t) − e(− (λ1 +λ2 +λ5 )t) + e(− (λ3 +λ6 )t) + e(− (λ4 +λ6 )t)
⎜− e 3 4 6 − e 1 2 3 4 5 6 + e 1 2 3 5 6
(− (λ +λ +λ )t) (− (λ +λ +λ +λ +λ +λ )t) (− (λ +λ +λ +λ +λ )t) ⎟ (−
A8 = ⎜
⎝ +e(− (λ1 +λ2 +λ4 +λ5 +λ6 )t) + e(− (λ1 +λ3 +λ4 +λ5 +λ6 )t) − e(− (λ1 +λ3 +λ5 +λ6 )t)
⎟e

(λ7 +λ9 )t)

(− (λ1 +λ4 +λ5 +λ6 )t) (− (λ2 +λ3 +λ4 +λ5 +λ6 )t) (− (λ2 +λ3 +λ5 +λ6 )t) (− (λ2 +λ4 +λ5 +λ6 )t)
− e +e − e − e

⎛ ⎞
e(− (λ1 +λ5 )t) + e(− (λ2 +λ5 )t) − e(− (λ1 +λ2 +λ5 )t) + e(− (λ3 +λ6 )t) + e(− (λ4 +λ6 )t)
⎜ − e(− (λ3 +λ4 +λ6 )t) − e(− (λ1 +λ2 +λ3 +λ4 +λ5 +λ6 )t) + e(− (λ1 +λ2 +λ3 +λ5 +λ6 )t) + ⎟ (−
A9 = ⎜
⎝ e(− (λ1 +λ2 +λ4 +λ5 +λ6 )t) + e(− (λ1 +λ3 +λ4 +λ5 +λ6 )t) − e(− (λ1 +λ3 +λ5 +λ6 )t)
⎟e

(λ7 +λ8 )t)

(λ1 +λ4 +λ5 +λ6 )t) (λ2 +λ3 +λ4 +λ5 +λ6 )t) (λ2 +λ3 +λ5 +λ6 )t) (λ2 +λ4 +λ5 +λ6 )t)
− e(− + e(− − e(− − e(−

9
L. Chen et al. Reliability Engineering and System Safety 222 (2022) 108435

Fig. 9. Maintenance cost-based Birnbaum importance of the nine components

C1 (t) c1 t C2 (t) c2 t C3 (t) c3 t


I1MCIM (t) = = , I MCIM (t) = IIM = , I MCIM (t) = IIM =
I1IIM (t) A1 e(− λ1 t) 2 I2 (t) A2 e(− λ2 t) 3 I3 (t) A3 e(− λ3 t)
C4 (t) c4 t C5 (t) c5 t C6 (t) c6 t
I4MCIM (t) = IIM = , I MCIM (t) = IIM =
(− λ4 t) 5
, I MCIM (t) = IIM =
(− λ5 t) 6
I4 (t) A4 e I5 (t) A5 e I6 (t) A6 e(− λ6 t)
( ) ( ) ( )
C7 (t) cs,7 + c7 t MCIM C8 (t) cs,8 + c8 t MCIM C9 (t) cs,9 + c9 t
I7MCIM (t) = IIM = , I 8 (t) = IIM = , I 9 (t) = IIM =
I7 (t) A7 e(− λ7 t) I8 (t) A8 e(− λ8 t) I9 (t) A9 e(− λ9 t)

Assuming the time interval is [0, 8], c1 = 15, c2 = 15, c3 = 15, c4 = Figs. 6 and 7 reflect the trend of the maintenance cost-based Birn­
15, c5 = 10, c6 = 15, c7 = 5, c8 = 30, c9 = 30 (all unit prices are taken in baum importance and MCIM of the nine components.
K RMB). The cost of system failure caused by components 7, 8, and 9 are From Figs. 6 and 7, the relative importance of components has
cs,7 = 50, cs,8 = 300, cs,9 = 300, which are much higher than the changed more significantly over time. The MCIM of components 1, 2, 3,
maintenance cost associated with maintaining a single component. The and 4 starts with a relatively high level, then decreases to a certain level
failure intensities of the system components are λ1 = λ2 = λ3 = λ4 = 0.1, and then began to increase. The MCIM of components 5 and 6 has slowly
λ5 = λ6 = 0.08, λ7 = 0.05, λ8 = λ9 = 0.08 per year. increased over time, but due to the difference in their maintenance costs,
Figs. 4 and 5 show the Birnbaum importance and IIM of components. the MCIM of component 6 is always higher than that of component 5.
From Figs. 4 and 5, the Birnbaum importance and IIM of components Component 7 has the lowest overall importance based on the mainte­
7, 8, and 9 decrease with time. The importance of components 1, 2, 3, 4, nance cost in the first half of the interval due to its low maintenance cost,
5, and 6 slowly increases with time, and the increasing trends of com­ and its MCIM slowly increases over time. The overall MCIM of compo­
ponents 1, 2, 3, and 4 are consistent because their failure intensities are nents 8, 9 has continuously increased from a lower level to the highest
the same. The growth rates of components 5 and 6 also remain the same by the end of the interval considered in this study.
because they have the same failure intensity. Since Birnbaum impor­ In summary, as compared to the existing Birnbaum importance and
tance and IIM do not take into account the maintenance cost, although IIM that fail to consider maintenance costs-a crucial factor in actual
component 5 and component 6 have different maintenance costs, there maintenance practices, the proposed maintenance cost-based impor­
is no difference in their importance levels, because their age distribution tance measures can reflect the impact of different types of maintenance
and location in the system are the same. However, different mainte­ costs, thus can better guide the practical maintenance work more
nance costs must be handled with different priorities in the maintenance effectively.
practice, so it is crucial to incorporate maintenance costs in the impor­
tance analysis.

10
L. Chen et al. Reliability Engineering and System Safety 222 (2022) 108435

Fig. 10. MCIM of components

7.2. Hybrid maintenance strategy


C2 (t) c2 λ2 t BM,c C4 (t) c4 λ4 t BM,c C6 (t) c6 λ6 t
I2BM,c (t) = = ,I (t) = BM = ,I (t) = BM =
In this case study, components 2, 4, 6, 8 and 9 adopt the failure- I2BM (t) A2 4 I4 (t) A4 6 I6 (t) A6
( ) ( )
triggered replacement strategy, while components 1, 3, 5 and 7 adopt 8
C (t) cs,8 + c8 λ8 t BM,c 9
C (t) cs,9 + c9 λ9 t
the age-based replacement strategy. I8BM,c (t) = BM = , I9 (t) = BM =
I8 (t) A8 I9 (t) A9
The Birnbaum importance measures has nothing to do with the
maintenance strategy adopted by the component.
where
According to (2), the IIMs of the five components are

dR2 (t) ∂Rs (t) dR4 (t) ∂Rs (t)


I2IIM (t) = − ⋅ = λ2 e(− λ2 t)
⋅I2BM (t), I4IIM (t) = − ⋅ = λ4 e(− λ4 t)
⋅I4BM (t)
dt ∂R2 (t) dt ∂R4 (t)
dR6 (t) ∂Rs (t) dR8 (t) ∂Rs (t)
I6IIM (t) = − ⋅ = λ6 e(− λ6 t)
⋅I6BM (t), I8IIM (t) = − ⋅ = λ8 e(− λ8 t)
⋅I8BM (t)
dt ∂R6 (t) dt ∂R8 (t)
dR9 (t) ∂Rs (t)
I9IIM (t) = − ⋅ = λ9 e(− λ9 t)
⋅I9BM (t)
dt ∂R9 (t)

According to (4), the maintenance cost-based Birnbaum importance


measure of the five components 2, 4, 6, 8 and 9 are

( λ5 t) (λ1 +λ5 )t)


)
e(− − e(− − e(− (λ1 +λ3 +λ4 +λ5 +λ6 )t) + e(− (λ1 +λ3 +λ5 +λ6 )t) +
A2 = − (λ1 +λ4 +λ5 +λ6 )t) (− (λ3 +λ4 +λ5 +λ6 )t) (− (λ3 +λ5 +λ6 )t) (− (λ4 +λ5 +λ6 )t) e(− (λ7 +λ8 +λ9 )t)
e(− +e − e − e

11
L. Chen et al. Reliability Engineering and System Safety 222 (2022) 108435

( λ6 t) (λ3 +λ6 )t)


)
e(− − e(− − e(− (λ1 +λ2 +λ3 +λ5 +λ6 )t) + e(− (λ1 +λ2 +λ5 +λ6 )t) + (λ7 +λ8 +λ9 )t)
A4 = − e(−
e(− (λ1 +λ3 +λ5 +λ6 )t)
+e (− (λ2 +λ3 +λ5 +λ6 )t)
− e(− (λ1 +λ5 +λ6 )t)
− e(− (λ2 +λ5 +λ6 )t)

⎛ ⎞
e(− λ3 t) + e(− λ4 t) − e(− (λ3 +λ4 )t) − e(− (λ1 +λ2 +λ3 +λ4 +λ5 )t) + e(− (λ1 +λ2 +λ3 +λ5 )t)

A6 = +e (− (λ1 +λ2 +λ4 +λ5 )t)
+e (− (λ1 +λ3 +λ4 +λ5 )t)
− e(− (λ1 +λ3 +λ5 )t)
− e(− (λ1 +λ4 +λ5 )t) ⎠ (−
e (λ7 +λ8 +λ9 )t)

+e(− (λ2 +λ3 +λ4 +λ5 )t) − e(− (λ2 +λ3 +λ5 )t) − e(− (λ2 +λ4 +λ5 )t)

⎛ ⎞
e(− (λ1 +λ5 )t) + e(− (λ2 +λ5 )t) − e(− (λ1 +λ2 +λ5 )t) + e(− (λ3 +λ6 )t) + e(− (λ4 +λ6 )t)
⎜ − e(− (λ3 +λ4 +λ6 )t) − e(− (λ1 +λ2 +λ3 +λ4 +λ5 +λ6 )t) + e(− (λ1 +λ2 +λ3 +λ5 +λ6 )t) ⎟ (−
A8 = ⎜
⎝ +e(− (λ1 +λ2 +λ4 +λ5 +λ6 )t) + e(− (λ1 +λ3 +λ4 +λ5 +λ6 )t) − e(− (λ1 +λ3 +λ5 +λ6 )t)
⎟e

(λ7 +λ9 )t)

(− (λ1 +λ4 +λ5 +λ6 )t) (− (λ2 +λ3 +λ4 +λ5 +λ6 )t) (− (λ2 +λ3 +λ5 +λ6 )t) (− (λ2 +λ4 +λ5 +λ6 )t)
− e +e − e − e

⎛ ⎞
e(− (λ1 +λ5 )t) + e(− (λ2 +λ5 )t) − e(− (λ1 +λ2 +λ5 )t) + e(− (λ3 +λ6 )t) + e(− (λ4 +λ6 )t)
⎜ − e(− (λ3 +λ4 +λ6 )t) − e(− (λ1 +λ2 +λ3 +λ4 +λ5 +λ6 )t) + e(− (λ1 +λ2 +λ3 +λ5 +λ6 )t) + ⎟ (−
A9 = ⎜
⎝ e(− (λ1 +λ2 +λ4 +λ5 +λ6 )t) + e(− (λ1 +λ3 +λ4 +λ5 +λ6 )t) − e(− (λ1 +λ3 +λ5 +λ6 )t)
⎟e

(λ7 +λ8 )t)

(− (λ1 +λ4 +λ5 +λ6 )t) (− (λ2 +λ3 +λ4 +λ5 +λ6 )t) (− (λ2 +λ3 +λ5 +λ6 )t) (− (λ2 +λ4 +λ5 +λ6 )t)
− e +e − e − e

According to (9), the MCIM of the five components 2, 4, 6, 8 and 9 According to (24), the IIM of the four components 1, 3, 5 and 7 are
are

C2 (t) c2 t C4 (t) c4 t
I2MCIM (t) = = , I MCIM (t) = IIM =
I2IIM (t) A2 e(− λ2 t) 4 I4 (t) A4 e(− λ4 t)
( ) ( )
C6 (t) c6 t C8 (t) cs,8 + c8 t MCIM C9 (t) cs,9 + c9 t
I6MCIM (t) = IIM = , I MCIM
(t) = = , I (t) = =
I6 (t) A6 e(− λ6 t) 8 I8IIM (t) A8 e(− λ8 t)
9
I9IIM (t) A9 e(− λ9 t)

∫ T
1( )
A1 R(x)dx 1 − e(− λ1 t) A1
E[U]I1BM (T) λ
I1IIM (T) = 2
= [∫ 0
]2 = 1 ( )
1
( )
ECL (T) T
(− λ1 t)
R(t)dt + τf F(T) + τp R(T) 1 − e + τf 1 − e(− λ1 t) + τp e(− λ1 t)
λ1
0
∫T
1( )
BM A3 R(x)dx A3 1 − e(− λ3 t)
E[U]I (T) λ
I3IIM (T) = 3
= [∫ 0
]2 = 1 ( )
3
( )
ECL2 (T) T
(− λ3 t)
R(t)dt + τf F(T) + τp R(T) 1 − e + τf 1 − e(− λ3 t) + τp e(− λ3 t)
λ3
0
∫T
1( )
BM A5 R(x)dx A5 1 − e(− λ5 t)
IIM E[U]I5 (T) 0 λ5
I5 (T) = = [∫ ]2 = 1 ( ) ( )
ECL2 (T) T
R(t)dt + τf F(T) + τp R(T) 1 − e(− λ5 t) + τf 1 − e(− λ5 t) + τp e(− λ5 t)
λ5
0
∫T
1( )
A 7 R(x)dx A7 1 − e(− λ7 t)
IIM E[U]I7BM (T) 0 λ7
I7 (T) = = [∫ ]2 = 1 ( ) ( )
ECL2 (T) T
R(t)dt + τf F(T) + τp R(T) 1 − e(− λ7 t) + τf 1 − e(− λ7 t) + τp e(− λ7 t)
λ7
0

12
L. Chen et al. Reliability Engineering and System Safety 222 (2022) 108435

( )
J∞ (t)t t cf F1 (t) + cp R1 (t)
t cf 1 − e(− λ1 t) + cp e(− λ1 t)
I1BM,c (t) = = ⋅∫ ⋅ =
A1 A1 T
A1 1 ( ) ( )
R1 (ξ)dξ + τf F1 (t) + τp R1 (t) 1 − e(− λ1 t) + τf 1 − e(− λ1 t) + τp e(− λ1 t)

0
λ1
( )
J∞ (t)t t c F
f 3 (t) + c R
p 3 (t) t cf 1 − e(− λ3 t) + cp e(− λ3 t)
I3BM,c (t) = = ⋅∫ T = ⋅ ( ) ( )
A3 A3 A3 1
R3 (ξ)dξ + τf F3 (t) + τp R3 (t) 1 − e(− λ3 t) + τf 1 − e(− λ3 t) + τp e(− λ3 t)

0
λ3
( )
J∞ (t)t t cf F5 (t) + cp R5 (t) t cf 1 − e(− λ5 t) + cp e(− λ5 t)
I5BM,c (t) = = ⋅∫ T = ⋅ ( ) ( )
A5 A5 A5 1
R5 (ξ)dξ + τf F5 (t) + τp R5 (t) 1 − e(− λ5 t) + τf 1 − e(− λ5 t) + τp e(− λ5 t)

0
λ5

According to (4) and (25), the maintenance cost-based Birnbaum


importance measures of the four components 1, 3, 5 and 7 are

J∞ (t)t t cf F7 (t) + cp R7 (t)


I7BM,c (t) = = ⋅∫ T
A7 A7 R7 (ξ)dξ + τf F7 (t) + τp R7 (t)
0
( )
t cf 1 − e(− λ7 t) + cp e(− λ7 t)
= ⋅1
A7 λ7 (1 − e(− λ7 t) ) + τf (1 − e(− λ7 t) ) + τp e(− λ7 t)

where

( λ5 t) (λ2 +λ5 )t)


)
e(− − e(− − e(− (λ2 +λ3 +λ4 +λ5 +λ6 )t) + e(− (λ2 +λ3 +λ5 +λ6 )t) +
A1 = − (λ2 +λ4 +λ5 +λ6 )t) (− (λ3 +λ4 +λ5 +λ6 )t) (− (λ3 +λ5 +λ6 )t) (− (λ4 +λ5 +λ6 )t) e(− (λ7 +λ8 +λ9 )t)
e(− +e − e − e

( λ6 t) (λ4 +λ6 )t)


)
e(− − e(− − e(− (λ1 +λ2 +λ4 +λ5 +λ6 )t) + e(− (λ1 +λ2 +λ5 +λ6 )t) +
A3 = − (λ1 +λ4 +λ5 +λ6 )t) e(− (λ7 +λ8 +λ9 )t)
e(− + e(− (λ2 +λ4 +λ5 +λ6 )t) − e(− (λ1 +λ5 +λ6 )t) − e(− (λ2 +λ5 +λ6 )t)

⎛ ⎞
e(− λ1 t) + e(− λ2 t) − e(− (λ1 +λ2 )t) − e(− (λ1 +λ2 +λ3 +λ4 +λ6 )t) + e(− (λ1 +λ2 +λ3 +λ6 )t)
A5 = ⎝ +e 1 2 4 6 + e 1 3 4 6 − e 1 3 6 − e 1 4 6 ⎠e(−
(− (λ +λ +λ +λ )t) (− (λ +λ +λ +λ )t) (− (λ +λ +λ )t) (− (λ +λ +λ )t) (λ7 +λ8 +λ9 )t)

+e(− (λ2 +λ3 +λ4 +λ6 )t) − e(− (λ2 +λ3 +λ6 )t) − e(− (λ2 +λ4 +λ6 )t)

⎛ ⎞
e(− (λ1 +λ5 )t) + e(− (λ2 +λ5 )t) − e(− (λ1 +λ2 +λ5 )t) + e(− (λ3 +λ6 )t) + e(− (λ4 +λ6 )t)
⎜ − e(− (λ3 +λ4 +λ6 )t) − e(− (λ1 +λ2 +λ3 +λ4 +λ5 +λ6 )t) + e(− (λ1 +λ2 +λ3 +λ5 +λ6 )t) + ⎟ (−

A7 = ⎝ (− (λ1 +λ2 +λ4 +λ5 +λ6 )t) ⎟e (λ8 +λ9 )t)
e + e(− (λ1 +λ3 +λ4 +λ5 +λ6 )t) − e(− (λ1 +λ3 +λ5 +λ6 )t) ⎠
(λ1 +λ4 +λ5 +λ6 )t) (λ2 +λ3 +λ4 +λ5 +λ6 )t) (λ2 +λ3 +λ5 +λ6 )t) (λ2 +λ4 +λ5 +λ6 )t)
− e(− + e(− − e(− − e(−

13
L. Chen et al. Reliability Engineering and System Safety 222 (2022) 108435

According to (26), the MCIM of the four components 1, 3, 5 and 7 are

[∫ T ]
[ ]
t cf F(T) + cp R(T) R(t)dt + τf F(T) + τp R(T)
ECC(T)ECL(T) 0
I1MCIM (t) = t⋅ = ∫T
E[U]I1BM (T)
A1 R(x)dx
0
[ ]
[ ( λ1 t)
) ] 1( (− λ1 t)
) ( )
t cf 1 − e(− + cp e 1 − e(− λ1 t) + τf 1 − e(− λ1 t) + τp e(− λ1 t)
λ1
=
1( )
A1 1 − e(− λ1 t)
λ1
[∫ T ]
[ ]
t cf F(T) + cp R(T) R(t)dt + τf F(T) + τp R(T)
ECC(T)ECL(T) 0
I3MCIM (t) = t⋅ = ∫T
E[U]I3BM (T)
A3 R(x)dx
0
[ ]
[ ( λ3 t)
) ] 1( ) ( )
t cf 1 − e(− + cp e(− λ3 t) 1 − e(− λ3 t) + τf 1 − e(− λ3 t)
+ τp e(− λ3 t)
λ3
=
1( )
A3 1 − e(− λ3 t)
λ3

[∫ T ]
[ ]
t cf F(T) + cp R(T) R(t)dt + τf F(T) + τp R(T)
ECC(T)ECL(T) 0
I5MCIM (t) = t⋅ = ∫T
E[U]I5BM (T)
A5 R(x)dx
0
[ ]
[ ( λ5 t)
) ] 1( ) ( )
t cf 1 − e(− + cp e(− λ5 t) 1 − e(− λ5 t) + τf 1 − e(− λ5 t) + τp e(− λ5 t)
λ5
=
1( )
A5 1 − e(− λ5 t)
λ5
[∫ T ]
[ ]
t cf F(T) + cp R(T) R(t)dt + τf F(T) + τp R(T)
ECC(T)ECL(T) 0
I7MCIM (t) = t⋅ = ∫T
E[U]I7BM (T)
A7 R(x)dx
0
[ ]
[ ( λ7 t)
) (− λ7 t)
] 1( ) ( )
t cf 1 − e(− + cp e 1 − e(− λ7 t) + τf 1 − e(− λ7 t)
+ τp e(− λ7 t)
λ7
=
1( )
A7 1 − e(− λ7 t)
λ7

importance does not consider the impact of component maintenance on


Assuming the time interval is [0, 8], c1 = 15, c2 = 15, c3 = 15, c4 = the component importance. Only the system’s mechanism and compo­
15, c5 = 10, c6 = 15, c7 = 5, c8 = 30, c9 = 30 (all unit prices are taken in nent failure intensity affect the ranking of Birnbaum’s importance.
K RMB). The cost of system failure caused by components 7, 8, and 9 are We show the importance analysis results using IIM, maintenance
cs,7 = 50, cs,8 = 300, cs,9 = 300, which are much higher than the cost-based Birnbaum importance, and MCIM in Figs. 8-10. The IIM
maintenance cost associated with maintaining single component. Pre­ changes of components are shown in Fig. 8. While the failure intensities
ventive maintenance cost of components 1, 3, 5 and 7 arecp1 = 1, cp3 = of components 1 and 2 are the same, component 1 adopts the age-based
1, cp5 = 0.8, cp7 = 0.4, and their preventive maintenance time isτp1 = τp3 replacement strategy, and component 2 adopts the failure- triggered
= τp5 = τp7 = 0.1. The maintenance cost of components 1, 3, 5 and 7 replacement strategy.
following their failure are cf1 = 5, cf3 = 5, cf5 = 4, cf7 = 2, and their From Fig. 8, the IIM of component 1 increases until it reaches a
maintenance time after the failure areτf1 = τf3 = τf5 = τf7 = 0.5. Thus, certain height and then slowly decreases, while component 2 slowly
the time and cost of post-failure maintenance are 5 times that of pre­ increases and then slowly decreases. This is because the maximum age of
ventive maintenance. Component failure intensity λ1 = λ2 = λ3 = λ4 = the components under the age-based replacement strategy is limited to
0.1, λ5 = λ6 = 0.08, λ7 = 0.05, λ8 = λ9 = 0.08 per year. the age threshold T, and the components exceeding T have undergone
As discussed at the beginning of Section 7.2, the Birnbaum impor­ preventive maintenance. Therefore, the maintenance of the components
tance values of the components under the hybrid maintenance strategy under the age-based replacement strategy is more frequent and the
are the same as those under the single failure-triggered replacement failure intensity is higher than that under the failure-triggered replace­
strategy presented in Section 7.1 (Fig. 4). This is because Birnbaum ment strategy, resulting in greater importance values.

14
L. Chen et al. Reliability Engineering and System Safety 222 (2022) 108435

The maintenance cost-based Birnbaum importance values of the nine Acknowledgments


components are shown in Fig. 9.
From Fig. 9, the importance values of components 1, 3 drop from a The authors gratefully acknowledge the financial supports for this
high level to a certain height over time, and then begin to rise. Similarly, research from the National Natural Science Foundation of China (nos.
the importance values of components 2, 4 drop from a high level to a 72071182, 61807031, U1904211), the Ministry of education’s human­
certain height over time, and then begin to increase rapidly. Over time, ities and social sciences planning fund (no. 20YJA630012), the Program
the importance values of components 2, 4 exceed those of components 1, for Science&Technology Innovation Talents in Universities of Henan
3. It shows that the initial maintenance cost of the system under the age- Province (no. 22HASTIT022), the Program for Young Backbone Teach­
based replacement strategy has a great influence on the importance, but ers in Universities of Henan Province (no. 2021GGJS007), Key Science
in the long run, this influence becomes gradually smaller. This conclu­ and Technology Program of Henan Province (nos. 222102520019,
sion can also be reflected by comparing the importance trends of 132102210560, 162102210004), and Key Science and Technology
component 5 and component 6. Component 7 has a low maintenance Special Projects of Henan Province (no. 201111210800).
cost, so it has a low importance level, and its importance value does not
change much over time. The importance values of components 8 and 9 References
tend to increase exponentially over time, and eventually rise to the
highest level during the time interval considered in this case study. [1] Birnbaum ZW. On the importance of different components in a multi-component
system. New York: Academic Press; 1969.
Fig. 10 shows that the trends of components MCIM over time, which
[2] Dui HY, Si SB, Wu SM, Yam RCM. An importance measure for multistate systems
are similar to those in Fig. 9 using the maintenance cost-based Birnbaum with external factors. Reliab Eng Syst Saf 2017;167:49–57.
importance measure. [3] Si SB, Levitin G, Dui HY, Sun SD. Component state-based integrated importance
measure for multi-state systems. Reliab Eng Syst Saf 2013;116:75–83.
[4] Xing LD, Dugan JB. Analysis of generalized phased-mission system reliability,
8. Conclusions and future work performance, and sensitivity. IEEE Trans Reliab 2002;51(2):199–211.
[5] Dui HY, Zheng XQ, Wu SM. Resilience analysis of maritime transportation systems
based on importance measures. Reliab Eng Syst Saf 2021;209:107461.
Considering the influence of different maintenance strategies and [6] Yeh W-C. Novel algorithm for computing all-pairs homogeneity-arc binary-state
costs of system components on component importance, this paper con­ undirected network reliability. Reliab Eng Syst Saf 2021;216:107950.
tributes by proposing a new measure of MCIM and its evaluation method [7] Yeh W-C. Novel binary-addition tree algorithm (BAT) for binary-state network
reliability problem. Reliab Eng Syst Saf 2021;208:107448.
under different maintenance strategy conditions, including the failure- [8] Yeh W-C. A quick BAT for evaluating the reliability of binary-state networks. Reliab
triggered replacement strategy, the age-based replacement strategy, Eng Syst Saf 2021;216:107917.
and the hybrid strategy. In contrast to the traditional measures that [9] Wei GZ, Zhao XJ, He SG, He Z. Reliability modeling with condition-based
maintenance for binary-state deteriorating systems considering zoned shock
reflect only the failure intensity and structure location of the compo­ effects. Comput Ind Eng 2019;130:282–97.
nents in their importance values, the proposed importance measures [10] Levitin G, Lisnianski A. Importance and sensitivity analysis of multi-state systems
also take into account the component maintenance cost as well as using the universal generating function method. Reliab Eng Syst Saf 1999;65(3):
271–82.
maintenance cost associated with the system failure, providing more [11] Levitin G, Podofillini L, Zio E. Generalised importance measures for multi-state
effective guidance in the practical system maintaining activities. As elements based on performance level restrictions. Reliab Eng Syst Saf 2003;82(3):
revealed from the case study of an example propeller plane system, we 287–98.
[12] Levitin G. Protection Survivability Importance in Systems with Multilevel
found that 1) as compared with the failure-triggered strategy, under the
Protection. Qual Reliab Eng Int 2004;20:727–38.
age-based replacement strategy, the system components undergo more [13] Levitin G, Ben-Haim H. Importance of protections against intentional attacks.
frequent maintenance and have higher costs of maintenance, and 2) in Reliab Eng Syst Saf 2008;93(4):639–46.
the case of the age-based replacement strategy being adopted, the initial [14] Wu SM. Joint importance of multistate systems. Comput Ind Eng 2005;49(1):
63–75.
maintenance cost of system components has a great influence on its [15] Dui HY, Tian TZ, Zhao JB, Wu SM. Comparing with the joint importance under
importance, but this influence shows a diminishing trend over time. consideration of consecutive-k-out-of-n system structure changes. Reliab Eng Syst
In this work, we are considering only binary-state components, each Saf 2022;219:108255.
[16] Shrestha A, Xing LD, Coit DW. An efficient multistate multivalued decision
being characterized by its failure intensity. In the future, we are inter­ diagram-based approach for multistate system sensitivity analysis. IEEE Trans
ested in extending the proposed importance measures and analysis to Reliab 2010;59:581–92.
multi-state systems that perform at multiple levels ranging from perfect [17] Xiahou TF, Liu Y, Jiang T. Extended composite importance measures for multi-state
systems with epistemic uncertainty of state assignment. Mech Syst Sig Process
function to complete failure. This can be done by considering the state 2018;109:305–29.
distribution probability and state transition probability of a component [18] Xu ZP, Ramirez-Marquez JE, Liu Y, Xiahou TF. A new resilience-based component
in the multi-state system. It is used to analyze how the state transition of importance measure for multi-state networks. Reliab Eng Syst Saf 2020;193:
106591.
the component affects the overall performance of the system. [19] Levitin G, Lisnianski A. Optimization of imperfect preventive maintenance for
multi-state systems. Reliab Eng Syst Saf 2000;67(2):193–203.
[20] Levitin G, Finkelstein M, Dai YS. Optimization of cyclic preventive replacement in
Author statement
homogeneous warm-standby system with reusable elements exposed to shocks.
Reliab Eng Syst Saf 2021;207:107351.
Liwei Chen and Chunchun Cheng performed the experiments and [21] Levitin G, Xing LD, Xiang YP. Optimal multiple replacement and maintenance
scheduling in two-unit systems. Reliab Eng Syst Saf 2021;213:107803.
analyzed the data; Hongyan Dui and Liudong Xing proposed the idea of
[22] Wu SM, Chen Y, Wu QT, Wang ZL. Linking component importance to optimisation
this paper and revised the methodology and model; All authors have of preventive maintenance policy. Reliab Eng Syst Saf 2016;146:26–32.
contributed to the editing and proofreading of this paper. [23] Han X, Wang ZL, Xie M, He YH, Li Y, Wang WZ. Remaining useful life prediction
and predictive maintenance strategies for multi-state manufacturing systems
considering functional dependence. Reliab Eng Syst Saf 2021;210:107560.
[24] de Pater I, Mitici M. Predictive maintenance for multi-component systems of
Declaration of Competing Interest repairables with Remaining-Useful-Life prognostics and a limited stock of spare
components. Reliab Eng Syst Saf 2021;214:107761.
[25] Zhang N, Cai KQ, Zhang J, Wang T. A condition-based maintenance policy
The authors report no conflict of interest and have received no considering failure dependence and imperfect inspection for a two-component
payment in preparation of this manuscript. system. Reliab Eng Syst Saf 2022;217:108069.

15
L. Chen et al. Reliability Engineering and System Safety 222 (2022) 108435

[26] Zhang C, Chen RT, Wang SP, Dui HY, Zhang YD. Resilience efficiency importance [30] Dui HY, Chen SS, Zhou YJ, Wu SM. Maintenance analysis of transportation
measure for the selection of a component maintenance strategy to improve system networks by the traffic transfer principle considering node idle capacity. Reliab
performance recovery. Reliab Eng Syst Saf 2022;217:108070. Eng Syst Saf 2022;221:108386.
[27] Nguyen K-A, Do P, Grall A. Joint predictive maintenance and inventory strategy for [31] Si SB, Dui HY, Zhao XB, Zhang SG, Sun SD. Integrated importance measure of
multi-component systems using Birnbaum’s structural importance. Reliab Eng Syst component states based on loss of system performance. IEEE Trans Reliab 2012;61
Saf 2017;168:249–61. (1):192–202.
[28] Chen LW, Gao YS, Dui Y, Xing LD. Importance measure-based maintenance [32] Dui HY, Xu Z, Chen LW, Xing LD, Liu B. Data-driven maintenance priority and
optimization strategy for pod slewing system. Reliab Eng Syst Saf 2021;216: resilience evaluation of performance loss in a main coolant system. Mathematics
108001. 2022;10(4):563.
[29] Wu SM, Coolen FPA. A cost-based importance measure for system components: an
extension of the Birnbaum importance. Eur J Oper Res 2013;225(1):189–95.

16

You might also like