Multi-objective Flexible Job Shop Scheduling Transportation Problem: Simulation Approach
Résumé
Influenced by a capitalized concept, “maximum utility with minimum cost”, more and more
industries in manufacturing area have focused on flexibility and adopted themselves to flexible job
shop environments. In traditional Job Shop Scheduling Problem (JSP), there are n jobs and m
machines in manufacturing system, where each job consists of j operations that need to be executed
on machines by a given order. [1]. It extends the assumption that only one machine is able to run a
particular operation. Since JSP can be considered as a special case of FJSP and JSP is well-known
NP-hard, FJSP is also regarded as an NP-hard problem.
Considering its computational complexity, various metaheuristic have been extensively applied
to FJSP. Zendieh et al. [2] proposed a Genetic Algorithm for FJSP by using several different
rules for generating the initial population and several strategies for producing new population for
next generation. Gao et al. [3] introduced a Pareto-based grouping discrete harmony search
algorithm (PGDHS) to solve FJSP. Chamber et al. [4] extended their Tabu Search strategy
previously described for job shops and applied it to FJSP. Besides metaheuristic algorithm,
researchers also employ traditional polynomial algorithms and hybrid algorithms, like GA combined
with a variable neighborhood descent [5], in FJSP.
The objective of Flexible Job Shop Scheduling Problem is to determine a feasible schedule S by
minimizing a given objective function [6]. In early work, the wide-used objective is from a single
dimension, considering only one objective once. Recently, researchers have addressed their study on
multiple objectives. Karthikeyan et al. [7] introduced a hybrid discrete firefly algorithm
(HDFA) in which the objectives include the minimization of makespan, maximal workload and total
workload of machines. Gao et al. [5] proposed a Discrete Harmony Search algorithm for FJS with
weighted combination of makespan, the mean of earliness and tardiness criteria.
In this paper, we propose a multi-objective dynamic scheduling algorithm based on entropy to
optimize the transportation task in a flexible job shop manufacturing environment. The system is
composed of machines, input/output buffers, transporters and traceable products, where the machines
are not identical and have ability to execute more than one operation. A simulation based on multiagent approach is used in order to evaluate proposed solutions regarding multiple objectives, such as
makespan, routing flexibility. For programming, we use Netlogo 6.0.1 to code the simulation model
with real production data from a Normandy company in France. Netlogo is a multi-agent
programmable modeling environment, which is able to present the layout of system and all results
directly to industrial users. In previous study, part of research study have used Matlab (Karthikeyan
et al. [7]) or C ++ (Zadieh et al. [2]) to test their algorithms in case environment. However, few of
them derives their algorithms from simulation of cases. This study is up to fill this gap and aims to
offer a more intuitive way for industrial users who may not have programming background [8].
Origine : Fichiers produits par l'(les) auteur(s)