Академический Документы
Профессиональный Документы
Культура Документы
a r t i c l e i n f o a b s t r a c t
Keywords: Preventive maintenance may be performed on a few selected components when a component fails. Importance
Preventive maintenance measures can be used to identify the most important component that requires maintenance. However, this process
Maintenance cost involves two problems: (a) the preventive maintenance time of the selected component may be bigger than the
System reliability
maintenance time of the failed component; (b) the most important component may incur the highest maintenance
Importance measure
cost. Traditional importance measures do not consider the possible effect of maintenance time and cost, which
significantly affect the improvement of system reliability. Given the joint effect of component maintenance cost
and time on system reliability, this study proposes a cost-based integrated importance measure (IIM) to identify
the component or group of components that may be selected for preventive maintenance. The characteristics of
cost-based IIM are examined to determine the relationships among failure rates, shape parameters, and the scale
parameters of different components. Finally, an application to a wind turbine system is used to illustrate its usage.
© 2017 Elsevier Ltd. All rights reserved.
∗
Corresponding author.
E-mail addresses: duihongyan@zzu.edu.cn (H. Dui), sisb@nwpu.edu.cn (S. Si), mery@cityu.edu.hk (R.C.M. Yam).
http://dx.doi.org/10.1016/j.ress.2017.05.025
Available online xxx
0951-8320/© 2017 Elsevier Ltd. All rights reserved.
Please cite this article as: H. Dui et al., Reliability Engineering and System Safety (2017), http://dx.doi.org/10.1016/j.ress.2017.05.025
JID: RESS
ARTICLE IN PRESS [m5GeSdc;May 12, 2017;12:34]
H. Dui et al. Reliability Engineering and System Safety 000 (2017) 1–7
[17,30] ignore maintenance time and maintenance cost. According to Eqs. (1) and (2), the reliability changes of a component
• Explanation for maintenance time or a few components will change system reliability.
When Δt → 0, we obtain
Preventive maintenance is performed on two selected components 𝑛
𝑅(𝑡 + Δ𝑡) − 𝑅(𝑡) 𝑑𝑅(𝑡) ∑ 𝑑 𝑅𝑖 (𝑡) 𝜕𝑅(𝑡)
when a component fails. We assume that the maintenance time of the lim = = . (3)
failed component is t, and the preventive maintenance time of the two
Δ𝑡→0 Δ𝑡 𝑑𝑡 𝑖=1
𝑑𝑡 𝜕 𝑅𝑖 (𝑡)
selected components is t1 , t2 , respectively. If t1 > t (t2 > t) and the se- Eq. (3) is the expression of the total derivative of the system relia-
lected components are the critical components, then the selected com- bility function attributed to time t.
ponents for preventive maintenance are unsuitable. This finding is at- The latter part of the right side of Eq. (3) is the Birnbaum importance
tributed to the fact that the selected components affect the boot time of component i, that is, 𝐼𝑖𝐵𝑀 (𝑡) = 𝜕𝜕𝑅 (𝑡)
𝑅 (𝑡)
, which evaluates the effect of
𝑖
of the system after maintaining the failed component. Furthermore, if
component reliability on system reliability.
t1 > 2t2 , then a new component can be performed during preventive
Given the joint effect of probability distributions and transition rates
maintenance to complement spare time t1 –t2 .
of the component states on the system reliability, Si et al. [26] propose
• Explanation for maintenance cost the IIM of component i, as shown in Eq. (4):
𝜕𝑅(𝑡)
How do we determine the components that should be selected if sys- 𝐼𝑖𝐼 𝐼 𝑀 (𝑡) = 𝑅𝑖 (𝑡)𝜆𝑖 (𝑡) , (4)
𝜕 𝑅𝑖 (𝑡)
tem reliability improvements are similar after maintaining two different
components? The same reliability improvement of the system may be where 𝜆i (t) is the failure rate of component i.
𝑑 𝑅 (𝑡)∕𝑑𝑡
related to different maintenance costs of components. For example, the Given that 𝜆𝑖 (𝑡) = − 𝑅𝑖 (𝑡) , the expression of IIM can be converted
𝑖
reliability improvement of a water pump in a water supply system will into
lead to significant improvement of system reliability. However, the cost 𝑑 𝑅𝑖 (𝑡) 𝜕𝑅(𝑡)
of maintaining a water pump is bigger than the cost of maintaining a 𝐼𝑖𝐼 𝐼 𝑀 (𝑡) = − . (5)
𝑑𝑡 𝜕 𝑅𝑖 (𝑡)
switchboard.
Which component or group of components should be maintained un- Based on Eqs. (3) and (5),
der these circumstances? In preventive maintenance, the IIM does not 𝑛 𝑛
𝑑𝑅(𝑡) ∑ 𝑑 𝑅𝑖 (𝑡) 𝜕𝑅(𝑡) ∑
consider the possible effect of maintenance time and cost on system re- = =− 𝐼𝑖𝐼 𝐼 𝑀 (𝑡). (6)
𝑑𝑡 𝑖=1
𝑑𝑡 𝜕 𝑅𝑖 (𝑡) 𝑖=1
liability. However, this approach can identify the component or group
of components that may be selected for maintenance. This study con- According to Eq. (6), the opposite of the IIM value of component
siders the joint effect of component maintenance cost and time on the i is a weight of the total change in R(t) due to component i. 𝑑𝑅𝑑𝑡(𝑡) < 0
2
JID: RESS
ARTICLE IN PRESS [m5GeSdc;May 12, 2017;12:34]
H. Dui et al. Reliability Engineering and System Safety 000 (2017) 1–7
because system reliability R(t) is a decreasing function with time t. The Based on Definition 1, the component with minimum cost-based IIM
IIM of component I represents the change of system reliability in unit ( min {𝐼𝑖𝐼 𝐼 𝑀 ,𝐶 (𝑡)}) should be first performed on the preventive main-
𝑖=1,2,…,𝑛
time due to component i. tenance.
IIM is additive according to Eq. (6). Suppose we are interested in the When a component fails, the 𝐼𝑖𝐼 𝐼 𝑀 ,𝐶 (𝑡) can be used to identify the
effect of a group of components i1 , i2 , …, im on the system reliability in components that may be selected for preventive maintenance. By means
unit time. Then, of 𝐼𝑖𝐼 𝐼 𝑀 ,𝐶 (𝑡), the reliability of the system can be maximally improved in
𝑑 𝑅𝑖1 (𝑡) 𝜕𝑅(𝑡) 𝑑 𝑅𝑖2 (𝑡) 𝜕𝑅(𝑡) 𝑑 𝑅𝑖𝑚 (𝑡) 𝜕𝑅(𝑡) the same repair time. Maintenance cost is at the minimum.
𝐼𝑖𝐼 𝐼,𝑖𝑀,…,𝑖 (𝑡) = − − −⋯− When a critical component fails, the system also fails. Then, 𝐼𝑖𝐼 𝐼 𝑀 ,𝐶 (𝑡)
1 2 𝑚 𝑑𝑡 𝜕 𝑅𝑖1 (𝑡) 𝑑𝑡 𝜕 𝑅𝑖2 (𝑡) 𝑑𝑡 𝜕 𝑅𝑖𝑚 (𝑡)
can be performed on all other components. When a non-critical compo-
= 𝐼𝑖𝐼 𝐼 𝑀 (𝑡) + 𝐼𝑖𝐼 𝐼 𝑀 (𝑡) + ⋯ + 𝐼𝑖𝐼 𝐼 𝑀 (𝑡). (7) nent has failed, the system does not fail. 𝐼𝑖𝐼 𝐼 𝑀 ,𝐶 (𝑡) can be performed on
1 2 𝑚
all other non-critical components to ensure smooth system operation.
Based on Eqs. (1), (2), (6), and (7), the change of system reliability When a group of components needs to be identified for preventive
is attributed to the change of a component or a group of components. maintenance, the following definition is obtained based on Eq. (7).
ventive maintenance time of component 4 is bigger than the repair time 𝐶(𝑖1 , 𝑖2 , … , 𝑖𝑚 , 𝑡) is the cost for maintaining multiple components i1 ,
𝐶(𝑖1 ,𝑖2 ,…,𝑖𝑚 ,𝑡)
of component 1, component 4 is rendered unsuitable because the long i2 , …, im . According to Eq. (7), 𝐼𝑖𝐼 𝐼,𝑖𝑀,…,,𝐶
𝑖 (𝑡) = 𝐼 𝐼 𝑀 𝐼𝐼𝑀 𝐼𝐼𝑀 .
1 2 𝑚 𝐼𝑖 (𝑡)+𝐼𝑖 (𝑡)+⋯+𝐼𝑖 (𝑡)
repair time may result in high system unavailability, thereby increasing 1 2 𝑚
According to Definition 2, multi-components can be identified to per-
cost. Moreover, if the maintenance costs of components 5, 6, 7, and 8
form preventive maintenance, improve system reliability, and further
are different, the order of the selected component for preventive main-
reduce the maintenance cost.
tenance is also different.
The number of components for preventive maintenance [31] is re-
The preceding analysis shows that the importance of component
lated to the constraint of total maintenance cost. Preventive mainte-
maintenance priority of Wu cannot be used in these scenarios. We pro-
nance can be performed on all other components when a critical compo-
vide the extended importance in the following discussion.
nent fails. When a non-critical component fails, preventive maintenance
can be performed on all other non-critical components. We assume that
3.1. Considering change in system reliability change during repair
the total maintenance cost is C, and the order of the identified com-
ponents performed on preventive maintenance is components 1, 2, …,
According to Section 2, system reliability in repair time Δt of a failed
n. The number of components for preventive maintenance is given as
component may be approximately reduced by Δ𝑡𝐼𝑖𝐼 𝐼 𝑀 (𝑡) because of com- 𝑁 𝑁+1
∑ ∑
ponent i based on the Taylor expansion of R(t). The component that 𝑁 = {𝑁| 𝐶(𝑖, 𝑡) ≤ 𝐶 ≤ 𝐶(𝑖, 𝑡)}.
brings the maximum decrement of system reliability should be first per- 𝑖=1 𝑖=1
The failure of a component may lead to system failure. The cost of Example 1. In a series system comprising n components, system relia-
𝑛
∏ 𝑛
∏ 𝑛
∑
component failure of component i includes the cost of system mainte- 𝑡 𝑡
bility is 𝑅(𝑡) = 𝑅𝑖 (𝑡) = exp(− ∫0 𝜆𝑖 (𝑢)𝑑𝑢) = exp(− ∫0 𝜆𝑖 (𝑢)𝑑𝑢).
nance, failure cost of component i, and maintenance cost of component 𝑖=1 𝑖=1 𝑖=1
i [20].
We assume that the maintenance costs of all components are
Total maintenance cost attributed to component i, C(i, t) is given by
the same, 𝐶(𝑖, 𝑡) = 𝐶(𝑡). Then, the cost-based IIM of component i is
𝐶(𝑡) 𝐶(𝑡)
𝐼𝑖𝐼 𝐼 𝑀 ,𝐶 (𝑡) = 𝑡 𝜕𝑅(𝑡) = 𝑛
∑
.
𝐶(𝑖, 𝑡) = 𝐶𝑠,𝑖 (𝑡) + 𝐶𝑖 (𝑡), (8) 𝜆𝑖 (𝑡) exp(− ∫0 𝜆𝑖 (𝑢)𝑑𝑢) 𝜕 𝑅 (𝑡)
𝑖 𝜆𝑖 (𝑡) exp(−
𝑡
∫0 𝜆𝑘 (𝑢)𝑑𝑢)
𝑘=1
where Ci (t) is the cost of maintaining component i at time t, and Cs, i (t) A comparison of the cost-based IIM of two different components i
is the cost of maintaining the system caused by the failure of component and j is given as:
i at time t, which means economic loss because of system unavailability. 𝑛
∑ 𝑡
𝜆𝑗 (𝑡) exp(− ∫0 𝜆𝑘 (𝑢)𝑑𝑢)
In Eq. (8), component failure does not necessarily lead to system 𝐼𝑖𝐼 𝐼 𝑀 ,𝐶 (𝑡) 𝑘=1 𝜆𝑗 (𝑡)
failure. For example, in a parallel system, one component failure cannot = 𝑛
= .
𝐼𝑗𝐼 𝐼 𝑀 ,𝐶 (𝑡) ∑ 𝑡 𝜆𝑖 (𝑡)
lead to the system failure. 𝐶𝑠,𝑖 (𝑡) = 0 when component i fails and if the 𝜆𝑖 (𝑡) exp(− ∫0 𝜆𝑘 (𝑢)𝑑𝑢)
system still works. 𝑘=1
3
JID: RESS
ARTICLE IN PRESS [m5GeSdc;May 12, 2017;12:34]
H. Dui et al. Reliability Engineering and System Safety 000 (2017) 1–7
𝛼 −1
(1+ 𝑢 ) 𝑗 𝑒−𝛽𝑗 𝑢 𝑑𝑢
∞
𝜆 (𝑡) ∫ Component 1
𝜆𝑖 (𝑡) > 𝜆𝑗 (𝑡) ⇔ 𝜆 𝑖 (𝑡) > 1 ⇔ 0∞ 𝑡𝑢 𝛼𝑖 −1 −𝛽 𝑢 > 1 ⇔
𝑗 ∫0 (1+ 𝑡 ) 𝑒 𝑖 𝑑𝑢
( ) Bearing A Component 3 Component 4 Component 5
∞ 𝑢 𝛼𝑗 −1 −𝛽𝑗 𝑢 𝑢 𝛼𝑖 −1 −𝛽𝑖 𝑢
∫0 (1 + 𝑡 ) 𝑒 − (1 + 𝑡 ) 𝑒 𝑑𝑢 > 0.
Main Shaft Gearbox Generator
When 𝛼 i < 𝛼 j , 𝛽 i > 𝛽 j , we have 𝜆i (t) > 𝜆j (t). Therefore, we can under-
Bearing B
stand that if 𝛼 i < 𝛼 j , 𝛽 i > 𝛽 j , then 𝐼𝑖𝐼 𝐼 𝑀 ,𝐶 (𝑡) < 𝐼𝑗𝐼 𝐼 𝑀 ,𝐶 (𝑡).
Component 2
(2) In a series system, when the lifetime distribution function of
component i follows Weibull distribution with shape parameter Fig. 1. Subsystem of a wind turbine.
4
JID: RESS
ARTICLE IN PRESS [m5GeSdc;May 12, 2017;12:34]
H. Dui et al. Reliability Engineering and System Safety 000 (2017) 1–7
-5
x 10
9
component1
8 component2
component3
7
component4
6
component5
5
IIM
0
0 1000 2000 3000 4000 5000 6000 7000
t
10
x 10
3
component1
component2
2.5 component3
component4
2
component5
cost-based IIM
1.5
0.5
0
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
t
components 1 and 2 is higher than that of components 4 and 5 based the analysis in reference [20]. In reference [20], cost-based importance
on Eq. (12); hence, the IIM curves of components 1 and 2 intersect with is the cost of component i divided by the reliability of component i, and
that of components 4 and 5. The relationships among the failure rates the reliability of component 3 > the reliability of component 5 > the
of components 3, 4, and 5 are 𝜆3 (t) > 𝜆5 (t) > 𝜆4 (t). Therefore, the or- reliability of component 4. In this study, the importance is the cost of
der of the effects of components 3, 4, and 5 on the system reliability is component i divided by the IIM of component i, and the IIM of com-
component 3, 5, and 4. ponent 3 > the IIM of component 5 > the IIM of component 4. Hence,
Based on Eq. (9), the cost-based IIM of component i is 𝐼𝑖𝐼 𝐼 𝑀 ,𝐶 (𝑡) = they result in the same conclusion of the order of cost-based importance.
𝐶(𝑖) However, the two importance measures are different. According to the
,𝑖 = 1, 2, 3, 4, 5, as shown in Fig 3.
𝐼𝑖𝐼 𝐼 𝑀 (𝑡)
IIM equation, the value of IIM value changes when the failure rate of
According to Eq. (14), C(4) > C(5) > C(3). From Fig. 2, we obtain
component changes. For example, when 𝜆3 (t) ≪ 𝜆5 (t) ≪ 𝜆4 (t), the IIM
𝐼3𝐼 𝐼 𝑀 (𝑡) > 𝐼5𝐼 𝐼 𝑀 (𝑡) > 𝐼4𝐼 𝐼 𝑀 (𝑡). The component with the minimum cost-
of component 3 < the IIM of component 5 < the IIM of component 4.
based IIM should be first performed on the maintenance. Therefore, the
This finding will result in conclusions that differ from those in reference
maintenance order of components is component 3, 5, and 4, as shown in
[20].
Fig. 3. 𝐼1𝐼 𝐼 𝑀 ,𝐶 (𝑡) and 𝐼2𝐼 𝐼 𝑀 ,𝐶 (𝑡) are highest because the effects of compo-
Based on Eq. (10), the cost-based IIM of a group of components i1 ,
nents 1 and 2 on the system reliability are smallest at the beginning of
i2 is
the system lifetime. With the increase of time, the effect of components
𝐶(1) + 𝐶(2) 𝐶(3) + 𝐶(4)
4 and 5 on the system reliability gradually reduces, as shown in Fig 2; 𝐼1𝐼,𝐼2𝑀 ,𝐶 (𝑡) = , 𝐼 𝐼 𝐼 𝑀 ,𝐶 (𝑡) = ,
𝐼1𝐼 𝐼 𝑀 (𝑡) + 𝐼2𝐼 𝐼 𝑀 (𝑡) 3,4 𝐼3𝐼 𝐼 𝑀 (𝑡) + 𝐼4𝐼 𝐼 𝑀 (𝑡)
hence, the cost-based IIM of components 4 and 5 increases as shown in 𝐶(3) + 𝐶(5) 𝐶(4) + 𝐶(5)
(16)
Fig. 3. The cost-based IIM curves of components 1 and 2 intersect with 𝐼3𝐼,𝐼5𝑀 ,𝐶 (𝑡) = , 𝐼 𝐼 𝐼 𝑀 ,𝐶 (𝑡) = .
𝐼3𝐼 𝐼 𝑀 (𝑡) + 𝐼5𝐼 𝐼 𝑀 (𝑡) 4,5 𝐼4𝐼 𝐼 𝑀 (𝑡) + 𝐼5𝐼 𝐼 𝑀 (𝑡)
the ones of components 4 and 5. The maintenance order of components
In Fig. 4, we discuss the effect of a group of two components on the
is 3, 1, 2, 5, and 4 because the component with the minimum cost-based
system reliability and identify the group of components that should be
IIM should be first performed on the maintenance.
first selected for preventive maintenance. Components 1 and 2 are in
Fig. 3 shows that the result is the same conclusion of the order of
the parallel part. Thus, components 1 and 2 are in a group, and the rest
the cost-based importance, that is, component 4 > 5 > 3 compared with
of the components can be in a group.
5
JID: RESS
ARTICLE IN PRESS [m5GeSdc;May 12, 2017;12:34]
H. Dui et al. Reliability Engineering and System Safety 000 (2017) 1–7
10
x 10
3
components 1,2
components 3,4
2.5 components 3,5
components 4,5
1.5
0.5
0
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
t
The maintenance costs of components 3 and 5 are small. Fig. 2 shows [2] Kuo W, Zhu X. Importance measures in reliability, risk, and optimization: principles
that the effects of components 3 and 5 on the system reliability are big. and applications. UK: Wiley; 2012.
[3] Kuo W, Zhu X. Some recent advances on importance measures in reliability. IEEE
Therefore, based on Eq. (16), the cost-based IIM of the group of compo- Trans Reliab 2012;61:344–60.
nents 3 and 5 is small as shown in Fig. 4. Components 3 and 5, as well [4] Kuo W, Zhu X. Relations and generalizations of importance measures in reliability.
as components 1 and 2, belong in another group. The effects of com- IEEE Trans Reliab 2012;61:659–74.
[5] Cai B, Liu Y, Fan Q. Multi-source information fusion based fault diagnosis of ground–
ponents 1 and 2 on system reliability are smallest at the beginning of source heat pump using Bayesian network. Appl Energy 2014;114:1–9.
system lifetime as shown Fig. 2. Hence, 𝐼1𝐼,𝐼2𝑀 ,𝐶 (𝑡) is highest at the initial [6] Peng R, Zhai Q, Xing L, Yang J. Reliability of demand-based phased-mission systems
stage. With the increase of time, the effects of components 3 and 5 on subject to fault level coverage. Reliab Eng Syst Saf 2014;121:18–25.
[7] Liu Y, Zuo MJ, Li Y, Huang H. Dynamic reliability assessment for multi-state systems
the system reliability are gradually reduced as shown in Fig. 2. Thus, the utilizing system-level inspection data. IEEE Trans Reliab 2015;64:1287–99.
cost-based IIM of the group of components 3 and 5 increased as shown [8] Jiang T, Liu Y. Parameter inference for non-repairable multi-state system reliability
in Fig. 4. The cost-based IIM curve of the group of components 1 and 2 models by multi-level observation sequences. Reliab Eng Syst Saf 2016. doi:10.1016/
j.ress.2016.11.019.
intersects with the one of the group of components 3 and 5. As shown
[9] Wu S, Chan L. Performance utility-analysis of multi-state systems. IEEE Trans Reliab
in Figs. 3 and 4, 𝐼4𝐼 𝐼 𝑀 ,𝐶 (𝑡) > 𝐼3𝐼,𝐼5𝑀 ,𝐶 (𝑡). Therefore, maintenance order is 2003;52:14–21.
the group of components 1 and 2, the group of components 3 and 5, and [10] Ramirez-Marquez JE, Coit DW. Composite importance measures for multi-state sys-
individual component 4 when the effect of a group of components on tems with multi-state components. IEEE Trans Reliab 2005;54:517–29.
[11] Ramirez-Marquez JE, Coit DW. Multi-state component criticality analysis for relia-
the system reliability is considered. bility improvement in multi-state systems. Reliab Eng Syst Saf 2007;92:1608–19.
[12] Ramirez-Marquez JE, Rocco CM, Gebre BA, Coit DW, Tortorella M. New in-
5. Conclusions sights on multi-state component criticality and importance. Reliab Eng Syst Saf
2006;91:894–904.
[13] Natvig B. On the reduction in remaining system lifetime due to the failure of a
This study considers the joint effect of component reliability and specific component. J Appl Probab 1982;19:642–52.
maintenance cost on system reliability in a maintenance time. This study [14] Natvig B. New light on measures of importance of system components. Scand J Stat
proposes a cost-based IIM to identify the component or group of com- 1985;12:43–54.
[15] Natvig B, Gåsemyr J. New results on the Barlow–Proschan and Natvig measures of
ponents that should be first selected for preventive maintenance. When component importance in nonrepairable and repairable systems. Methodol Comput
a few components fail, importance measures can be used to identify the Appl Probab 2009;11:603–20.
essential components that may affect system reliability. System relia- [16] Natvig B, Eide KA, Gåsemyr J, Huseby AB, Isaksen SL. Simulation based analysis and
an application to an offshore oil and gas production system of the Natvig measures of
bility improvements may be the same while two failed components are component importance in repairable systems. Reliab Eng Syst Saf 2009;94:1629–38.
being repaired. However, the maintenance costs of the two components [17] Gao XL, Cui LR, Li JL. Analysis for joint importance of components in a coherent
are different. The cost-based IIM can be very useful under this circum- system. Eur J Oper Res 2007;182:282–99.
[18] Levitin G, Podofillini L, Zio E. Generalized importance measures for multi-state ele-
stance, and the component with minimum cost-based IIM should be first
ments based on performance level restrictions. Reliab Eng Syst Saf 2003;82:287–98.
selected for maintenance. [19] Peng H, Coit DW, Feng Q. Component reliability criticality or importance measures
Future work can consider the economic dependence of multi- for systems with degrading components. IEEE Trans Reliab 2012;61:391–408.
component systems in preventive maintenance policies. Maintenance [20] Wu S, Coolen FPA. A cost-based importance measure for system components: an
extension of the Birnbaum importance. Eur J Oper Res 2013;225:189–95.
cost can consider different types, such as the constraint of other re- [21] Borgonovo E, Apostolakis GE. A new importance measure for risk-informed deci-
sources (finance and external factors) and opportunity cost. sion-making. Reliab Eng Syst Saf 2001;72:193–212.
[22] Borgonovo E, Aliee H, Glaß M, Teich J. A new time-independent reliability impor-
Acknowledgments tance measure. Eur J Oper Res 2016;254:427–42.
[23] Borgonovo E. Differential, criticality and Birnbaum importance measures: an appli-
cation to basic event, groups and SSCs in event trees and binary decision diagrams.
This work was supported by the National Natural Science Foundation Reliab Eng Syst Saf 2007;92:1458–67.
of China (nos. 71501173, 71271170, and 71631001) and a grant from [24] Zhai Q, Yang J, Xie M, Zhao Y. Generalized moment-independent importance mea-
sures based on Minkowski distance. Eur J Oper Res 2014;239:449–55.
the Research Grants Council of the Hong Kong Special Administrative [25] Tyrväinen T. Risk importance measures in the dynamic flowgraph methodology.
Region, China (Project no. 11203815) (CityU 9042183). Reliab Eng Syst Saf 2013;118:35–50.
[26] Si S, Dui H, Zhao X, Zhang S, Sun S. Integrated importance measure of component
References states based on loss of system performance. IEEE Trans Reliab 2012;61:192–202.
[27] Dui H, Si S, Cui L, Cai Z, Sun S. Component importance for multi-state system life-
[1] Birnbaum ZW. On the importance of different components in a multi-component times with renewal functions. IEEE Trans Reliab 2014;63:105–17.
system. New York: Academic Press; 1969.
6
JID: RESS
ARTICLE IN PRESS [m5GeSdc;May 12, 2017;12:34]
H. Dui et al. Reliability Engineering and System Safety 000 (2017) 1–7
[28] Si S, Levitin G, Dui H, Sun S. Component state-based integrated importance measure [31] Peng R, Xie M, Ng SH, Levitin G. Element maintenance and allocation for linear
for multi-state systems. Reliab Eng Syst Saf 2013;116:75–83. consecutively connected systems. IIE Trans 2012;44:964–73.
[29] Dui H, Si S, Zuo M, Sun S. Semi-Markov process-based integrated importance mea- [32] Andrawus JA. Maintenance optimization for wind turbines Ph.D. thesis. Robert Gor-
sure for multi-state systems. IEEE Trans Reliab 2015;64:754–65. don University; 2008.
[30] Wu S, Chen Y, Wu Q, Wang Z. Linking component importance to optimisation of
preventive maintenance policy. Reliab Eng Syst Saf 2016;146:26–32.