Applied Soft Computing 19 (2014) 93101

Contents lists available at ScienceDirect

Applied Soft Computing

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

Improved cuckoo search algorithm for hybrid ow shop scheduling

problems to minimize makespan
M.K. Marichelvam a, , T. Prabaharan b , X.S. Yang c
a
Department of Mechanical Engineering, Kamaraj College of Engineering and Technology, Virudhunagar, Tamilnadu 626001, India
b
Department of Mechanical Engineering, Mepco Schlenk Engineering College, Sivakasi, Tamilnadu 626005, India
c
School of Science and Technology, Middlesex University, London NW4 4BT, UK

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

Article history: The multistage hybrid ow shop (HFS) scheduling problems are considered in this paper. Hybrid owshop
Received 30 May 2012 scheduling problems were proved to be NP-hard. A recently developed cuckoo search (CS) metaheuristic
Received in revised form 27 January 2014 algorithm is presented in this paper to minimize the makespan for the HFS scheduling problems. A
Accepted 3 February 2014
constructive heuristic called NEH heuristic is incorporated with the initial solutions to obtain the optimal
Available online 13 February 2014
or near optimal solutions rapidly in the improved cuckoo search (ICS) algorithm. The proposed algorithm
is validated with the data from a leading furniture manufacturing company. Computational results show
Keywords:
that the ICS algorithm outperforms many other metaheuristics.
Hybrid ow shop (HFS)
Scheduling 2014 Elsevier B.V. All rights reserved.
NP-hard
Improved cuckoo search (ICS)
Metaheuristics

1. Introduction
Scheduling is a process of allocating resources over time to
perform a collection of required tasks [1]. It is a decision making
process. Effective scheduling plays a vital role for the growth of
any kind of industries. Different types of scheduling problems were
addressed in the literature with different performance measures.
Single machine scheduling, ow shop scheduling, parallel machine
scheduling, job shop scheduling, HFS scheduling and project sched-
uling are some of the important types of scheduling environments.
Among them HFS scheduling problems are considered in this paper.
Minimization of makespan, total ow time, mean ow time, total
earliness, total tardiness, weighted earliness and tardiness and
number of tardy jobs are some of the important objective func-
tions of scheduling problems. The objective of this paper is to
minimize makespan. Many researchers concentrated on the HFS
scheduling problems for the past several decades since it was
proposed by Arthanari and Ramamurthy [2] due to their practi-
cal importance. Many industries like steel, ceramic, electronics,
chemical and textile industries resemble the HFS environment. The
HFS environment can be assumed to be the combination of ow
shop and parallel machines. In the HFS environment, machines are
arranged into several stages in series. Each stage has one or more
Corresponding author. Tel.: +91 9751043410; fax: +91 04549 378171.
E-mail address: mkmarichelvamme@gmail.com (M.K. Marichelvam).
identical machines in parallel. A job must pass through all stages.
Moreover, the job must be processed by exactly one machine at
every stage. The layout of a k stage hybrid owshop scheduling
is shown in Fig. 1 where Mmk represents the machine in the mth
parallel and kth stage [3].
Researchers have proposed many different approaches to solve
the HFS problems. Exact solution methods, heuristics and meta-
heuristics are the three important approaches addressed in the
literature [4]. One of most important, exact solution methods used
for the HFS problem is the branch & bound algorithm [58]. Port-
mann and Vignier [9] improved the performance of the branch
& bound algorithm by hybridizing the genetic algorithm (GA).
Neron et al. [10] solved the HFS scheduling problems using branch
and bound algorithm. They applied energetic reasoning and global
operations for enhancing the efciency of branch and bound algo-
rithm to minimize the makespan. The HFS scheduling problems
were proved to be NP-hard by Gupta [11] and Hoogeveen et al. [12].
Hence, many heuristics and metaheuristics have been developed
to search for optimal or near-optimal solutions for solving the HFS
scheduling problems. Gupta and Tunc [13] proposed four heuris-
tics to minimize makespan for two stage HFS scheduling problems
with separable setup and removal times. Lee and Vairaktarakis [14]
addressed heuristics to minimize makespan for solving the HFS
problems. Hunsucker and Shah [15] evaluated the performance of
different dispatching rules for a constrained HFS scheduling envi-
ronment. Brah and Loo [16] developed a heuristic procedure to
solve ow shop scheduling problems with multiple processors.

http://dx.doi.org/10.1016/j.asoc.2014.02.005
1568-4946/ 2014 Elsevier B.V. All rights reserved.
94 M.K. Marichelvam et al. / Applied Soft Computing 19 (2014) 93101
M11
Start M21
Mm1
Fig. 1. Layout of the hybrid ow shop environment.
Negenman [17] presented a local search method to solve HFS prob-
lems. Soewnadi and Elmaghraby [18] suggested several heuristics
for the three-stage HFS with the multiple processors to minimize
makespan. Oguz et al. [19] also proposed some heuristics to solve
multiprocessor task scheduling in a two-stage ow shop scheduling
problems. He and Babayan [20] solved three stage exible ow
shop scheduling problems using an agent based approach. Kypari-
sis and Koulamas [21] developed some heuristics to minimize the
makespan for the multistage exible ow shop scheduling problem
with uniform parallel machines in each stage.
Recently, researchers have applied many metaheuristics to solve
the HFS problems. Engin and Dyen [22] proposed an articial
immune system algorithm for solving the HFS problems to min-
imize the makespan. Zandieh et al. [23] also presented an immune
algorithm approach for solving the HFS problems. They considered
the sequence-dependent setup times. Ying [24] proposed a new
hybrid immune algorithm and iterated greedy algorithm to min-
imize the makespan in the multistage HFS scheduling problems.
Yang et al. [25] solved the ow shop scheduling with multiple
processor problems by using a tabu search simulation optimiza-
tion approach. They have considered a case study problem of a
multilayer ceramic capacitor manufacturing company to minimize
tardiness. Bozejko et al. [26] addressed a parallel tabu search algo-
rithm to solve the HFS scheduling problems. Many researchers have
applied different versions of genetic algorithm (GA) to solve the HFS
problems [2732]. Ying and Lin [33] developed an ant colony sys-
tem heuristic for the multistage HFS problem with multiprocessor
tasks to minimize the makespan. They tested the algorithm over
the benchmark problems from the literature. Alaykyran et al. [34]
also suggested an ant colony optimization algorithm for solving
the HFS problems. Wang et al. [35] proposed a simulated anneal-
ing (SA) algorithm for the HFS scheduling problems to minimize the
makespan. Tseng and Liao [36] presented particle swarm optimiza-
tion (PSO) algorithm to solve the HFS problems. Recently, Liao et al.
[4] developed a PSO hybridized with bottleneck heuristic to solve
the HFS problems to minimize makespan. The different HFS envi-
ronments, constraints and solution methodologies can be found in
[37] and [38].
In this paper, a new metaheuristic algorithm called cuckoo
search (CS) algorithm is used to solve the HFS problems. Cuckoo
search algorithm is one of the population based metaheuristic
algorithms developed by Yang and Deb in 2009 [39]. Yang and
Deb [40] solved various optimization problems using the cuckoo
search algorithm. Yang and Deb [41] also proposed a multiobjec-
tive cuckoo search algorithm for design optimization problems.
Gandomi et al. [42] solved the structural optimization problems
M12 Mik
M22 M2k End
Mm2 Mmk
using cuckoo search algorithm. Walton et al. [43] proposed a mod-
ied cuckoo search algorithm to solve the optimization problems.
Layeb [44] addressed a novel quantum inspired cuckoo search for
knapsack problems. Durgun and Yildiz [45] used the cuckoo search
algorithm to solve design optimization problems for vehicle com-
ponents. Yildiz [46] determined the optimal machining parameters
for milling operations using the cuckoo search algorithm.
From the current literature, one can easily understand that
the cuckoo search algorithm is applied to solve different types
of optimization problems. However, the application of cuckoo
search algorithm to solve the scheduling problems is very limited.
Marichelvam [47] proposed an improved hybrid cuckoo search
algorithm for solving the permutation ow shop scheduling prob-
lems. Chandrasekaran and Simon [48] proposed a hybrid cuckoo
search algorithm integrated with fuzzy system for solving multi-
objective scheduling problems. Burnwal and Deb [49] solved
the exible manufacturing scheduling problem using the cuckoo
search algorithm. They compared the results with GA and PSO.
Hence in this paper, we apply the cuckoo search algorithm to solve
the HFS problems. Similar to many metaheuristic algorithms, the
initial solutions in the cuckoo search algorithm are also generated
randomly. But in this paper a constructive heuristic called NEH
heuristic developed by Nawaz et al. [50] is applied to generate one
of the initial solutions to improve the solution quality. In addition,
the parameters of the cuckoo search algorithm are tuned in this
paper to increase the efciency of the algorithm. Hence, in this
paper, an improved cuckoo search (ICS) algorithm is presented to
minimize makespan for the HFS scheduling problems.
The paper is organized as follows: the problem denition is pre-
sented in Section 2. The CS and ICS algorithms are explained in
Section 3. Computational results are presented in Section 4. Con-
clusions and future research opportunities will be discussed in
Section 5.
2. Problem denition
The HFS scheduling problems considered in this work consist
of k stages in series. Each stage i (i = 1,2,. . .,k) consists of m iden-
tical machines in parallel. A set of N jobs j (j = 1,2,. . .,N) has to be
processed on anyone of the machines available at each stage. The
processing times of job j at various stages are P1j ,P2j ,. . .,Pkj , and it
may be zero for some jobs as the jobs are not processed in some
stages. Processing time is the time required to process a job in
a stage. The objective is to minimize the makespan (Cmax ). The
makespan for a scheduling problem is dened as the completion
 M.K. Marichelvam et al. / Applied Soft Computing 19 (2014) 93101
time of the last job to leave the production system [1]. Makespan
is used for measuring the system utilization.
The makespan should be at least equal to the completion time
of the last job at the last stage. Completion time is the moment at
which the job is completed in a stage. Completion time of the job j
at stage s is the sum of the starting time of job j at stage s and the
processing time of job j at stage s. Processing of a job can be started
in the rst stage only if the job is ready. The ready time of the job j is
the moment at which the job j is available for processing. Each job
should be assigned to exactly one machine at each stage. Processing
of a job can be started in a stage only when it has been completed
at the previous stage. The makespan value should be positive. The
formulation of mathematical model can be found in [51].
In order to solve the HFS scheduling problem, the following
assumptions are considered in this paper:
1. The number of stages and the number of machines at each stage
are known in advance.
2. The numbers of jobs, their processing times are known in
advance and are xed.
3. All the jobs and the machines are available at time zero.
4. No pre-emption is allowed.
5. The setup and transportation times of the jobs are independent
of the sequence and are included in the processing times.
6. Each machine can process only one job at a time.
7. All the machines are available for the entire period of scheduling
(no machine breakdown).
The assumptions 1, 2, 3, 4 and 6 are common for most of the real
cases. For instance, the rst assumption is applicable for all indus-
trial scheduling problems as the scheduler knows the number of
stages and the number of machines at each stage in advance. How-
ever, the assumptions 5 and 7 might not be real. These assumptions
were included to simplify the problem.
3. Cuckoo search algorithm
Cuckoo search (CS) algorithm is a new nature-inspired meta-
heuristic algorithm developed by Yang and Deb. CS algorithm was
inspired by the obligate brood parasitic behaviour of some cuckoo
species in combination with the Lvy ight behaviour of some birds
and fruit ies in nature. The breeding behaviour and the Lvy ights
will be discussed in the following sections. The details may also be
found in [39].
3.1. Cuckoo breeding behaviour
Some of the cuckoo species lay their eggs in the nests of other
host birds. The cuckoos often select the recently spawned nests
instinctly. They may remove others eggs to increase the hatching
probability of their own eggs. Some host birds can engage direct
conict with the intruding cuckoos. If a host bird discovers the eggs
are not their owns, they will either throw these alien eggs away or
simply abandon its nest and build a new nest elsewhere. Some cuck-
oos have evolved in such a way that female parasitic cuckoos are
often very specialized in the mimicry in colour and pattern of the
eggs of a few chosen host birds. This will reduce the probability of
their eggs being abandoned. This also increases their reproductiv-
ity. Furthermore, the timing of egg-laying of some cuckoos is also
amazing. The 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 rst cuckoo chick is hatched, the rst
instinct action it will take is to evict the host eggs by blindly pro-
pelling the eggs out of the nest. This will increase the cuckoo chicks
share of food provided by its host bird. Moreover, a cuckoo chick
95
can also mimic the call of host chicks to gain access to more feeding
opportunity.
3.2. Lvy ights
In general, animals search for food in a random or quasi-random
manner. The foraging path of any animal is effectively a random
walk as the next move is based on the current location or state
and the transition probability to the next location. The probabil-
ity for choosing depends implicitly on a probability which can be
modelled mathematically. The ight behaviour of many animals
and insects has demonstrated the typical characteristics of Lvy
ights [39]. Hence, such behaviour has been applied to optimiza-
tion and optimal search, and preliminary results show its promising
capability.
3.3. Cuckoo search algorithm
Based on the brooding behaviour of cuckoos, we can now
describe the basic steps of cuckoo search using the following three
idealized rules/approximations:
1. Each cuckoo lays one egg (solution) at a time, and dumps its egg
in a randomly chosen nest. That is, an egg represents a solution.
As there is one egg in one nest, then we can use the simplest case
that an egg is equivalent to a nest and a solution.
2. The best nests with high quality eggs/solutions will carry out to
the next generation. Here is the best means the solution with the
best (minimum) objective values (for minimization problems).
3. The egg laid by a cuckoo can be discovered by the host bird with
a probability Pa and a nest will then be built. That is to say, a
fraction Pa of the n nests being replaced by new nests (with new
random solutions at new locations).
For minimization problems the quality or tness function value
may be the reciprocal of the objective function. Each egg in a nest
represents a solution and the cuckoo egg represents a new solution.
Therefore, there is no difference between an egg, a nest and solu-
tion. The pseudo code of the cuckoo search algorithm is presented
in Fig. 2.
When generating new solutions xa (t+1) for ath cuckoo a Lvy
ight is performed using the following equation:
xat+1 = xat + s H(pa ) (xjt xkt ), (1)
where xjt and xkt are two different solutions selected randomly by
random permutation, H(u) is a Heaviside function, is a random
number drawn from a uniform distribution, and s is the step size.
On the other hand, the global random walk is carried out by using
Lvy ights
 () sin(/2) 1
xit+1 = xit + L(s, ), L(s, ) = ,
 s1+
(s  s0 > 0). (2)
where gamma (lamda) is the Gamma function and lamda the ran-
dom step length. Lvy ights essentially provide a random walk
while their random steps are drawn from a Lvy distribution for
large steps (1 <  3). Here the step size vector L is drawn from the
power-law distribution as given in (2).
The cuckoo search algorithm consists of three parameters Pa
(probability of worse nests to be abandoned), (step size), and 
(random step length). Among them the parameters Pa and are
very important to obtain better solutions. In the literature these
parameters of the cuckoo search algorithm are kept constant. But,
in this paper the parameters Pa , , and  are not kept constant. The
96 M.K. Marichelvam et al. / Applied Soft Computing 19 (2014) 93101

Start
Objective function f(x), x = (x1, x2, ..., xd)
Generate initial population of n host nests xi (i = 1, 2,..., n)
While (t <Max Generation)
Get a cuckoo randomly (say a) by Lvy flights
evaluate its quality/fitness Fa [proportional to f(x)]
Choose a nest among n (say b) randomly
if (Fa is better than Fb),
replace solution b by the new solution;
end
Rank the fitness of the solutions/nests
Abandon a fraction Pa of worse nests
Keep the best solutions (or nests with top quality solutions)
Find the current best solution
end while
Post process results and visualization
End

Fig. 2. Pseudo code of the cuckoo search (CS) algorithm.

parameters of Pa and are determined by the design of experi-
ments. The values used in our simulations are based on extensive
parametric studies [39], and we will choose n = 1540,  = 1.5, and
= 0.01*abs(Ub Lb), where Ub and Lb are the upper bound and
lower bound, respectively, of the problem of interest.
Most metaheuristic algorithms use uniform and Gaussian ran-
domization techniques. Here we have used Lvy ights. One of the
reasons is that Lvy ights are more efcient than random walks
based on Gaussian random steps. This is because the variance (rep-
resenting the average squared distance) of the Lvy ights increases
nonlinearly, faster than the linear relationship in Gaussian random
walks. In fact, the mean and variance can approach innite, and thus
Lvy ights are more suitable for explorative large-scale search.
Another reason to use Lvy ights is that the occasional larger
steps drawn from a Levy distribution can enable the search algo-
rithm to have the capability to escape from a local minimum, and
thus increase the probability of nding the true global optimality
[52,53].
3.4. Improved cuckoo search (ICS) algorithm
are arranged in the decreasing order of the sum. Then the steps 2
and 3 of the NEH heuristic are applied. The solution obtained by
the NEH heuristic is used as an initial solution and the remaining
solutions are generated randomly.
As for any metaheuristic algorithm, if we can start with bet-
ter solutions that are closer to the potential optimal solutions, we
can save some computational time. Thus, it requires fewer func-
tion evaluations than those random initial solutions. This is, in fact,
to incorporate one known design or knowledge of the problem
of interest. A simple way is to use a two-stage procedure using
two different methods; that is to run one method such as NEH
heuristic to get some crude solutions which are better than ini-
tial, random solutions. Then, we use these solutions obtained by
the rst method as the new initialized solutions for cuckoo search,
and consequently run CS to get the nal better and/or optimal solu-
tions. It can be expected that this two-stage method can produce
better results than each individual method alone. In fact, from our
simulations given later, this ICS can indeed obtain excellent results
with superior efciency.

Similar to many metaheuristic algorithms the solutions are gen-
erated randomly in the cuckoo search algorithm. But, studies have
reported that the performance of the metaheuristic algorithms
might be improved by generating some of the initial solutions by
using some constructive heuristics that incorporate the knowledge
of specic type of problems of interest. Hence in this paper, we try
to incorporate a constructive heuristic called as NEH heuristic into
the standard cuckoo search algorithm so as to improve the solu-
tion quality for the HFS scheduling problems. The NEH heuristic
was originally developed to minimize the makespan for n-job m-
machine owshop scheduling problems. The NEH heuristic consists
of the following three steps:
Step 1: The sum of processing times of all the jobs on all
machines is calculated. Then the jobs are sorted in decreasing order
of the sum. Let the sequence be j1 , j2 ,. . ., jn .
Step 2: The rst two jobs j1 and j2 are taken and ordered them so
that the makespan of the sequence given by the ordering of these
two jobs is minimized.
Step 3: The remaining jobs j3 , j4 ,. . ., jn are inserted in each possi-
ble position in the sequence that results in the smallest makespan
value.
Ruiz and Maroto [54] proved that the NEH heuristic is the
best constructive heuristic to minimize the makespan for the ow
shop scheduling problems. Hence, it is planned to implement the
NEH heuristic to solve the HFS scheduling problems. The sum of
processing times for all the jobs in all stages is calculated. The jobs
4. Computational results
The proposed ICS algorithm for the HFS problem was coded in
C++ and run on a Pentium 4 3.4 GHz PC with 512 MB memory.
4.1. Case study problem
Now we consider a real-world application. The scheduling prob-
lem of a furniture manufacturing company is considered in this
paper. The company is one of the leading manufacturers in this
eld in India. The company produces a variety of furniture products.
Among them, the ten drawer horizontal media storage cabinets are
considered in this research work. The ten drawer horizontal media
storage cabinets consist of 20 different parts. Each part may be con-
sidered as a job. Hence, we consider 20 jobs in this problem. The
parts are produced in batches. The batch size is 60. The production
system consists of ve different stages. The ve stages are punching,
bending, welding, power pressing and drilling. Each stage consists
of the different number of identical parallel machines. The layout
of the case study company is shown in Fig. 3.
The processing time is the time required for a job to be processed
in different stages. The processing time taken by each job is mea-
sured many times and the average time is taken in this work. The
processing time (in seconds) of all 20 jobs in different stages are
given in Table 1. The rst job needs two operations namely punch-
ing and bending. The processing time for these two operations is 54
 M.K. Marichelvam et al. / Applied Soft Computing 19 (2014) 93101 97

1
1 1
2
1
2 3 2

4
3 3 2
5

4 6 4 3

7
5 5
8

Punching Bending Welding Power pressing Drilling

stage stage stage stage stage

Fig. 3. Layout of the case study company.

and 48 s, respectively. The processing time is zero for other oper-
ations. The 16th job only needs the drilling operation. Hence, the
processing time is zero for all other jobs.
In the furniture company, neither a heuristic nor an algorithm
is applied to obtain the sequence. Instead the sequence is obtained
only by following the First In First Out (FIFO) sequence. Hence, the
company sequence is 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
18 19 20. The makespan for this sequence is 13640 s. In order to
minimize the makespan value the ICS algorithm is applied. By this
ICS algorithm, a new sequence (15 8 5 18 19 9 12 20 4 6 7 3 16
13 14 17 2 11 10 1) is obtained and the makespan is minimized to
10700 s. In order to compare the performance of the proposed ICS
with other existing algorithms, we have solved the same problem
using genetic algorithm (GA), simulated annealing (SA), ant colony
optimization (ACO), particle swarm optimization (PSO), and the
standard cuckoo search (CS) addressed in the literature. The results
obtained by the ICS algorithm for the case study problem is com-
pared with GA [29], SA [35], ACO [33], PSO [36], CS [49] algorithms
and NEH heuristic [50]. We tested the algorithms with the similar
parameters used in the respective literature papers and the same

Table 1
Processing times (in seconds) of different jobs.

Stages Punching stage Bending stage Welding stage Power pressing stage Drilling stage
Job number

1 54 48 0 0 0
2 35 63 16 0 0
3 59 22 61 0 0
4 60 0 0 0 0
5 22 36 0 0 0
6 57 11 72 0 0
7 31 0 0 0 0
8 19 0 0 68 0
9 48 72 59 0 0
10 57 20 24 0 0
11 0 0 0 91 0
12 0 0 0 28 0
13 47 76 0 0 0
14 0 0 0 64 0
15 0 0 0 29 0
16 72 53 0 0 39
17 0 0 0 78 0
18 0 0 0 60 0
19 0 0 0 42 0
20 0 0 0 75 0

Table 2
Result comparison of case study problem.

Algorithm Best sequence Best makespan (in seconds) Best CPU time (in seconds)

FIFOa 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 13640 b
ICS 15 8 5 18 19 9 12 20 4 6 7 3 16 13 14 17 2 11 10 1 10700 8.58
GA 14 8 16 1 7 4 17 13 11 2 15 12 3 10 5 9 20 19 6 18 10892 13.28
SA 19 8 14 4 13 6 18 9 15 11 20 16 3 2 7 17 12 1 5 10 10920 12.25
ACO 14 5 13 4 15 20 19 12 1 16 18 11 6 7 8 10 17 2 9 3 10968 12.32
PSO 12 4 19 5 14 17 3 2 15 6 9 11 16 10 18 13 8 7 1 20 10860 11.12
CS 18 20 16 12 1 9 17 14 2 19 15 10 7 13 8 11 5 6 4 3 10776 9.82
NEH 20 19 8 7 11 12 3 5 4 18 13 17 9 16 6 10 14 1 2 15 11180 18.64
a
Company sequence.
b
In the industry, the sequence is obtained manually. Hence, there is no computational time.
98 M.K. Marichelvam et al. / Applied Soft Computing 19 (2014) 93101

Table 3 Table 4
Result comparison of case study problem. Percentage deviation in makespan.

Algorithm Makespan (in seconds) CPU time (in seconds) Problem FIFO GA SA ACO PSO CS NEH

Average Standard Average Standard Case study 21.55 1.76 2.01 2.44 1.47 0.71 4.29
deviation deviation

FIFOa 13640 b b b
ICS 10777 21.875 8.58 0.053 Table 5
GA 10979 29.863 13.28 0.073 Factor levels for the design of experiments.
SA 11007 28.067 12.25 0.063
S. No. Factors Levels
ACO 11039 21.194 12.32 0.054
PSO 10934 24.818 11.12 0.068 1 M 5 (small size problems)
CS 10853 22.033 9.82 0.052 10 (medium size
NEH 11252 20.142 18.64 0.062 problems)
a
20 (large size
Company sequence.
b
problems)
In the industry, the sequence is obtained manually. Hence, there is no compu-
2 N 10, 50 and 100
tational time.
3 Pa Uniform [0.01, 0.5]
4 Uniform [0.05, 1.0]
5  Uniform [1.5, 2.5]
number of iterations. The parameters are presented in Appendix A.
The results are presented in Table 2. From the results in the above
table, it can be concluded that the ICS algorithm performs better ICS algorithm, we adapted percentage deviation as a performance
than the GA, SA, ACO, PSO, CS algorithms and NEH heuristics. The measure. The percentage deviation (PD) is calculated by using the
ICS algorithm provides better results than the CS algorithm. This is following formula:
due to the NEH constructive heuristics used to generate one of the
CX CICS
initial solutions. PD = 100 (3)
CX
Monte-Carlo simulations were conducted 20 times for the case
study problem. The mean and standard deviation of the makespan where CX = makespan obtained by FIFO, GA, SA, ACO, PSO, CS
were calculated. Moreover, the mean and standard deviation algorithms and NEH.
of the computational time were also calculated. The results are Heuristics
presented in Table 3. In order to evaluate the performance of the CICS = makespan obtained by ICS algorithm.

Table 6
Makespan comparison of different algorithms.

Problem No. M mk n Pa  Best makespan (in seconds)

GA SA ACO PSO CS ICS NEH

1 10 2 10 0.05 0.27 1.67 2590 2592 2594 2580 2540 2530 2632
2 10 5 10 0.22 0.45 1.87 3114 3118 3120 3090 3080 3060 3142
3 10 5 10 0.31 0.13 2.09 3514 3524 3532 3471 3460 3430 3546
4 10 2 10 0.34 0.52 2.17 3060 3080 3080 3030 3030 3010 3110
5 10 5 10 0.42 0.63 2.45 2376 2380 2390 2343 2340 2320 2408
6 10 5 50 0.32 0.07 1.84 6284 6288 6292 6208 6192 6172 6382
7 10 2 50 0.19 0.22 2.32 7286 7292 7298 7268 7242 7228 7386
8 10 2 50 0.44 0.17 2.46 8172 8184 8186 8142 8128 8116 8340
9 10 5 50 0.07 0.55 1.68 7052 7056 7062 7028 7024 7008 7168
10 10 5 50 0.16 0.73 2.08 5372 5386 5384 5352 5342 5326 5439
11 10 2 100 0.46 0.82 1.62 13104 13114 13126 12272 12258 12212 13270
12 10 5 100 0.32 0.67 1.95 10929 10942 10954 10898 10872 10848 11181
13 10 5 100 0.24 0.83 2.04 11506 11524 11528 11472 11458 11436 11694
14 10 2 100 0.37 0.54 2.42 12124 12132 12142 12098 12082 12056 12268
15 10 2 100 0.27 0.23 2.34 11522 11526 11528 11495 11472 11458 11656

Table 7
Makespan comparison of different algorithms.

Problem No. M mk n Pa  Average makespan (in seconds)

GA SA ACO PSO CS ICS NEH

1 10 2 10 0.05 0.27 1.67 2611 2613 2611 2598 2558 2548 2649
2 10 5 10 0.22 0.45 1.87 3139 3143 3140 3112 3102 3082 3162
3 10 5 10 0.31 0.13 2.09 3542 3552 3555 3495 3485 3454 3569
4 10 2 10 0.34 0.52 2.17 3084 3105 3100 3051 3051 3031 3130
5 10 5 10 0.42 0.63 2.45 2395 2399 2406 2359 2357 2336 2424
6 10 5 50 0.32 0.07 1.84 6334 6338 6333 6251 6236 6216 6423
7 10 2 50 0.19 0.22 2.32 7344 7350 7346 7319 7293 7279 7434
8 10 2 50 0.44 0.17 2.46 8237 8249 8240 8199 8186 8174 8394
9 10 5 50 0.07 0.55 1.68 7108 7112 7108 7077 7074 7058 7214
10 10 5 50 0.16 0.73 2.08 5415 5429 5419 5389 5380 5364 5474
11 10 2 100 0.46 0.82 1.62 13209 13219 13212 12358 12345 12299 13356
12 10 5 100 0.32 0.67 1.95 11016 11030 11026 10974 10949 10925 11253
13 10 5 100 0.24 0.83 2.04 11598 11616 11604 11552 11539 11517 11770
14 10 2 100 0.37 0.54 2.42 12221 12229 12222 12182 12168 12142 12347
15 10 2 100 0.27 0.23 2.34 11614 11618 11604 11575 11553 11539 11731
 M.K. Marichelvam et al. / Applied Soft Computing 19 (2014) 93101 99

Table 8
Makespan comparison of different algorithms.

Problem No. M mk n Pa  Standard deviation of makespan

GA SA ACO PSO CS ICS NEH

1 10 2 10 0.05 0.27 1.67 7.102 6.663 5.013 5.897 5.193 5.172 4.742
2 10 5 10 0.22 0.45 1.87 8.538 8.015 6.029 7.064 6.297 6.256 5.660
3 10 5 10 0.31 0.13 2.09 9.634 9.058 6.826 7.934 7.075 7.012 6.389
4 10 2 10 0.34 0.52 2.17 8.388 7.918 5.952 6.926 6.194 6.153 5.603
5 10 5 10 0.42 0.63 2.45 6.514 6.117 4.620 5.355 4.785 4.742 4.339
6 10 5 50 0.32 0.07 1.84 17.228 16.162 12.159 14.190 12.659 12.618 11.497
7 10 2 50 0.19 0.22 2.32 19.976 18.743 14.104 16.614 14.805 14.776 13.307
8 10 2 50 0.44 0.17 2.46 22.405 21.035 15.821 18.612 16.618 16.593 15.025
9 10 5 50 0.07 0.55 1.68 19.334 18.136 13.647 16.065 14.360 14.328 12.913
10 10 5 50 0.16 0.73 2.08 14.729 13.844 10.404 12.233 10.921 10.889 9.798
11 10 2 100 0.46 0.82 1.62 35.928 33.708 25.367 28.053 25.060 24.967 23.907
12 10 5 100 0.32 0.67 1.95 29.964 28.127 21.170 24.911 22.226 22.178 20.143
13 10 5 100 0.24 0.83 2.04 31.547 29.621 22.280 26.223 23.424 23.380 21.068
14 10 2 100 0.37 0.54 2.42 33.241 31.184 23.466 27.653 24.701 24.648 22.101
15 10 2 100 0.27 0.23 2.34 31.590 29.626 22.280 26.275 23.453 23.424 20.998

Table 9
CPU time comparison of different algorithms.

Problem No. M mk n Pa  Best CPU time (in seconds)

GA SA ACO PSO CS ICS NEH

1 10 2 10 0.05 0.27 1.67 5.22 5.38 5.42 4.80 4.62 4.58 7.12
2 10 5 10 0.22 0.45 1.87 5.46 5.68 5.82 5.06 4.98 4.92 7.68
3 10 5 10 0.31 0.13 2.09 5.82 5.88 5.94 5.42 5.34 5.04 8.12
4 10 2 10 0.34 0.52 2.17 5.48 5.73 5.84 5.04 4.84 4.78 7.32
5 10 5 10 0.42 0.63 2.45 4.72 4.78 4.83 4.68 4.53 4.36 6.78
6 10 5 50 0.32 0.07 1.84 6.58 6.74 6.84 6.36 6.28 6.24 9.46
7 10 2 50 0.19 0.22 2.32 7.08 7.14 7.28 6.84 6.72 6.58 9.72
8 10 2 50 0.44 0.17 2.46 7.84 7.92 7.94 7.56 7.42 7.24 10.96
9 10 5 50 0.07 0.55 1.68 7.04 7.12 7.16 6.92 6.84 6.56 10.42
10 10 5 50 0.16 0.73 2.08 6.94 6.98 7.03 6.74 6.68 6.58 10.66
11 10 2 100 0.46 0.82 1.62 11.76 11.84 11.88 11.56 11.48 11.38 16.28
12 10 5 100 0.32 0.67 1.95 9.86 9.94 9.98 9.72 9.67 9.59 14.38
13 10 5 100 0.24 0.83 2.04 10.74 10.78 10.82 10.58 10.46 10.34 15.68
14 10 2 100 0.37 0.54 2.42 11.84 11.88 11.94 11.68 11.56 11.44 17.08
15 10 2 100 0.27 0.23 2.34 10.86 10.92 10.94 10.64 10.58 10.46 15.96

The PD for the case study problem is presented in Table 4.
The ICS algorithm improves the existing company result by
21.55%. That is the direct labour cost may be reduced by 21.55% by
this ICS algorithm. In addition to this, the level of customer satisfac-
tion is also improved. This would be useful for the further growth
of the company.
4.2. Additional performance evaluation
We conduct a set of experiments to evaluate the performance
of the proposed ICS algorithm. The different factor levels for the
design of experiments are given in Table 5. The job processing
time is uniformly distributed over [1,50] seconds. The number

Table 10
CPU time comparison of different algorithms.

Problem No. M mk n Pa  Average CPU time (in seconds)

GA SA ACO PSO CS ICS NEH

1 10 2 10 0.05 0.27 1.67 5.30 5.46 5.50 4.88 4.70 4.66 7.20
2 10 5 10 0.22 0.45 1.87 5.54 5.76 5.91 5.14 5.07 5.01 7.77
3 10 5 10 0.31 0.13 2.09 5.91 5.97 6.03 5.51 5.43 5.13 8.21
4 10 2 10 0.34 0.52 2.17 5.56 5.81 5.93 5.12 4.92 4.86 7.40
5 10 5 10 0.42 0.63 2.45 4.79 4.85 4.90 4.76 4.61 4.44 6.86
6 10 5 50 0.32 0.07 1.84 6.68 6.84 6.94 6.47 6.39 6.35 9.57
7 10 2 50 0.19 0.22 2.32 7.19 7.25 7.39 6.95 6.84 6.69 9.83
8 10 2 50 0.44 0.17 2.46 7.96 8.04 8.06 7.69 7.55 7.37 11.08
9 10 5 50 0.07 0.55 1.68 7.15 7.23 7.27 7.04 6.96 6.67 10.54
10 10 5 50 0.16 0.73 2.08 7.05 7.08 7.13 6.85 6.80 6.69 10.78
11 10 2 100 0.46 0.82 1.62 11.94 12.01 12.06 11.75 11.68 11.58 16.46
12 10 5 100 0.32 0.67 1.95 10.01 10.09 10.13 9.88 9.84 9.76 14.54
13 10 5 100 0.24 0.83 2.04 10.90 10.94 10.98 10.76 10.64 10.52 15.86
14 10 2 100 0.37 0.54 2.42 12.02 12.06 12.12 11.87 11.76 11.64 17.27
15 10 2 100 0.27 0.23 2.34 11.03 11.08 11.10 10.82 10.76 10.64 16.14
100 M.K. Marichelvam et al. / Applied Soft Computing 19 (2014) 93101

Table 11
CPU time comparison of different algorithms.

Problem No. M mk n Pa  Standard deviation of CPU time

GA SA ACO PSO CS ICS NEH

1 10 2 10 0.05 0.27 1.67 0.029 0.028 0.024 0.030 0.025 0.029 0.024
2 10 5 10 0.22 0.45 1.87 0.030 0.029 0.026 0.031 0.027 0.032 0.026
3 10 5 10 0.31 0.13 2.09 0.032 0.030 0.026 0.034 0.029 0.032 0.028
4 10 2 10 0.34 0.52 2.17 0.030 0.029 0.026 0.031 0.026 0.031 0.025
5 10 5 10 0.42 0.63 2.45 0.026 0.024 0.022 0.029 0.025 0.028 0.023
6 10 5 50 0.32 0.07 1.84 0.036 0.034 0.030 0.040 0.034 0.040 0.032
7 10 2 50 0.19 0.22 2.32 0.039 0.037 0.032 0.042 0.037 0.042 0.033
8 10 2 50 0.44 0.17 2.46 0.043 0.041 0.035 0.047 0.040 0.047 0.037
9 10 5 50 0.07 0.55 1.68 0.039 0.036 0.032 0.043 0.037 0.042 0.035
10 10 5 50 0.16 0.73 2.08 0.038 0.036 0.031 0.042 0.036 0.042 0.036
11 10 2 100 0.46 0.82 1.62 0.064 0.061 0.053 0.072 0.062 0.073 0.055
12 10 5 100 0.32 0.67 1.95 0.054 0.051 0.044 0.060 0.053 0.062 0.049
13 10 5 100 0.24 0.83 2.04 0.059 0.055 0.048 0.066 0.057 0.067 0.053
14 10 2 100 0.37 0.54 2.42 0.065 0.061 0.053 0.073 0.063 0.074 0.058
15 10 2 100 0.27 0.23 2.34 0.059 0.056 0.049 0.066 0.058 0.067 0.054

of machines mk at kth stage is chosen randomly over [2,5]. The
parameters Pa , , and  are also taken from the values presented
in Table 5. In total, we have tested 15 problems with different
parameters. Monte-Carlo simulations were conducted 20 times
for the random problems also. The results obtained by the ICS
algorithm are compared with GA, SA, ACO, PSO, CS algorithms and
NEH heuristics. The best makespan comparison is presented in
Table 6. The average makespan comparison is presented in Table 7.
The standard deviation of makespan is presented in Table 8. The
best, average and standard deviation of the computational time
are presented in Tables 911 respectively.
From the results, it is observed that the proposed ICS algorithm
provides optimal results with minimum computational time.
5. Conclusions
Scheduling problems under complex constraints are usually
challenging to solve. In this paper, we have used an improved
cuckoo search (ICS) algorithm to tackle such ow-job scheduling
problems. In this ICS algorithm, a constructive heuristic called NEH
heuristic is used for initialization so as to obtain the optimal or
near optimal solutions rapidly. The algorithm has been tested with
real-world application data from a furniture manufacturing com-
pany. Moreover, we have conducted the design of experiments with
different parameters to evaluate the performance. Computational
results and detailed comparison studies with GA, ACO, SA, and PSO
show that the ICS algorithm outperforms many other metaheuristic
algorithms and the NEH heuristic in the literature.
Though the proposed algorithm has been tested with random
problem instances, it would be interesting to see the result for
benchmark problems for large-scale HFS scheduling problems. In
the future, we will investigate other constructive heuristics to build
the initial solutions. We will also apply this proposed algorithm
to other HFS problems related with due-date-related performance
measures. Further research should consider to extend to HFS prob-
lems where machines are not identical at each stage. We have made
several assumptions in this paper. Some of these assumptions can
be dropped and we can solve the problems using the ICS algorithm.
That is the proposed algorithm could be applied to solve the HFS
scheduling problems with setup and transportation times. We can
consider the maintenance of machines too. This may be another
possible scope.
Acknowledgements
The authors are very grateful to the anonymous reviewers for
their insightful comments and helpful suggestions.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.asoc.2014.02.005.
References
[1] M. Pinedo, Scheduling Theory, Algorithms and Systems, Prentice Hall, Engle-
wood Clis, NJ, 1995.
[2] T.S. Arthanari, K.G. Ramamurthy, An extension of two machines sequencing
problem, Operations Research 8 (1971) 1022.
[3] C. Kahraman, O. Engin, I. Kaya, R.E. ztrk, Multiprocessor task scheduling in
multistage hybrid ow-shops: a parallel greedy algorithm approach, Applied
Soft Computing Journal 10 (2010) 12931300.
[4] C.J. Liao, E. Tjandradjaja, T.P. Chung, An approach using particle swarm
optimization and bottleneck heuristic to solve hybrid ow shop scheduling
problem, Applied Soft Computing Journal 12 (2012) 17551764.
[5] S.A. Brah, J.L. Hunsucker, Branch and bound algorithm for the ow shop
with multiple processors, European Journal of Operational Research 51 (1991)
8899.
[6] F. Riane, A. Artiba, S.E. Elmaghraby, A Hybrid three stage ow shop problem:
efcient heuristics to minimize makespan, European Journal of Operational
Research 109 (1998) 321329.
[7] O. Moursli, Y. Pochet, A branch and bound algorithm for the hybrid ow shop,
International Journal of Production Economics 64 (2000) 113125.
[8] H. Allaoui, A. Artiba, Scheduling two stage hybrid ow shop with availability
constraints, Computers & Operations Research 33 (2006) 13991419.
[9] M.C. Portmann, A. Vignier, D. Dardilhac, D. Dezalay, Branch and bound crossed
with GA to solve hybrid owshops, European Journal of Operational Research
107 (1998) 389400.
[10] E. Neron, P. Baptise, J.N.D. Gupta, Solving hybrid ow shop problem using ener-
getic reasoning and global operations, Omega 29 (2001) 501511.
[11] J.N.D. Gupta, Two-stage, hybrid owshop scheduling problem, Journal of the
Operational Research Society 39 (1988) 359364.
[12] J.A. Hoogeveen, J.K. Lenstra, B. Vettman, Pre-emptive scheduling in a two-stage
multiprocessor ow shop is NP-hard, European Journal of Operational Research
89 (1996) 172175.
[13] J.N.D. Gupta, E.A. Tunc, Scheduling a two-stage hybrid owshop with separable
setup and removal times, European Journal of Operational Research 77 (1994)
415428.
[14] C.-Y. Lee, G.L. Vairaktarakis, Minimizing makespan in hybrid owshops, Oper-
ational Research Letters 16 (1994) 149158.
[15] J.L. Hunsucker, J.R. Shah, Comparative performance analysis of priority rules
in a constrained ow-shop with multiple processors environment, European
Journal of Operational Research 72 (1994) 102114.
[16] S.A. Brah, L.L. Loo, Heuristics for scheduling in a ow shop with multiple pro-
cessors, European Journal of Operational Research 113 (1999) 113122.
[17] E.G. Negenman, Local search algorithms for multiprocessor ow shop sched-
uling problem, European Journal of Operational Research 128 (2001) 147158.
[18] H. Soewandi, S.E. Elmaghraby, Sequencing three-stage exible owshops with
identical machines to minimize makespan, IIE Transactions 33 (2001) 985993.
[19] C. Oguz, M.F. Ercan, T.C.E. Cheng, Y.F. Fung, Heuristic algorithms for multipro-
cessor task scheduling in a two-stage hybrid ow-shop, European Journal of
Operational Research 149 (2003) 390403.
[20] D. He, A. Babayan, Solving the n-job 3-stage exible ow shop scheduling
problem using an agent-based approach, International Journal of Production
Research 42 (2004) 13991415.
[21] G.J. Kyparisis, C. Koulamas, Flexible ow shop scheduling with uniform parallel
machines, European Journal of Operational Research 168 (2006) 985997.
 M.K. Marichelvam et al. / Applied Soft Computing 19 (2014) 93101
[22] O. Engin, A. Dyen, A new approach to solve hybrid ow shop scheduling prob-
lems by articial immune system, Future Generation Computer Systems 20
(2004) 10831095.
[23] M. Zandieh, S.M.T.F. Ghami, S.M.M. Husseini, An immune algorithm approach to
hybrid ow shops scheduling with sequence-dependent setup times, Applied
Mathematics and Computation 180 (2006) 111127.
[24] K.C. Ying, Minimising makespan for multistage hybrid owshop scheduling
problems with multiprocessor tasks by a hybrid immune algorithm, European
Journal of Industrial Engineering 6 (2012) 199215.
[25] T. Yang, Y. Kuo, I. Chang, Tabu-search simulation optimization approach for
ow-shop scheduling with multiple processors a case study, International
Journal of Production Research 42 (2004) 40154030.
[26] W. Bozejko, J. Pempera, C. Smutnicki, Parallel tabu search algorithm for the
hybrid ow shop problem, Computers & Industrial Engineering 65 (2013)
466474.
[27] S.H.H. Edwin, N. Ansari, H. Ren, A genetic algorithm for multiprocessor sched-
uling, IEEE Transactions on Parallel and Distributed Systems 5 (1994) 113120.
[28] S.F.S. Serifoglu, G. Ulusoy, Multiprocessor task scheduling in multistage hybrid
ow shops: a genetic algorithm approach, Journal of the Operational Research
Society 55 (2004) 504512.
[29] C. Oguz, M.F. Ercan, A genetic algorithm for hybrid ow shop scheduling with
multiprocessor tasks, Journal of Scheduling 8 (2005) 323351.
[30] R. Ruiz, C. Maroto, A genetic algorithm for hybrid owshops with sequence
dependent setup times and machine eligibility, European Journal of Opera-
tional Research 169 (2006) 781800.
[31] E. Rashidi, M. Jahandar, M. Zandieh, An improved hybrid multi-objective par-
allel genetic algorithm for hybrid ow shop scheduling with unrelated parallel
machines, International Journal of Advanced Manufacturing Technology 49
(2010) 11291139.
[32] O. Engin, G. Ceran, M.K. Yilmaz, An efcient genetic algorithm for hybrid ow
shop scheduling with multiprocessor task problems, Applied Soft Computing
11 (2011) 30563065.
[33] K.C. Ying, S.W. Lin, Multiprocessor task scheduling in multistage hybrid ow-
shops: an ant colony system approach, International Journal of Production
Research 44 (2006) 31613177.
[34] K. Alaykyran, O. Engin, A. Dyen, Using ant colony optimization to solve hybrid
ow shop scheduling problems, International Journal of Advanced Manufac-
turing Technology 35 (2007) 541550.
[35] H.M. Wang, F.D. Chou, F.C. Wu, A simulated annealing for hybrid ow shop
scheduling with multiprocessor tasks to minimize makespan, International
Journal of Advanced Manufacturing Technology 53 (2011) 761776.
[36] C.T. Tseng, C.J. Liao, A particle swarm optimization algorithm for hybrid ow-
shop scheduling with multiprocessor tasks, International Journal of Production
Research 46 (2008) 46554670.
[37] R. Linn, W. Zhang, Hybrid ow shop scheduling: a survey, Computers and
Industrial Engineering 37 (1999) 5761.
[38] R. Ruiz, J.A. Vazquez-Rodriguez, The hybrid ow shop scheduling problem,
European Journal of Operational Research 205 (2010) 118.
[39] X.S. Yang, S. Deb, Cuckoo search via Lvy ights, in: NaBIC 2009: Proceedings of
the World Congress on Nature & Biologically Inspired Computing, Coimbatore,
India, 2009, pp. 210214.
[40] X.S. Yang, S. Deb, Engineering optimisation by cuckoo search, International
Journal of Mathematical Modelling and Numerical Optimisation 1 (2010)
330343.
[41] X.S. Yang, S. Deb, Multiobjective cuckoo search for design optimization,
Computers and Operations Research (2011), http://dx.doi.org/10.1016/j.cor.
2011.09.026.
[42] A. Gandomi, X.S. Yang, A. Alavi, Cuckoo search algorithm: a metaheuristic
approach to solve structural optimization problems, Engineering with Com-
puters 29 (1) (2013) 1735.
[43] S. Walton, O. Hassan, K. Morgan, M.R. Brown, Modied cuckoo search: a new
gradient free optimization algorithm, Chaos, Solitons & Fractals 44 (2011)
710718.
[44] A. Layeb, A novel quantum inspired cuckoo search for knapsack problems,
International Journal of Bio-Inspired Computation 3 (2011) 297305.
[45] I. Durgun, A.R. Yildiz, Structural design optimization of vehicle components
using cuckoo search algorithm, Materials Testing 54 (2012) 185188.
101
[46] A.R. Yildiz, Cuckoo search algorithm for the selection of optimal
machining parameters in milling operations, International Jour-
nal of Advanced Manufacturing Technology 64 (14) (2013) 5561,
http://dx.doi.org/10.1007/s00170-012-4013-7.
[47] M.K. Marichelvam, An improved hybrid Cuckoo Search (IHCS) metaheuristics
algorithm for permutation ow shop scheduling problems, International Jour-
nal of Bio-Inspired Computation 4 (2012) 200205.
[48] K. Chandrasekaran, S.P. Simon, Multi-objective scheduling problem: hybrid
approach using fuzzy assisted cuckoo search algorithm, Swarm and Evolution-
ary Computation 5 (2012) 116.
[49] S. Burnwal, S. Deb, Scheduling optimization of exible manufacturing system
using a cuckoo search based Approach, International Journal of Advanced Man-
ufacturing Technology (2012), http://dx.doi.org/10.1007/s00170-012-4061-z.
[50] M. Nawaz, E.E. Enscore Jr., I. Ham, A heuristic algorithm for the m-machine,
n-job owshop sequencing problem, OMEGA, The International Journal of Man-
agement Science 11 (1983) 9195.
[51] C.D. Paternina-Arboleda, J.R. Montoya-Torres, M.J. Acero-Dominguez, M.C.
Herrera-Hernandez, Scheduling jobs on a k-stage exible ow-shop, Annals
of Operations Research 164 (2008) 2940.
[52] X.S. Yang, Review of meta-heuristics and generalised evolutionary walk algo-
rithm, International Journal of Bio-Inspired Computation 3 (2011) 7784.
[53] X.S. Yang, Z.H. Cui, R.B. Xiao, A.H. Gandomi, M. Karamanoglu, Swarm Intel-
ligence and Bio-Inspired Computation: Theory and Applications, Elsevier,
London, 2013.
[54] R. Ruiz, C. Maroto, A comprehensive review and evaluation of permutation
owshop heuristics, European Journal of Operational Research 165 (2005)
479494.
M.K. Marichelvam received the B.E. degree in Mechani-
cal Engineering and M.E. degree in Industrial Engineering
from Madurai Kamaraj University, Madurai, Tamilnadu,
India, in 2000 and 2002, respectively. Currently he is
working as Assistant Professor in the Department of
Mechanical Engineering, Kamaraj College of Engineer-
ing and Technology, Virudhunagar, Tamilnadu, India.
His research interests include manufacturing scheduling,
multi-objective optimization and metaheuristics. He has
published more than ten papers in conferences and
referred international journals.
T. Prabaharan is currently working as Professor in the
Department of Mechanical Engineering, Mepco Schlenk
Engineering College, Sivakasi, Tamilnadu, India. He
received his Ph.D. from Anna University, Chennai, in the
eld of scheduling optimization for automated manu-
facturing environments. He has published 40 technical
papers in the referred journals and conferences. His
research interests are in the areas of scheduling, robotics
and production planning.
Xin-She Yang received his DPhil in applied mathematics
from Oxford University. After working at Cambridge Uni-
versity and then as Senior Research Scientist at National
Physical Laboratory, UK, for a few years, he is now Reader
in Middlesex University, UK, and Adjunct Professor at
Reykjavik University, Iceland. He has published 15 books
and more than 200 research papers. He is also the Editor-
in-Chief of Int. J. Mathematical Modelling and Numerical
Optimization (IJMMNO).