Computers & Industrial Engineering 64 (2013) 459–468

Contents lists available at SciVerse ScienceDirect

Computers & Industrial Engineering

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

Improved cuckoo search for reliability optimization problems

Ehsan Valian a, Saeed Tavakoli a,⇑, Shahram Mohanna a, Atiyeh Haghi b
a
Faculty of Electrical and Computer Engineering, University of Sistan and Baluchestan, Iran
b
Department of Management, University of Sistan and Baluchestan, Iran

a r t i c l e i n f o a b s t r a c t

Article history: An efficient approach to solve engineering optimization problems is the cuckoo search algorithm. It is a
Received 21 May 2011 recently developed meta-heuristic optimization algorithm. Normally, the parameters of the cuckoo
Received in revised form 5 March 2012 search are kept constant. This may result in decreasing the efficiency of the algorithm. To cope with this
Accepted 5 July 2012
issue, the cuckoo search parameters should be tuned properly. In this paper, an improved cuckoo search
Available online 31 August 2012
algorithm, enhancing the accuracy and convergence rate of the cuckoo search algorithm, is presented.
Then, the performance of the proposed algorithm is tested on some complex engineering optimization
Keywords:
problems. They are four well-known reliability optimization problems, a large-scale reliability optimiza-
Cuckoo search
Improved cuckoo search
tion problem as well as a complex system, which is a 15-unit system reliability optimization problem.
Large-scale Finally, the results are compared with those given by several well-known methods. Simulation results
Optimization demonstrate the effectiveness of the proposed algorithm.
Reliability problem Ó 2012 Elsevier Ltd. All rights reserved.

1. Introduction (Prasad and Kuo, 2000). To solve a category of reliability optimiza-

System reliability optimization plays a vital role in real-world
applications and has been studied for decades (Aponte and Sans-
everino, 2007; Chen, 2006; Chern and Jan, 1986; Coelho, 2009;
Elegbede, 2005; Gen and Kim, 1999; Gen and Yun, 2006; Kuo,
2007; Kuo and Prasad, 2000; Marseguerra et al., 2004; Meziane
et al., 2005; Painton and Campbell, 1995; Prasad and Kuo, 2000;
Ramirez-Marquez, 2008; Salazar et al., 2006; Yin et al., 2007; Yoko-
ta et al., 1996; Zou et al., 2011). To be more competitive in the mar-
ket, many designers have worked on improving the reliability of
manufacturing systems or product components. The reliability
problem can usually be formulated as a nonlinear programming
problem with several constraints such as cost, weight, and volume.
To attain higher system reliability, two main ways are usually
used. The first approach is to increase the reliability of system com-
ponents. It may improve the system reliability to some degree,
however, the required reliability enhancement may be never
achievable even though the most currently reliable elements are
used. The second approach is to use redundant components in dif-
ferent subsystems. By using this method, however, the cost,
weight, volume, etc., are also increased. This approach is named
reliability redundancy allocation problem (RAP) (Kuo & Prasad,
2000).
In order to cope with optimization problems arising in system
reliability, important contributions have been made since 1970
⇑ Corresponding author. Tel.: +98 541 805 6248; fax: +98 541 244 3388.
E-mail address: tavakoli@ece.usb.ac.ir (S. Tavakoli).
tion problems with multiple-choice constraints, Chern and Jan
(1986) developed a 2-phase solution method. They presented a
general model that can be stated as the problem of finding the
optimum number of redundancies maximizing the system
reliability. Prasad and Kuo (2000) offered a search method (P and
K-Algorithm) based on lexicographic order, and an upper bound
on the objective function for solving redundancy allocation prob-
lems in coherent systems. The main advantages of the P and
K-Algorithm are its simplicity and its applicability to a wide range
of complex optimization problems arising in system reliability de-
sign (Zou et al., 2011). A penalty guided artificial immune algo-
rithm to solve mixed-integer reliability design problems was
proposed in (Chen, 2006). To efficiently find the feasible optimal/
near optimal solution, it can search over promising feasible and
infeasible regions (Zou et al., 2011). Gen and Yun (2006) employed
a soft computing approach for solving a variety of reliability opti-
mization problems. To prevent the early convergence situation of
its solution, this method combined rough search and local search
techniques (Zou et al., 2011). Moreover, several optimization algo-
rithms based on swarm intelligence, such as particle swarm opti-
mization (Coelho, 2009; Elegbede, 2005; Wu et al., 2010; Yin
et al., 2007), genetic algorithm (Gen and Kim, 1999; Painton and
Campbell, 1995; Marseguerra et al., 2004), evolutionary algorithm
(Aponte and Sanseverino, 2007; Salazar et al., 2006; Ramirez-
Marquez, 2008), ant colony algorithm (Meziane et al., 2005), har-
mony search algorithms (HS) (Zou et al., 2010; Zou et al., 2011)
and artificial bee colony algorithm (Yeh and Hsieh, 2011), have been
employed to solve reliability problems. Kuo (2007) reviewed recent
advances in optimal RAP and summarized several techniques.

0360-8352/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.cie.2012.07.011
460 E. Valian et al. / Computers & Industrial Engineering 64 (2013) 459–468
The Cuckoo Search (CS) developed in 2009 by Yang and Deb
(2009, 2010), is a new meta-heuristic algorithm imitating animal
behavior. The optimal solutions obtained by the CS are far better
than the best solutions obtained by efficient particle swarm opti-
mizers and genetic algorithms (Yang & Deb, 2009). First, an Im-
proved Cuckoo Search (ICS) algorithm for optimization problems
is developed. Then, the performance of the proposed algorithm is
tested on some complex engineering optimization problems.
The paper is organized as follows. In Section 2, the procedure of
cuckoo search algorithm is briefly presented. In Section 3, the
improved cuckoo search algorithm is presented. To apply it to
reliability optimization problems, some preparation works are
done in Section 4. In Section 5, four reliability optimization
problems, a large-scale reliability optimization problem as well
as a complex system are introduced. In Section 6, a number of
simulations are carried out to test the performance and effective-
ness of the proposed algorithm in solving complex reliability
optimization problems. We end this paper with some conclusions
in Section 7.
2. Cuckoo search algorithm
In this section the cuckoo search algorithm is briefly reviewed.
2.1. Cuckoo breeding behavior
The CS was inspired by the obligate brood parasitism of some
cuckoo species by laying their eggs in the nests of host birds. Some
cuckoos have evolved in such a way that female parasitic cuckoos
can imitate the colors and patterns of the eggs of a few chosen host
species. This reduces the probability of the eggs being abandoned
and, therefore, increases their re-productivity (Payne, Sorenson, &
Klitz, 2005). It is worth mentioning that several host birds engage
direct conflict with intruding cuckoos. In this case, if host birds dis-
cover the eggs are not their own, they will either throw them away
or simply abandon their nests and build new ones, elsewhere.
Parasitic cuckoos often choose a nest where the host bird just
laid its own eggs. In general, the cuckoo eggs hatch slightly earlier
than their host eggs. Once the first cuckoo chick is hatched, his first
instinct action is to evict the host eggs by blindly propelling the
eggs out of the nest. This action results in increasing the cuckoo
chick’s share of food provided by its host bird (Payne et al.,
2005). Moreover, studies show that a cuckoo chick can imitate
the call of host chicks to gain access to more feeding opportunity.
The CS models such breeding behavior and, thus, can be applied
to various optimization problems. Yang and Deb (2009, 2010) dis-
covered that the performance of the CS can be improved by using
Lévy Flights instead of simple random walk.
2.2. Lévy flights
In nature, animals search for food in a random or quasi-random
manner. Generally, the foraging path of an animal is effectively a
random walk because the next move is based on both the current
location/state and the transition probability to the next location.
The chosen direction implicitly depends on a probability, which
can be modeled mathematically. Various studies have shown that
the flight behavior of many animals and insects demonstrates the
typical characteristic of Lévy flights (Brown, Liebovitch, & Glendon,
2007). A Lévy flight is a random walk in which the step-lengths are
distributed according to a heavy-tailed probability distribution.
After a large number of steps, the distance from the origin of the
random walk tends to a stable distribution.
2.3. Cuckoo search implementation
Each egg in a nest represents a solution, and a cuckoo egg
represents a new solution. The aim is to employ the new and
potentially better solutions (cuckoos) to replace not-so-good solu-
tions in the nests. In the simplest form, each nest has one egg. The
algorithm can be extended to more complicated cases in which
each nest has multiple eggs representing a set of solutions (Yang
& Deb, 2009, 2010). The CS is based on three idealized rules:
 Each cuckoo lays one egg at a time, and dumps it in a randomly
chosen nest.
 The best nests with high quality of eggs (solutions) will carry
over to the next generations.
 The number of available host nests is fixed, and a host can dis-
cover an alien egg with probability pa 2 [0, 1]. In this case, the
host bird can either throw the egg away or abandon the nest
to build a completely new nest in a new location (Yang &
Deb, 2010).
For simplicity, the last assumption can be approximated by a
fraction pa of the n nests being replaced by new nests, having
new random solutions. For a maximization problem, the quality
or fitness of a solution can simply be proportional to the objec-
tive function. Other forms of fitness can be defined in a similar
way to the fitness function in genetic algorithms (Yang & Deb,
2010).
Based on the above-mentioned rules, the basic steps of the CS
can be summarized as the pseudo code, as follows (Yang & Deb,
2010):
begin
Objective function f(x), x = (x1, . . . , xd)T
Generate initial population of
n host nests xi(i = 1, 2, . . . , n)
While (t < MaxGeneration) or (stop criterion)
Get a cuckoo randomly by Lévy flights
evaluate its quality/fitness Fi
Choose a nest among n (say.j) randomly
if (Fi > Fj),
replace j by the new solution;
end if
A fraction (pa) of worse nests
are abandoned and new ones are built;
Keep the best solutions
(or nests with quality solutions);
Rank the solutions and find the current best
end while
Postprocess results and visualization
end
When generating new solutions xi(t + 1) for the ith cuckoo, the
following Lévy flight is performed
xi ðt þ 1Þ ¼ xi ðtÞ þ a  Le
vy ðkÞ ð1Þ
where a > 0 is the step size, which should be related to the scale of
the problem of interest. The product  means entry-wise multipli-
cations location (Yang and Deb, 2009). In this research work, we
consider a Lévy flight in which the step-lengths are distributed
according to the following probability distribution
vy u ¼ t k ; 1 < k 6 3
Le ð2Þ
which has an infinite variance. Here, the consecutive jumps/steps of
a cuckoo essentially form a random walk process which obeys a
power-law step-length distribution with a heavy tail.
It is worth pointing out that, in the real world, if a cuckoo’s egg
is very similar to a host’s eggs, then this cuckoo’s egg is less likely
to be discovered, thus the fitness should be related to the
 E. Valian et al. / Computers & Industrial Engineering 64 (2013) 459–468 461

difference in solutions. Therefore, it is a good idea to do a random where k represents the penalty coefficient, and it is set to 105, in this
walk in a biased way with some random step sizes location (Yang & paper. As the maximization problem of F(x) is equivalent with the
Deb, 2010). minimization problem of F(x), Eq. (6) can be written as follows.

Xng

3. Improved cuckoo search Min FðxÞ ¼ f ðxÞ þ k maxð0; g i Þ ð7Þ

i¼1

The parameters pa, k and a introduced in the CS help the algo-

rithm to find globally and locally improved solutions, respectively. 4.2. Optimization process for discrete variables
The parameters pa and a are very important parameters in fine-
tuning of solution vectors, and can be potentially used in adjusting Many engineering optimization problems involve discrete vari-
convergence rate of algorithm. The traditional CS algorithm uses ables. In this paper, these discrete variables, representing the num-
fixed value for both pa and a. These values are set in the initializa- ber of components in the ith subsystem, are denoted with ni. The
tion step and cannot be changed during new generations. The main parameter ni is a real number and is directly transformed into
drawback of this method appears in the number of iterations to the nearest integer.
find an optimal solution. If the value of pa is small and the value
of a is large, the performance of the algorithm will be poor and
leads to considerable increase in number of iterations. If the value 5. Case studies: reliability optimization problems
of pa is large and the value of a is small, the speed of convergence is
high but it may be unable to find best solutions. To evaluate the performance of the proposed approach on reli-
The key difference between the ICS and CS is in the way of ability optimization problems, four case studies are considered.
adjusting pa and a. To improve the performance of the CS algorithm They are a complex (bridge) system, a series system, a series–par-
and eliminate the drawbacks resulted from considering fixed val- allel system, and an overspeed protection system.
ues of pa and a, the ICS algorithm uses variables pa and a. In the
early generations, the values of pa and a must be big enough to en- 5.1. Case study 1: a complex (bridge) system
force the algorithm to increase the diversity of solution vectors.
However, these values should be decreased in final generations This case study (Chen, 2006; Coelho, 2009; Gen & Yun, 2006;
to result in a better fine-tuning of solution vectors. The values of Hikita, Nakagawa, & Harihisa, 1992; Hsieh, Chen, & Bricker, 1998;
pa and a are dynamically changed with the number of generation Wu et al., 2010; Yeh & Hsieh, 2011; Zou et al., 2010) is a nonlinear
and expressed in Eqs. (3)–(5), where NI and gn are the number of mixed-integer programming problem for a complex (bridge)
total iterations and the current iteration, respectively. system with five subsystems, as shown in Fig. 1.
The problem is formulated as follows
ðPamax  Pamin Þ
Pa ðgnÞ ¼ Pamax   gn ð3Þ
NI Max f ðr; nÞ ¼ R1 R2 þ R3 R4 þ R1 R4 R5 þ R2 R3 R5
aðgnÞ ¼ amax expðc:gnÞ ð4Þ R1 R2 R3 R4  R1 R2 R3 R5  R1 R2 R4 R5
 
Ln aamax
min R1 R3 R4 R5  R2 R3 R4 R5 þ 2R1 R2 R3 R4 R5
c¼ ð5Þ s:t:
NI
X
m
g 1 ðr; nÞ ¼ wi v 2i n2i  V 6 0
i¼1
4. Some preparation works for applying the ICS to reliability
Xm  bi
optimization problems g 2 ðr; nÞ ¼ 1000
ai  lnðr ½ni þ expð0:25ni Þ  C 6 0
iÞ
i¼1
4.1. Conversion of constrained optimization problems to X
m
unconstrained ones g 3 ðr; nÞ ¼ wi ni expð0:25ni Þ  W 6 0
i¼1

The general mathematical model of reliability optimization 0 6 r i 6 1; ni 2 Z þ ; 16i6m

problems can be formulated as
where m is the number of subsystems in the system. ni, ri, and
Max f ðxÞ qi = 1  ri are the number of components, the reliability of each
ð5Þ component, and the failure probability of each component in
s:t: : g j ðxÞ 6 0 j ¼ 1; 2; . . . ; ng  n 
subsystem i, respectively. Ri ðni Þ ¼ 1  qi i is the reliability of
where f(x) is the reliability function, gj(x) is the jth resource con-
straint, and ng is the number of constraints. There is a big difference
between unconstrained optimization problems and constrained
ones. The global best solution of an unconstrained optimization
problem is the solution which has the maximum value of the objec- 1 2
tive function. However, it is difficult to find a balance between the
constraints and the value of the objective function for a constrained
optimization problem. As a result, it is a challenging task to deter-
mine the global best solution vector for such a problem. 5
To avoid violation of constraints, unfeasible solutions should be
adapted to feasible solutions. To do this, a penalty function is em-
ployed to convert the constrained optimization problem in Eq. (5)
to the unconstrained one in Eq. (6)
3 4
ng
X
Max FðxÞ ¼ f ðxÞ  k maxð0; g i Þ ð6Þ
i¼1 Fig. 1. The schematic diagram of complex (bridge) system.
462 E. Valian et al. / Computers & Industrial Engineering 64 (2013) 459–468

subsystem i, whilst f(r, n) is the system reliability. For each compo- Table 1
nent in subsystem i, wi, vi, and ci are the weight, volume, and cost, Data used in complex (bridge) and series systems.

respectively. V, C, and W refer to the upper limit on the sum of i 105ai bi wiv 2i wi V C W
the subsystems’ products of volume and weight, the upper limit
1 2.330 1.5 1 7
on the cost of the system, and the upper limit on the weight of 2 1.450 1.5 2 8
the system, respectively. The parameters ai and bi are physical fea- 3 0.541 1.5 3 8 110 175 200
tures of system components. Constraint g1(r, n) is a combination of 4 8.050 1.5 4 6
weight, redundancy allocation, and volume. g2(r, n) and g3(r, n) are 5 1.950 1.5 2 9

constraints on the cost and weight. The input parameters of the

complex (bridge) system are given in Table 1.

1 2 3 4 5
5.2. Case study 2: a series system
Fig. 2. The schematic diagram of a series system.
This case study (Chen, 2006; Hikita et al., 1992; Hsieh et al.,
1998; Kuo, Hwang, & Tillman, 1978; Wu et al., 2010; Xu, Kuo, &
Lin, 1990; Yeh & Hsieh, 2011) is a nonlinear mixed-integer pro- 5.4. Case study 4: an overspeed system for a gas turbine
gramming problem for a series system with five subsystems, as
shown in Fig. 2. Overspeed detection is continuously provided by the electrical
The problem formulation is as follows: and mechanical systems. When an overspeed occurs, it is necessary
to cut off the fuel supply. For this purpose, 4 control valves (V1–V4)
Y
m must be closed. The control system is modeled as a 4-stage series
Max f ðr; nÞ ¼ Ri ðni Þ system, as shown in Fig. 4. The objective is to determine an optimal
i¼1
level of ri and ni at each stage i so that the system reliability is
s:t:
maximized.
X
m
g 1 ðr; nÞ ¼ wi v 2i n2i  V 6 0 This case study (Chen, 2006; Coelho, 2009; Dhingra, 1992; Wu
i¼1 et al., 2010; Yeh & Hsieh, 2011; Yokota et al., 1996; Zou et al.,
X
m  bi 2010, 2011) is formulated as follows
1000
g 2 ðr; nÞ ¼ ai  lnðr Þ
½ni þ expð0:25ni Þ  C 6 0
i
i¼1 Y
m
X
m Max f ðr; nÞ ¼ ½1  ð1  r i Þni 
g 3 ðr; nÞ ¼ wi ni expð0:25ni Þ  W 6 0 i¼1
i¼1 s:t:
þ
0 6 r i 6 1; ni 2 Z ; 16i6m X
m
g 1 ðr; nÞ ¼ v i n2i  V 6 0
This case study has three nonlinear constraints, which are the same i¼1
as those of the 1st case study. Also, the input parameters of the ser- Xm

ies system are the same as those of the complex (bridge) system. g 2 ðr; nÞ ¼ Cðr i Þ½ni þ expð0:25ni Þ  C 6 0
i¼1
Xm
g 3 ðr; nÞ ¼ wi ni expð0:25ni Þ  W 6 0
5.3. Case study 3: a series–parallel system i¼1

0:5 6 r i 6 1  106 ; r i 2 Rþ ; 1 6 ni 6 10; ni 2 Z þ

This case study (Chen, 2006; Hikita et al., 1992; Hsieh et al.,
1998; Kuo et al., 1978; Wu et al., 2010; Xu et al., 1990; Yeh and where ri, ni, vi, and wi refer to the reliability of each component, the
Hsieh, 2011) is a nonlinear mixed-integer programming problem number of redundant components, the product of weight and vol-
for a series–parallel system with five subsystems, as shown in ume per element, and the weight of each component, all in stage
Fig. 3. The reliability function reported by (Gen and Kim, 1999; i. The term exp(0.25ni) accounts for the interconnecting hardware.
Xu et al., 1990; Yeh and Hsieh, 2011) was f(r,n) = 1  (1  R1R2) The cost of each component with reliability ri at subsystem i is given
 bi
(1  (1  R3)(1  R4)R5). In fact, it is wrong (Wu et al., 2010), and T
by Cðr i Þ ¼ ai  lnðr . Parameters ai and bi are constants represent-
i Þ
the right problem formulation is as follows:
ing the physical characteristic of each component at stage i. T is the
operating time during which the component must not fail. The in-
Max f ðr; nÞ ¼ 1  ð1  R1 R2 Þð1  ðR3 þ R4  R3 R4 ÞR5 Þ
s:t:
X
m
g 1 ðr; nÞ ¼ wi v 2i n2i  V 6 0 1 2
i¼1
Xm  bi
1000
g 2 ðr; nÞ ¼ ai  lnðr iÞ
½ni þ expð0:25ni Þ  C 6 0
i¼1
X
m
g 3 ðr; nÞ ¼ wi ni expð0:25ni Þ  W 6 0 3
i¼1

0 6 r i 6 1; ni 2 Z þ ; 16i6m 5

4
This case study has the same nonlinear constraints as those of the
1st case study, but it has different input parameters, given in Table
2. Fig. 3. The schematic diagram of a series–parallel system.
 E. Valian et al. / Computers & Industrial Engineering 64 (2013) 459–468 463

Table 2 error h, which implies 33% of the minimum requirement of each re-
Data used in series–parallel system. source (lj), is available for optimization. For the reliability system
i 105ai bi wi v 2i wi V C W with n subsystems, the average minimum resource requirements
Pn
1 2.500 1.5 2 3.5 j¼1 g ij ðlj Þ; ði ¼ 1; . . . ; 4Þ and the average values of which is repre-
 h
Pn
2 1.450 1.5 4 4 sented by bi ¼ 1 þ 100 j¼1 g ij ðlj Þ; ði ¼ 1; . . . ; 4Þ. In this way, we set
3 0.541 1.5 5 4 180 175 100
the available system resources for reliability system with 36, 38,
4 0.541 1.5 8 3
5 2.100 1.5 4 4.5
40, 42 and 50 subsystems, respectively, as shown in Table 5.
The best solutions obtained by the above two approaches (Gen
& Yun, 2006; Prasad & Kuo, 2000) are reported in Table 6. Here,
VTV represents the variables which get value 2 in optimum condi-
tions and the other variables are equal to 1.
Gas Turbine

Mechanical 5.6. Case study 6: 15-unit system reliability optimization problem

and
electrical Considering a 15-unit structure, shown in Fig. 5, the optimiza-
V1 V2 V3 V4
overspeed tion problem is defined as follows (Agrwal & Vikas, 2010).
detection
Max f ðxÞ ¼ R1 R2 R3 R4 þ R9 R10 R11 R12 R13 R14 R15
ðQ 1 þ R1 Q 2 þ R1 R2 Q 3 þ R1 R2 R3 Q 4 þ R1 R2 R3 R4 Q 5 þ R1 R2 R3 R4 R5 Q 6 Þ
Air Fuel Mixture þR3 R4 R5 R6 R7 R9 R10 ðQ 11 þ R11 Q 12 þ R11 R12 Q 13 þ R11 R12 R13 Q 14 þ R11 R12 R13 R14 Q 15 Þ
ðQ 1 þ R1 Q 2 Þ þ ððQ 1 þ R1 Q 2 ÞðQ 3 þ R3 Q 4 þ R3 R4 Q 7 Þ þ R1 R2 Q 7 ðQ 3 þ R3 Q 4 ÞÞ
ðQ 13 þ R13 Q 14 þ R13 R14 Q 15 ÞR5 R6 R8 R9 R10 R11 R12 þ R1 R2 R5 R6 R7 R8 R11 R12
Fig. 4. The schematic diagram of an overspeed system for a gas turbine. ðR9 R10 þ Q 9 þ R9 Q 10 ÞðQ 3 þ R3 Q 4 ÞðQ 13 þ R13 Q 14 þ R13 R14 Q 15 Þ þ ðQ 5 þ R5 Q 6 Þ
ððQ 7 þ R7 Q 11 þ R7 R11 Q 12 ÞðQ 9 þ R9 Q 10 Þ þ R9 R10 ðQ 11 þ R11 Q 12 ÞÞ
R1 R2 R3 R4 R8 R13 R14 R15 þ R1 R2 R7 R11 R12 R13 R14 R15 ðQ 9 þ R9 Q 10 Þ
put parameters defining the overspeed protection system are given
ðQ 3 þ R3 Q 4 þ R3 R4 Q 5 þ R3 R4 R5 Q 6 Þ þ R3 R4 R7 R8 R9 R10 R13 R14 R15 ðQ 1 þ R1 Q 2 Þ
in Table 3.
ðQ 11 þ R11 Q 12 ÞðQ 5 þ R5 Q 6 Þ
s:t:
5.5. Case study 5: large-scale system reliability problem X
15
g y ðr;nÞ ¼ cyi xi 6 by y ¼ 1;2;...;m
i¼1
Recently, Prasad and Kuo (2000) developed P and K-Algorithm xi 2 Z þ ; i ¼ 1;2;...;15:
for solving large-scale system reliability optimization design prob-
lems. Gen and Yun (2006) developed a soft computing approach to P
where by ¼ d  15 i¼1 cyi with d = rand(1.5, 3.5). The parameters
solve the same problem. The considered mathematical formula- Ri ðxi Þ ¼ 1  ð1  r i Þxi and Qi = 1  Ri refer to the reliability and unre-
tions are as follows: liability of subsystem i, respectively. The coefficients cyi and ri are
Y
n generated from uniform distributions in [0, 100] and [0.6, 0.85],
Max f ðr; nÞ ¼ ½1  ð1  r j Þxj  respectively. The parameter m refers to the number of constraints.
j¼1 Two sets of problems are considered by taking m = 1 and m = 5.
s:t: The objective function f(x) was obtained using the recursive disjoint
X
m
 X
n
g 1 ðr; nÞ ¼ ai x2i  1 þ 100
h
aj l2j 6 0
i¼1 j¼1 Table 4
X
n
x   X
n  Data used in Large-scale system reliability problem.
h l
g 2 ðr; nÞ ¼ bj exp 2j  1 þ 100 bj exp 2j 6 0
j¼1 j¼1
j 1  rj aj bj cj dj j 1  rj aj bj cj dj

X
n
 X
n 1 0.005 8 4 13 26 26 0.029 8 1 18 35
g 3 ðr; nÞ ¼ cj xj  1 þ 100
h
cj lj 6 0 2 0.026 10 4 16 32 27 0.022 8 3 16 32
j¼1 j¼1 3 0.035 10 4 12 23 28 0.017 9 3 15 29
4 0.029 6 3 12 24 29 0.002 10 1 18 35
X
n pffiffiffi
 X
n pffiffiffi
h 5 0.032 7 1 13 26 30 0.031 9 2 19 37
g 4 ðr; nÞ ¼ dj xj  1 þ 100 dj lj 6 0
6 0.003 10 4 16 31 31 0.021 7 5 15 28
j¼1 j¼1
7 0.020 9 2 19 38 32 0.023 9 5 11 22
1 6 xj 6 10; j ¼ 1; 2; . . . ; n: 8 0.018 9 3 15 29 33 0.030 6 3 15 29
9 0.004 7 4 12 23 34 0.026 7 3 14 27
Table 4 shows the design data of reliability systems for case study 5. 10 0.038 6 4 16 31 35 0.009 6 5 15 29
The component reliabilities are generated from the uniform distri- 11 0.028 6 5 14 28 36 0.019 10 5 17 33
bution in [0.95, 1.0]. The coefficients aj, bj, cj and dj are generated 12 0.021 10 3 15 30 37 0.005 9 5 19 37
13 0.039 9 1 17 34 38 0.019 10 5 11 22
from uniform distributions in [6, 10], [1, 5], [11, 20] and [21, 40],
14 0.013 10 4 20 39 39 0.002 6 2 17 34
respectively. Here, lj represents the lower bound of xj. The tolerance 15 0.038 7 4 14 28 40 0.015 8 3 17 33
16 0.037 10 2 13 25 41 0.023 10 5 17 33
17 0.021 10 1 15 29 42 0.040 8 3 18 35
Table 3 18 0.023 8 3 19 38 43 0.012 8 1 18 35
Data used in overspeed protection system. 19 0.027 10 5 18 36 44 0.026 6 4 19 38
20 0.028 7 4 13 26 45 0.038 6 4 13 26
i 105ai bi vi wi V C W T
21 0.030 6 2 15 30 46 0.015 8 1 19 37
1 1.0 1.5 1 6 22 0.027 6 2 12 24 47 0.036 7 4 14 28
2 2.3 1.5 2 6 23 0.018 7 2 20 40 48 0.032 10 2 19 37
3 0.3 1.5 3 8 250 400 500 1000 h 24 0.013 8 5 19 38 49 0.038 8 3 15 30
4 2.3 1.5 2 7 25 0.006 9 5 15 39 50 0.013 10 2 11 22
464 E. Valian et al. / Computers & Industrial Engineering 64 (2013) 459–468

Table 5 For the first four case studies, Tables 8–11 compare the best re-
Available system resources for each system in case study 5. sults obtained by the ICS with those provided by other methods re-
n i 1 2 3 4 ported in the literatures (Chen, 2006; Coelho, 2009; Dhingra, 1992;
36 bi 391 257 738 1454 Hsieh et al., 1998; Kuo et al., 1978; Wu et al., 2010; Xu et al., 1990;
38 bi 416 278 778 1532 Yokota et al., 1996; Zou et al., 2010; Zou et al., 2011). ‘Slack’ repre-
40 bi 435 289 823 1621 sents unused resources.
42 bi 458 306 870 1712 An improvement index is required to measure the improvement
50 bi 543 352 1040 2048
of the best solutions found by the proposed approach in compari-
son with those given by the other methods. This index, which has
been called maximum possible improvement (MPI) (Coelho, 2009),
Table 6 is as follows
The optimization results of case study 5 with different dimensions.

n VTV in optimum f(x)

 
fICS  fother
MPI ð%Þ ¼ ð8Þ
36 {5, 10, 15, 21, 33} 0.519976 1  fother
38 {10, 13, 15, 21, 33} 0.510989
40 {5, 10, 13, 15, 33} 0.503292
42 {4, 10, 11, 15, 21, 33} 0.479664 where fICS represents the best system reliability obtained by the ICS
50 {4, 10, 15, 21, 33, 45, 47} 0.405390 algorithm, while fother denotes the best system reliability obtained
by any other method. Considering this index, the solutions obtained
by the ICS are clearly better than those given by other methods. It
should be noted that in high reliability applications, it is often dif-
product method given by Abraham (1979). The random data used in ficult to obtain even very small improvements in reliability.
two sets of problems is given in Table 14. For the complex (bridge) system, as mentioned in Table 8, the
best results reported by Hikita et al. (1992), Hsieh et al. (1998),
6. Simulation results, analysis and discussion Chen (2006), Coelho (2009), Zou et al. (2010), Wu et al. (2010),
and Yeh and Hsieh (2011), are 0.9997894, 0.99987916, 0.999
The parameters of the CS and ICS algorithms used for reliability 88921, 0.99988957, 0.99988962, 0.99988963 and 0.99988962
optimization problems are shown in Table 7. Here, ‘SD’ represents respectively. The result provided by the ICS is better than the above
the standard deviation based on fifty converged objective function seven results. The improvement indices are 47.597341%,
values. ‘NFOS’ represents the number of feasible optimal solutions 8.672625%, 0.388122%, 0.063389%, 0.018120%, 0.009060% and
found in fifty runs. 0.018120%, respectively.

1 2 3 4 5 6

7 8 15

9 10 11 12 13 14

Fig. 5. The schematic diagram of 15-unit system.

Table 7
Parameter for CS and ICS.

Example Population size Number of iterations Pa for CS a for CS Pa for ICS a for ICS
Case study 1 10 1000 0.25 1 Pa(max) = 0.5 a(max) = 0.5
Pa(min) = 0.005 a(min) = 0.01
Case study 2 10 1000 0.25 1 Pa(max) = 0.5 a(max) = 0.5
Pa(min) = 0.005 a(min) = 0.01
Case study 3 10 1000 0.25 1 Pa(max) = 0.5 a(max) = 0.5
Pa(min) = 0.005 a(min) = 0.01
Case study 4 10 1000 0.25 1 Pa(max) = 0.5 a(max) = 0.5
Pa(min) = 0.005 a(min) = 0.01
Case study 5-(36 DIM) 10 2500 0.25 1 Pa(max) = 0.5 a(max) = 0.5
Pa(min) = 0.005 a(min) = 0.01
Case study 5-(38 DIM) 10 2500 0.25 1 Pa(max) = 0.5 a(max) = 0.5
Pa(min) = 0.005 a(min) = 0.01
Case study 5-(40 DIM) 10 2500 0.25 1 Pa(max) = 0.5 a(max) = 0.5
Pa(min) = 0.005 a(min) = 0.01
Case study 5-(42 DIM) 10 2500 0.25 1 Pa(max) = 0.5 a(max) = 0.5
Pa(min) = 0.005 a(min) = 0.01
Case study 5-(50 DIM) 10 5000 0.25 1 Pa(max) = 0.5 a(max) = 0.5
Pa(min) = 0.005 a(min) = 0.01
Case study 6-(15  1) 10 500 0.25 1 Pa(max) = 0.5 a(max) = 0.5
Pa(min) = 0.005 a(min) = 0.01
Case study 6-(15  1) 10 500 0.25 1 Pa(max) = 0.5 a(max) = 0.5
Pa(min) = 0.005 a(min) = 0.01
 E. Valian et al. / Computers & Industrial Engineering 64 (2013) 459–468 465

Table 8
Case study 1: Bridge (complex) system.

Parameter Hikita et al. (1992) Hsieh et al. (1998) Chen (2006) Coelho (2009) Zou et al. (2010) Wu et al. (2010) Yeh and Hsieh (2011) ICS
f(r, n) 0.9997894 0.99987916 0.99988921 0.99988957 0.99988962 0.99988963 0.99988962 0.99988964
n1 3 3 3 3 3 3 3 3
n2 3 3 3 3 3 3 3 3
n3 2 3 3 2 2 2 2 2
n4 3 3 3 4 4 4 4 4
n5 2 1 1 1 1 1 1 1
r1 0.814483 0.814090 0.812485 0.826678 0.82883148 0.82868361 0.828087 0.828094038
r2 0.821383 0.864614 0.867661 0.857172 0.85836789 0.85802567 0.857805 0.858004485
r3 0.896151 0.890291 0.861221 0.914629 0.91334996 0.91364616 0.914240 0.914162924
r4 0.713091 0.701190 0.713852 0.648918 0.64779451 0.64803407 0.648146 0.647907792
r5 0.814091 0.734731 0.756699 0.715290 0.70178737 0.70227595 0.704163 0.704565982
MPI (%) 47.597341 8.672625 0.388122 0.063389 0.018120 0.009060 0.018120 –
Slack (g1) 18 18 19 5 5 5 5 5
Slack (g2) 1.854075 0.376347 0.001494 0.000339 0.00004063 0.00000359 0.00002500 0.00007929
Slack (g3) 4.264770 4.264770 4.264770 1.560466 1.56046629 1.56046629 1.56046628 1.560466288

Table 9
Case study 2: Series system.

Parameter Kuo et al. (1978) Xu et al. (1990) Hikita et al. (1992) Hsieh et al. (1998) Chen (2006) Wu et al. (2010) Yeh and Hsieh (2011) ICS
f(r, n) 0.92975 0.931677 0.931363 0.931578 0.931678 0.931680 0.931682 0.931682387
n1 3 3 3 3 3 3 3 3
n2 3 2 2 2 2 2 2 2
n3 2 2 2 2 2 2 2 2
n4 3 3 3 3 3 3 3 3
n5 2 3 3 3 3 3 3 3
r1 0.77960 0.77939 0.777143 0.779427 0.779266 0.78037307 0.779399 0.779416938
r2 0.80065 0.87183 0.867514 0.869482 0.872513 0.87178343 0.871837 0.871833278
r3 0.90227 0.90288 0.896696 0.902674 0.902634 0.90240890 0.902885 0.902885082
r4 0.71044 0.71139 0.717739 0.714038 0.710648 0.71147356 0.711403 0.711393868
r5 0.85947 0.78779 0.793889 0.786896 0.788406 0.78738760 0.787800 0.787803712
MPI (%) 2.750748 0.007904 0.465347 0.152583 0.006440 0.003513 0.000006 –
Slack (g1) 27 27 27 27 27 27 27 27
Slack (g2) 0.000010 0.013773 0.000000 0.121454 0.001559 0.121454 0.0002184 0.000000265
Slack (g3) 10.57248 7.518918 7.518918 7.518918 7.518918 7.518918 7.5189182 7.518918241

Table 10
Case study 3: Series–parallel system.

Parameter Hikita et al. (1992) Hsieh et al. (1998) Chen (2006) Wu et al. (2010) ICS
f(r, n) 0.99996875 0.99997418 0.99997658 0.99997664 0.999976649
n1 3 2 2 2 2
n2 3 2 2 2 2
n3 1 2 2 2 2
n4 2 2 2 2 2
n5 3 4 4 4 4
r1 0.838193 0.785452 0.812485 0.81918526 0.819927087
r2 0.855065 0.842998 0.843155 0.84366421 0.845267657
r3 0.878859 0.885333 0.897385 0.89472992 0.895491554
r4 0.911402 0.917958 0.894516 0.89537628 0.895440692
r5 0.850355 0.870318 0.870590 0.86912724 0.868318775
MPI (%) 25.280000 9.566228 0.298890 0.042808 –
Slack (g1) 53 40 40 40 40
Slack (g2) 0.000000 1.194440 0.002627 0.000561 0.0000161
Slack (g3) 7.110849 1.609289 1.609289 1.609289 1.6092890

For the series system, Table 9 shows that the best results re-
ported by Kuo et al. (1978), Xu et al. (1990), Hikita et al. (1992),
Hsieh et al. (1998), Chen (2006), Wu et al. (2010), and Yeh and Hsieh
(2011), are 0.92975, 0.931677, 0.931363, 0.931578, 0.931678,
0.931680 and 0.931682 respectively. The result given by the ICS
is better than the above-mentioned results. The improvement indi-
ces are 2.750748%, 0.007904%, 0.465347%, 0.152583%, 0.006440%,
0.003513% and 0.000006%, respectively.
A same investigation has been done for the series–parallel sys-
tem. Table 10 shows that the ICS leads to better results and the
improvement indices are 25.280000%, 9.566228%, 0.298890%, and
0.042808% in comparison with the best results reported by Hikita
et al. (1992), Hsieh et al. (1998), Chen (2006), and Wu et al. (2010),
respectively.
Considering an overspeed system for a gas turbine, Table 11
shows that the best results reported by Dhingra (1992), Yokota
466 E. Valian et al. / Computers & Industrial Engineering 64 (2013) 459–468

Table 11
Case study 4: Over-speed system.

Parameter Dhingra (1992) Yokota et al. (1996) Chen (2006) Coelho (2009) Zou et al. (2010) Wu et al. (2010) Yeh and Hsieh (2011) ICS
f(r, n) 0.99961 0.999468 0.999942 0.999953 0.99995467 0.99995467 0.99995468 0.99995468
n1 6 3 5 5 5 5 5 5
n2 6 6 5 6 6 6 6 5
n3 3 3 5 4 4 4 4 4
n4 5 5 5 5 5 5 5 6
r1 0.81604 0.965593 0.903800 0.902231 0.90186194 0.90163164 0.901614 0.901614595
r2 0.80309 0.760592 0.874992 0.856325 0.84968407 0.84997020 0.849920 0.888223369
r3 0.98364 0.972646 0.919898 0.948145 0.94842696 0.94821828 0.948143 0.948141029
r4 0.80373 0.804660 0.890609 0.883156 0.88800590 0.88812885 0.888223 0.849920899
MPI (%) 88.4615 91.5414 22.4138 4.2553 0.012186 0.000910 – –
Slack (g1) 65 92 50 55 55 55 55 55
Slack (g2) 0.064 70.733576 0.002152 0.975465 0.00120356 0.000009 0.0000336 0.0000000096
Slack (g3) 4.348 127.583189 28.803701 24.801882 24.8018827 24.081883 24.801883 15.36346309

Table 12
The best results of case study 5 with different dimensions given by ICS algorithm.

n VTV in optimum f(x) Slack (g1) Slack (g2) Slack (g3) Slack (g4)
36 {5, 10, 15, 21, 33} 0.51997597 1 49.12576352 109 301.353247
38 {10, 13, 15, 21, 33} 0.51098860 1 53.63855081 115 317.0395385
40 {4, 10, 11, 21, 22, 33} 0.50599242 0 51.04714167 119 333.2405486
42 {4, 10, 11, 15, 21, 33} 0.47966355 2 52.71825039 129 52.71825039
50 {4, 10, 15, 21, 33, 42, 45} 0.40695474 0 61.95598259 154 423.9146468

Table 13
Comparison results for the large-scale system reliability problem.

Example Algorithm Best Worst Median SD

P5-(36 DIM) IPSO (Wu et al., 2010) 0.51997597 0.49705368 0.50874873 5.8624e003
ICS 0.51997597 0.51072253 0.51908590 2.39974e03
P5-(38 DIM) IPSO (Wu et al., 2010) 0.51098860 0.48707780 0.50557752 5.3420e003
ICS 0.51098860 0.50754594 0.51073726 7.77523e04
P5-(40 DIM) IPSO (Wu et al., 2010) 0.50599242 0.48016779 0.49772492 5.4393e003
ICS 0.50599242 0.50231712 0.50328548 3.23365e04
P5-(42 DIM) IPSO (Wu et al., 2010) 0.47966355 0.45788422 0.47090324 4.6902e003
ICS 0.47966355 0.47295763 0.47670740 1.19317e03
P5-(50 DIM) IPSO (Wu et al., 2010) 0.40657143 0.36682410 0.39338827 6.5878e003
ICS 0.40695475 0.40126529 0.40507236 2.16488e03

Table 14 Table 15
Data used in the 15-unit system reliability problem. Comparison results for the 15-unit system reliability problem.

i r c1 c2 c3 c4 c5 Example Algorithm Best Worst Median SD

1 0.6796 33.2468 35.6054 13.7848 44.1345 10.9891 P6-(15  1) GA 0.82700343 0.31578718 0.58041963 1.1514e01
2 0.7329 27.5668 44.9520 96.7365 25.9855 68.0713 ABC 0.95260704 0.93872370 0.94819588 3.4596e03
3 0.6688 13.3800 28.6889 85.8783 19.2621 1.0164 HS 0.97235030 0.96904606 0.97141480 7.0892e04
4 0.6102 0.4710 0.4922 63.0815 12.1687 29.4809 NGHS 0.97235030 0.96718809 0.97089223 9.9082e04
5 0.7911 51.2555 39.6833 78.5364 23.9668 59.5441 ICS 0.97235030 0.97193413 0.97233609 6.4903e05
6 0.8140 82.9415 59.2294 11.8123 28.9889 46.5904
P6-(15  5) GA 0.82381093 0.36520795 0.61764965 1.3423e01
7 0.8088 51.8804 78.4996 97.1872 47.8387 49.6226
ABC 0.95141221 0.93911154 0.94695160 2.9954e03
8 0.7142 77.9446 86.6633 45.0850 25.0545 59.2594
HS 0.95825279 0.95364976 0.95685535 1.0764e03
9 0.8487 26.8835 7.8195 3.6722 76.9923 87.4070
NGHS 0.95810719 0.95055900 0.95594432 1.8335e03
10 0.7901 85.8722 27.7460 55.3950 53.3007 55.3175
ICS 0.95825279 0.95800848 0.95812801 7.1692e05
11 0.6972 41.8733 90.4377 75.7999 95.0057 54.1269
12 0.6262 61.6181 58.0131 98.5166 97.9127 59.1341
13 0.6314 90.0418 77.8206 60.6308 37.2226 40.9427
14 0.6941 75.5947 36.4524 70.4654 96.9179 40.2141
15 0.6010 88.5974 61.0591 18.8802 42.1222 80.0045
seven results. The improvement indices are 88.461538%,
d – 3.2150 3.4710 3.3247 2.6236 3.4288
91.541353%, 22.413793%, 4.255319%, 0.012186%, 0.000910% and
0%, respectively.
For case study 5, the large-scale optimization problem, Table 6
et al. (1996), Chen (2006), Coelho (2009), Zou et al. (2010), Wu et shows the best results given by Prasad and Kuo (2000) and Gen and
al. (2010), and Yeh and Hsieh (2011), are 0.99961, 0.999468, Yun (2006). For this case study, Table 12 shows the best results
0.999942, 0.999953, 0.99995467, 0.99995467 and 0.99995468 provided by ICS algorithm. It can be seen that the ICS gives the best
respectively. The result provided by the ICS is better than the above results.
 E. Valian et al. / Computers & Industrial Engineering 64 (2013) 459–468 467

Table 16
Simulation results after 50 runs.

Example Algorithm Best Worst Median SD NFOS

Case study1 ICS 0.99988964 0.99983249 0.99987998 1.40652e05 50
CS 0.99988961 0.99967108 0.99986831 3.50850e05 50
Case study 2 ICS 0.93168239 0.92066034 0.92987132 1.99046e03 50
CS 0.93168068 0.91009077 0.92819071 5.09968e03 50
Case study 3 ICS 0.99997658 0.99994886 0.99997090 4.45034e06 50
CS 0.99997663 0.99972236 0.99995902 3.03977e05 50
Case study 4 ICS 0.99995468 0.99991746 0.99995335 4.45034e06 50
CS 0.99995467 0.99988279 0.99995164 8.59533e06 50
Case study 5-(36 DIM) ICS 0.51997597 0.51072253 0.51908590 2.39974e03 50
CS 0.51997597 0.47377699 0.50197913 8.80802e03 50
Case study 5-(38 DIM) ICS 0.51098860 0.50754594 0.51073726 7.77523e04 50
CS 0.50999639 0.46470218 0.49385964 9.17139e03 50
Case study 5-(40 DIM) ICS 0.50599242 0.50231712 0.50328548 3.23365e04 50
CS 0.50329249 0.45579344 0.48574531 9.50105e03 50
Case study 5-(42 DIM) ICS 0.47966355 0.47295763 0.47670740 1.19317e03 50
CS 0.47433072 0.43225203 0.45862087 8.40493e03 50
Case study 5-(50 DIM) ICS 0.40695475 0.40126529 0.40507236 2.16488e03 50
CS 0.40031471 0.35897796 0.38697083 8.10335e03 50
Case study 6-(15  1) ICS 0.97235030 0.97235030 0.97235030 6.49029e05 50
CS 0.97235030 0.95244364 0.96986428 4.85763e03 50
Case study 6-(15  5) ICS 0.95825279 0.95800848 0.95812801 7.16924e05 50
CS 0.95825279 0.95056430 0.9567778 2.475548e03 50

For all case studies, Table 16 includes the simulation results of
fifty independent runs given by the CS and ICS. It shows that the
ICS performs better than the CS in all criteria.
It can be seen in Table 13 that the best results of ICS and Wu
algorithms (Wu et al., 2010) are the same for n = 36, 38, 40, 42.
Here, ‘SD’ represents the standard deviation based on fifty con-
verged objective function values. Also the worst, median and SD re-
sults given by the ICS are better than those given by Wu for n = 36,
38, 40, and 42. Moreover, for n = 50 the ICS provides better results
in terms of all criteria in comparison with Wu algorithm.
For case study 6, the 15-unit system reliability optimization
problem, Table 15 shows the results given by Genetic Algorithm
(GA) (Holland, 1975), Artificial Bee Colony (ABC) (Karaboga,
2005), Harmony Search (HS) (Lee and Geem, 2005), Novel Global
Harmony Search (NGHS) (Zou et al., 2010) and ICS. The parameters
of GA, ABC, HS and NGHS algorithms are as follows.
 GA: Chromosome = 50, Crossover = 0.9, Mutation = 0.3, Genera-
tions = 2000.
 ABC: Colony = 20, Limit = 100, Iterations = 1000.
 HS: HMS = 5, HMCR = 0.9,PAR = 0.3, BW = 0.01.
 NGHS: HMS = 5, pm = 0.05.
Considering two sets of problems, the result provided by the ICS
is better than the results given by above-mentioned algorithms.
For all case studies, Table 16 includes the simulation results of
fifty independent runs provided by the CS and ICS. It shows that
the ICS performs better than the CS in all criteria. The parameter
‘NFOS’a represents the number of feasible optimal solutions found
in fifty runs.
According to the above simulations and comparisons, it can be
concluded that the ICS algorithm outperforms the other methods
in literature to find best solutions for the given reliability optimiza-
tion problems.
7. Conclusion
To enhance the accuracy and convergence rate of the cuckoo
search algorithm, an improved cuckoo search algorithm was
developed in this paper. Then, the proposed algorithm was tested
on four well-known reliability optimization problems. Simulation
results confirmed the effectiveness of the proposed algorithm in
comparison with several well-known methods. Moreover, the per-
formance of the proposed algorithm was tested on two complex
engineering optimization problems, namely a large-scale reliability
optimization problem as well as a 15-unit system. Simulation re-
sults demonstrated the superiority of the proposed algorithm over
other well-known methods.
References
Abraham, J. A. (1979). An improved algorithm for network reliability. IEEE
Transaction Reliability, 28, 58–61.
Agrwal, M., & Vikas, K. S. (2010). Ant colony approach to constrained redundancy
optimization in binary systems. Applied Mathematical Modelling, 34, 992–1003.
Aponte, D. E. S., & Sanseverino, C. M. R. (2007). Solving advanced multi-objective
robust designs by means of multiple objective evolutionary algorithms (MOEA):
A reliability application. Reliability Engineering and System Safety, 92(6),
:697–706.
Brown, C., Liebovitch, L. S., & Glendon, R. (2007). Lévy flights in Dobe Ju/hoansi
foraging patterns. Human Ecology, 35, 129–138.
Chen, T. C. (2006). IAs based approach for reliability redundancy alposition
problems. Applied Mathematics and Computation, 182(2), 1556–1567.
Chern, M. S., & Jan, R. H. (1986). Reliability optimization problems with multiple
constraints. IEEE Transactions on Reliability, 35(4), 431–436.
Coelho, L. S. (2009). An efficient particle swarm approach for mixed-integer
programming in reliability–redundancy optimization applications. Reliability
Engineering and System Safety, 94(4), 830–837.
Dhingra, A. K. (1992). Optimal apportionment of reliability & redundancy in series
systems under multiple objectives. IEEE Transactions on Reliability, 41(4),
576–582.
Elegbede, C. (2005). Structural reliability assessment based on particles swarm
optimization. Structural Safety, 27(2), 171–186.
Gen, M., & Kim, J. R. (1999). GA-based reliability design: State-of-the-art survey.
Computers and Industrial Engineering, 37(1–2), 151–155.
Gen, M., & Yun, Y. S. (2006). Soft computing approach for reliability optimization:
State-of-the-art survey. Reliability Engineering and System Safety, 91(9),
1008–1026.
Hikita, M., Nakagawa, H., & Harihisa, H. (1992). Reliability optimization of systems
by a surrogate constraints algorithm. IEEE Transactions on Reliability, 41(3),
473–480.
Holland, J. H. (1975). Adaptation in natural and artificial systems. Ann Arbor, MI:
University of Michigan Press.
Hsieh, Y. C., Chen, T. C., & Bricker, D. L. (1998). Genetic algorithm for reliability
design problems. Microelectronics Reliability, 38(10), 1599–1605.
Karaboga, D. (2005). An idea based on honeybee swarm for numerical optimization.
Technical report TR06, Erciyes University, Engineering Faculty, Computer
Engineering Department, 2005.
Kuo, W. (2007). Recent advances in optimal reliability allocation. IEEE Transactions
on Systems, Man, and Cybernetics – Part A: Systems and Humans, 37(2), 143–156.
Kuo, W., Hwang, C. L., & Tillman, F. A. (1978). A note on heuristic methods in optimal
system reliability. IEEE Transactions on Reliability Reliability, 27(5), 320–324.
Kuo, W., & Prasad, V. R. (2000). An annotated overview of system-reliability
optimization. IEEE Transaction on Reliability, 49(2), 176–187.
468 E. Valian et al. / Computers & Industrial Engineering 64 (2013) 459–468
Lee, K. S., & Geem, Z. W. (2005). A new meta-heuristic algorithm for continuous
engineering optimization, harmony search theory and practice. Computer
Methods in Applied Mechanics and Engineering, 194, 3902–3933.
Marseguerra, M., Zio, E., & Podofillini, L. (2004). Optimal reliability/availability of
uncertain systems via multi-objective genetic algorithms. IEEE Transactions on
Reliability, 53(3), 424–434.
Meziane, R., Massim, Y., Zeblah, A., Ghoraf, A., & Rahli, R. (2005). Reliability
optimization using ant colony algorithm under performance and cost
constraints. Electric Power Systems Research, 76(1–3), 1–8.
Painton, L., & Campbell, J. (1995). Genetic algorithms in optimization of system
reliability. IEEE Transactions on Reliability, 44(2), 172–178.
Payne, R. B., Sorenson, M. D., & Klitz, K. (2005). The cuckoos. Oxford University Press.
Prasad, V. R., & Kuo, W. (2000). Reliability optimization of coherent systems. IEEE
Transactions on Reliability, 49(3), 323–330.
Ramirez-Marquez, J. E. (2008). Port-of-entry safety via the reliability optimization
of container inspection strategy through an evolutionary approach. Reliability
Engineering and System Safety, 93(11), 1698–1709.
Salazar, D., Rocco, C. M., & Galvn, B. J. (2006). Optimization of constrained
multipleobjective reliability problems using evolutionary algorithms. Reliability
Engineering and System Safety, 91(9), 1057–1070.
Wu, P., Gao, L., Zou, D., & Li, S. (2010). An improved particle swarm optimization
algorithm for reliability problems. ISA Transactions, 50(1), 71–81.
Xu, Z., Kuo, W., & Lin, H. H. (1990). Optimization limits in improving system
reliability. IEEE Transactions on Reliability, 39(1), 51–60.
Yang, X.S., Deb, S. (2009). Cuckoo search via Lévy flights. In Proceedings of World
congress on nature & biologically inspired computing (NaBIC 2009, India) (pp. 210–
214).
Yang, X. S., & Deb, S. (2010). Engineering optimisation by cuckoo search.
International Journal of Mathematical Modelling and Numerical Optimisation,
1(4), 330–343.
Yeh, W. C., & Hsieh, T. J. (2011). Solving reliability redundancy allocation problems
using an artificial bee colony algorithm. Computers & Operations Research, 38,
1465–1473.
Yin, P. Y., Yu, S. S., Wang, P. P., & Wang, Y. T. (2007). Task alposition for maximizing
reliability of a distributed system using hybrid particle swarm optimization.
Journal of Systems and Software, 80(5), 724–735.
Yokota, T., Gen, M., & Li, H. H. (1996). Genetic algorithm for nonlinear mixed-integer
programming problems and its application. Computers and Industrial
Engineering, 30(4), 905–917.
Zou, D., Gao, L., Li, S., & Wu, J. (2011). An effective global harmony search algorithm
for reliability problems. Expert Systems with Applications, 38, 4642–4648.
Zou, D., Gao, L., Wu, J., Li, S., & Li, Y. (2010). A novel global harmony search algorithm
for reliability problems. Computers & Industrial Engineering, 58, 307–316.