TECHNICAL TRANSACTIONS CZASOPISMO TECHNICZNE

MECHANICS MECHANIKA

1-M/2013

PAWEŁ WOJAKOWSKI*, DOROTA WARŻOŁEK*

RESEARCH STUDY OF STATE-OF-THE-ART ALGORITHMS

FOR FLEXIBLE JOB-SHOP SCHEDULING PROBLEM

PRZEGLĄD WSPÓŁCZESNYCH ALGORYTMÓW

HARMONOGRAMOWANIA ZADAŃ Z MASZYNAMI
ALTERNATYWNYMI

Abstract
The paper discusses various approaches used to solve flexible job-shop scheduling problem
concentrating on formulations proposed in the last ten years. It mainly refers to the applied
metaheuristic techniques which have been exploited in this research area. A comparison of
presented approaches is attempted, some concluding insights are highlighted. Finally future
research directions are suggested.
Keywords
: flexible job-shop scheduling
, multi-objective optimization
, review
Streszczenie
W artykule opisano różne podejścia stosowane do rozwiązania problemu harmonogramowania
zadań z maszynami alternatywnymi. Skoncentrowano się na opracowaniach opublikowanych
w ostatnich dziesięciu latach. Głównie skupiono uwagę na podejściach wykorzystujących al-
gorytmy metaheurystyczne. Dokonano próby porównania merytorycznego dostępnych w lite-
raturze rozwiązań oraz wskazano kierunki dalszych prac.
Słowakluczowe
: harmonogramowanie zadań z maszynami alternatywnymi
, optymalizacja
wielokryterialna
, przegląd literatury

* PhD. Eng. Paweł Wojakowski, MSc. Eng. Dorota Warżołek, Institute of Production Engineering,
Faculty of Mechanical Engineering, Cracow University of Technology.
382

1. Introduction

One of the most popular research problem that is exploited in recent years by international
research centres is job-shop scheduling problem. Because of its combinatorial complex nature
of sequencing operations on available machines, it is quite difficult to obtain optimal solution
directed to specific criterion. Therefore, many evolutionary approaches emerged in order to
cope with this problem. These approaches provide sub-optimal satisfactory solutions with the
emphasis on, in the most cases, one chosen criterion achieved in deterministic conditions [1].
It is worth noticing that optimization process runs with considering many constraints. One of
them is that an operation can be processed only on one machine.
As an extension of this problem, flexible job-shop scheduling (FJS) problem has
increasingly focused research attention over last ten years. This is because of eliminating
explicitly onerous constraint which does not permit doing operations on alternative
machines. However, by the evolution to FJS problem, the level of computational complexity
has been deepened significantly. The main reason is concerned to necessity of combining
several optimization criteria such as minimizing the overall completion time (makespan),
total workload of machines and the workload of the most loaded machine [2]. The need for
developing new approaches in the domain of multi-objective optimization FJS problem has
been intensified.
This paper is prepared for the investigation how modern research solve FJS problem.
It is organized as follows. In section 2 the definition of flexibility of manufacturing systems
is described. In section 3 problem definition and formulation is discussed. Section 4 presents
solution methodologies for FJS problem and literature review based on the articles which
were published over last ten years. Finally, summary and directions for future work is covered
in section 5.

2. The definition of flexibility of manufacturing systems

The flexibility of manufacturing system can be defined as the ability of this system to
adapt quickly to both the changeable demands of a market (demand uncertainty) or resources
failures on the shopfloor (production uncertainty). This changeable conditions can be
expressed by disturbances which result in lower efficiency of the manufacturing system.
To cope with disturbances, management should identify and decide on the type and scope
of flexibility corresponding to the manufacturing system. This can gain benefits in terms of
increasing the value of total efficiency factor [3].
From the point of view of manufacturing systems, there is many kinds of flexibility
which can be considered. Two of them are crucial, namely, a) the flexibility expressed by the
possibility of machines rearrangement to produce part families according to group technology
concept [4], b) the flexibility expressed by the possibility of machines to produce different
parts without necessity to loss much time for reconfiguring machines [5]. Regarding the FJS
problem, the second one is more interesting one.
In the study of ref. [6], two types of the second kind of flexibility have been investigated.
Machine flexibility and routing flexibility have been selected and compared. According to the
description from ref [6], machine flexibility is the capability of a machine to perform different
 383

operations required by a given set of part types and includes quick machine setup and jig
changing. This type of flexibility can be achieved by the use of high-tech automatic tool-
-changing or jig-changing devices in conjunction with sufficient tools and fixtures magazines
as well as reconfigurable machines which allow to replace whole machine modules to perform
other operations [7]. Machine flexibility allows smaller batch sizes, causing shorter lead
times, higher machines utilization and reduced work-in-progress inventory level. Routing
flexibility, in turn, is the capability of processing a given set of part types using more than
one route (alternative routings) [8]. Routing flexibility assumes existence of alternative paths
which can be followed through the manufacturing system for a given process plan [9].
Routing flexibility leads directly to the problem of FJS. This problem concerns two major
steps: a) assignment of each operation to one of the alternative machines (as an assignment
sub-problem), and b) sequencing the set of operations on each machine which has been
previously assigned to perform these operations (as sequencing sub-problem) [10]. Routing
flexibility can be improved by having identical machines or multipurpose machines. It can
help to handle unplanned events such as machine breakdowns or rush orders [6].
FJS problem is strongly NP-hard. That’s why it focus great attention by very large number
of researchers. This paper describes only a handful of available research approaches which
are available in literature. This approaches are mainly directed towards FJS problem with
routing flexibility. The investigation has been conducted to find out the optimization methods
used for solving multiobjective FJS problem, what is presented in the section 4.

3. Problem definition and formulation

FJS problem is an extension of the classical job-shop scheduling (JS) problem. In turn,
job-shop scheduling is, next to single-machine scheduling, flow-shop scheduling and open-
shop scheduling, one of the four basic problems, which have been classified as the challenging
scheduling problems [11].
The classical JS problem considers N jobs to be processed on M machines assuming
that they have pre-determined sequences of operations and each operation is performed on
a predefined machine. It means that for each job, distinct routing is fixed and known in
advance. In general, this problem is to determine optimal schedule of jobs so that one or more
performance criteria could be achieved.
FJS problem is associated with two difficulties. The first one is to assign all operations
to relevant machines (selected from Mk ∈ Mi,j). The second one is to calculate of the starting
times of operations in order to determine appropriate (optimal) sequence of their execution
on each machine so that technological constraints were not violated and predefined objectives
could be obtained [12].
FJS problem belongs to class of NP-hard problems just as mentioned JS problem. This
means that along with the growth of problem size, the number of calculations which must
be done increase in an exponential manner, where N denotes problem size. It also is worth
noticing that it has more complex nature than JS problem because of enlarged searching
scope of potential solutions through reduce machine constraints. Bruker nad Schlie (1990)
were among the first who took up solution of this problem [13].
384

FJS problem can be defined as follows [14]:

– there are N jobs independent from other jobs, indexed by i, i = 1, …, N, where N denotes
total number of jobs;
– each job consists of a sequence of operations, denotes by J i (precedence constraints);
– there is the set of machines, indexed by M, M = M1, …, Mk, where Mk means total number
machines in the set and k denotes k-th machine;
– each operation is indexed by Oi,j, j =1, …, J i, where Oi,j and J i denote that J i-th operation
of job i and number of operations required for job i, respectively;
– each operation Oi,j can be processed on only one machine out of a set of given machines
which are able to perform it. The set of eligible machines is denoted by Mi,j and Mi,j ⊆ M
(routing constraints);
– there is a predefined set of processing times. For a given operation Oi,j and a given machine,
the processing time is denoted by ti,j,k;
moreover [14]:
– if each operation Oi,j can be assigned to each machine from the set of available machines
M, that is Mi,j =M, then it is called total flexibility FJS problem (T-FJS problem);
– if certain operations can be executed only by some machines from the set of available
machines, that is Mi,j ⊂ M, then it is called partial flexibility FJS problem (P-FJS
problem);
Additionally, the following assumptions are made during the process of solving this
problem[14]:
– none of operation Oi,j cannot be interrupted during its processing (non-preemption
condition);
– each machine Mk can perform no more than one operation at any time (resource
constraints);
– each machine is available to other operations immediately after operation which is
assigned as the last to be completed;
– all machines are available at t = 0;
– all jobs can be started at t = 0;
– machines are independent from each other;
– there are no precedence constraints among operations of different jobs;
– setting up times of machines and transportation times between work stations (operations)
are neglected;
– breakdowns are not considers;
– neither release times nor due dates are specified.
In the literature, the following performance criteria for FJS problem are often to be
minimized [11]:
– maximal completion time, i.e. makespan Cmax;
– weight or mean completion time;
– maximal machine workload, i.e. sum of processing times of operations on a critical
machine;
– total workload of machines, i.e. sum of working times over all machines;
– maximal tardiness;
– maximal lateness;
 385

– maximum earliness;
– (weight) number of tardy jobs;
– maximal cost.
The most popular optimization criterion in the case of both single objective and multi-
objective optimization of FJS problem is to minimize the makespan.

4. Solution methodologies for FJS problem

Due to the high complexity of FJS problem, we distinguish two classes of optimization
approaches used to solve this problem. The first one is mathematical modeling while the
second one is metaheuristic approach.
Mathematical modeling allows to obtain optimal solution for small size problems, in turn
metaheuristic approach is used to solve medium and large size problems. Metaheuristic
approaches allow to achieve near-optimal solution [12].
Taking mathematical models into considerations, there are three distinct ways of
formulating the sequencing problem using integer programming (IP). These three approaches
differ from each other the type of binary variable what stores information about sequence
of operations on individual machines. Other existing formulations are based on these three
groups of variables.
The first approach is based on the sequence-position variables. It was proposed by Wagner
in 1959. The second one relies on precedence variables, introduced by Manne in 1960. The
latter way is based on time-indexed variables, proposed by Bowman in 1959 [15].
The second class of optimization approaches directed towards solving FJS problem is
involved with the use of metaheuristics. Over the past decade, metaheuristics have been
intensively exploited for the combinational optimization of FJS problem. The FJS problem
is considered to be more complex and difficult to obtain optimized solution than JS problem.
Thus, many researchers willingly strive for metaheuristics and try to combine them applying
hybridization in the way of achieving optimal solution. Five of the most notable groups
of metaheuristics approaches using to solve FJS problem are: simulated annealing, tabu
search, evolutionary algorithms, ant colony optimization and particle swarm optimization
as well.
Simulated annealing (SA) and tabu search (TS) metaheuristics have a common
characteristics as a search process starts from one initial state which is initial solution and
follows through specific trajectory of solutions in order to find optimum one according to
neighbor searching. One of the oldest metaheuristics is SA. In the case of FJS problem it is
usually used to schedule operations on each machine after the process of assigning operations
on machines [16]. That approach has been presented by Xia and Wu [17]. Fattahi et al. in turn
proposed simulated annealing approach to solve FJS problem in the case if customer demand
can be released more than one for each job as an important and practical issue of FJS problem
[18]. Further, Dalfard and Mohammadi developed simulated annealing approach to solve
FJS problem with parallel machines and with regard of maintenance cost [19].
TS algorithm has been used as a first metaheuristic to solve FJS problem [20]. Based on
this approach more effective tabu search algorithm with advanced variable neighborhood
386

search has been developed [21]. It can be noticed, that after publication of ref. [20], tabu
search has become frequently used metaheuristics to solve FJS problem. Henceforth, it has
been frequently combined with other metaheuristics thereby forming hybrid approaches. The
popular way is to combine tabu search with genetic algorithm, where tabu search is used to
generate initial solution [22]. In other way, it can be used to solve more developed models
like this with transportation constraints and bounded processing times [23].
The most exploited metaheuristics in terms of FJS problem is definitely genetic algorithms.
It was first employed by well-known study of Kacem [24]. Pezzella et al. also solved FJS
problem by genetic algorithm [25]. Further, Bagheri et al. followed by Pezzella et al. and
proposed artificial immune algorithm to solve FJS problem [26].
The last group of approaches employspopulation- based methods, which are (except
of mentioned genetic algorithms) particle swarm optimization (PSO) and ant colony
optimization (ACO). As an example, Xing et al. proposed ACO algorithm to solve FJS
problem [27], whereas Moslehi used particle swarm optimization for the same target [28].

5. Conclusions and future work directions

In the paper, FJS problem is considered and possible methods to solve it are presented.
In addition, it is worth noting that despite the strong tendency to solve much more complex
and difficult variants of JS problem, so far no algorithm has been developed what gives
an optimal solution for the classical JS problem regardless of its size [29].
Since FJS problem is NP – hard problem, researches around the world focus their efforts
on the developing effective metaheuristics that will find a good solution for a given optimal
problem in acceptable time. In other words, metaheuristics attempt to achieve trade-off
between solution quality and search completeness within reasonable a time interval.
The article briefly discussed the five most commonly used group of algorithms for multi-
-objective optimization FJS problem and possible their hybrids. Generally speaking, the
hybrid algorithms as multi-objective optimization methods used to solve FJS problem are
becoming more and more popular and it can be suggested as one of directions of further
research.
Furthermore, it should be taken into account that in multi-objective problems with
conflicting objective functions, existing only one (optimal) solution by optimizing all
objective functions is almost impossible . Hence, in recent years, more studies have also
focused on Pareto – based approaches what provide a set of optimal solutions, instead
of a single optimal solution.

References

[1] Lazar I., Review on solving the job-shop scheduling problem: recent development ,
Transfer Inovacii, 23/2012, 55-60.
[2] Hsu T., Dupas R., Jolly D., Goncalves G.,Evaluation of mutation heuristics for the solving
of multiobjective flexible job-shop by an evolutionary , 2002 algorithm
IEEE International
Conference on Systems, Man and Cybernetics, Vol. 5, 2002.
 387

[3] Das S.K., Nagendra P., Investigations into the impactof flexibilityon manufacturing
performance , International Journal of Production Research, Vol. 31, No. 10, 1993, 2337-2354.
[4] Nomden G., v.d.Zee D.J., Virtual cellular manufacturing: Configuring routing , flexibility
International Journal of Production Economics, Vol. 112, 2008, 439-451.
[5] Pattanaik L.N., Jain P.K., Mehta N.K., Cell formation in the presence of reconfigurable
machines , International Journal of Advanced Manufacturing Technology, Vol. 34, 2007, 335-345.
[6] Tsubone H., Horikawa M.,A comparison between machine flexibility and routing , flexib
International Journal of Flexible Manufacturing Systems, Vol. 11, 1999, 83-101.
[7] Landers R.G., Min B.K., Koren Y.,Reconfigurable machine toolsAnnals – Manufacturing
, CIRP
Technology, Vol. 50, 2001, 269-274.
[8] Özgüven C., Özbakir L., Yavuz Y.,Mathematical models for job-shop scheduling problem
with routing and process plan , flexibility Applied Mathematical Modelling, Vol. 34, 2010,
1539-1548.
[9] Stecke K.E., Raman N., FMS planning decision, operating flexibilities and system , performan
IEEE Transactions on Engineering Management, Vol. 42, 1995, 82-90.
[10] Rossi A., Dini G., Flexible job-shop scheduling with routing flexibility and separable
times using ant colony optimization method
, Robotics and Computer-Integrated Manufacturing,
Vol. 23, 2007, 503-516.
[11] Wang X.J., Zhang C.Y., Gao L., Li P.G., A survey and future trend of study on multi-objective
scheduling , ICNC IEEE International Conference on Natural Computation, 2008, 382-391.
[12] Fattahi P., Mehrabad M S.,Jolai F., Mathematical modeling and heuristic approaches to flexibl
job shop scheduling problem , Journal of Intelligent Manufacturing, Vol. 18, 2007, 331-342.
[13] Bruker P., Schile R.,Job shop scheduling with multi-purpose machineVol. 45,
, Computing,
1990, 369-375.
[14] Motaghedi-Iarijani A., Sabri-Iaghaie K. & Heydari M., Solving flexible job shop scheduling with
multi objective approach , International Journal of Industry Engineering & Production Research,
ISSN: 2008-4889, Vol. 21, 2010, 197-209.
[15] Demir Y., Kürnat İgleyen S.,Evaluation of mathematical models for flexible job-shop schedu
problems , Applied Mathematical Modelling, Vol. 21, 2010, 197-209.
[16] Loukil T., Teghem J., Fortemps P.,A multi-objective production scheduling case study solved
simulated annealing , European Journal of Operational Research, Vol. 179, 2007, 709-722.
[17] Xia W., Wu Z., An effective hybrid optimization approach for multi-objective flexible job-s
scheduling problems , Computers & Industrial Engineering, Vol. 48, 2005, 409-425.
[18] Fattahi P., Jolai F., Arkat J.,Flexible job shop scheduling with overlapping, in operations
Applied
Mathematical Modelling, Vol. 33, 2009, 3076-3087.
[19] Dalfard V.M., Mohammadi G., Two meta-heuristic algorithms for solving multi-objective
flexible job-shop scheduling with parallel machine and maintenance constraints
, Computers and
Mathematics with Applications, Vol. 64, 2012, 2111-2117.
[20] Brandimarte P., Routing and scheduling in a flexible job-shop by tabuof search
, Annals
Operations Research, Vol. 41, 1993, 157-183.
[21] Li J.Q., Pan Q., Liang Y.C.,An effective hybrid tabu search algorithm for multi-objective flex
job-shop scheduling problems , Computers & Industrial Engineering, Vol. 59, 2010, 647-662.
[22] Vilcot G., Billaut J.C., A tabu search and a genetic algorithm for solving a bicriteria gene
job shop scheduling problem , European Journal of Operational Research, Vol. 190, 2008,
398-411.
[23] Zhang Q., Manier H., Manier M.A., A genetic algorithm with tabu search procedure for flexib
job-shop scheduling with transportation constraints and bounded, Computers processing times
& Operations Research, Vol. 39, 2012, 1713-1723.
388

[24] Kacem I., Hammadi S., Borne P., Pareto-optimality approach for flexible job-shop scheduling
problem: hybridization of evolutionary algorithms and fuzzy logic
, Mathematics and Computers
in Simulation, Vol. 60, 2002, 245-276.
[25] Pezzella F., Morganti G., Ciaschetti G., A genetic algorithm for the flexible job shop schedulin
problem , Computers & Operations Research, Vol. 35, 2008, 3202-3212.
[26] Bagheri A., Zandieh M., Mahdavi I., Yazdani M.,An artificial immune algorithm for the flexible
job-shop scheduling problem , Future Generation Computer Systems, Vol. 26, 2010, 533-541.
[27] Xing L.N., Chen Y.W., Wang P., Zhao S., Xiong J.,A knowledge-based ant colony optimization
for flexible job-shop scheduling ,problems Applied Soft Computing, Vol. 10, 2010, 888-896.
[28] Moslehi G., Mahnam M., A Pareto approach to multi-objective flexible job-shop sched
problem using particle swarm optimization and local search
, International Journal of Production
Economics, Vol. 129, 2011, 14-22.
[29] Baykasoglu A., Ozbakir L., Sonmez A.I., Using multiple objective tabu search and grammars
to model and solve multi-objective flexible job-shop scheduling , Journal ofproblems
Intelligent
Manufacturing, Vol. 15(6), 2004, 777-785.