Integrating sequence-dependent group scheduling problem and preventive maintenance in flexible flow shops

• Published in: International journal of advanced manufacturing technology 2015-03, Vol.77 (1-4), p.173-185
• Main Authors: Khamseh, Alireza, Jolai, Fariborz, Babaei, Morteza
• Format: Article
• Language: English
• Subjects: Algorithms; CAE) and Design; Central processing units; Completion time; Computer simulation; Computer-Aided Engineering (CAD; CPUs; Design parameters; Engineering; Genetic algorithms; Industrial and Production Engineering; Mechanical Engineering; Media Management; Order parameters; Original Article; Parameter robustness; Preventive maintenance; Production scheduling; Representations; Scheduling; Sequential scheduling; Simulated annealing; Taguchi methods
• ISSN: 0268-3768; 1433-3015
• DOI: 10.1007/s00170-014-6429-8

This study integrates flexible flow shop group scheduling problem with sequence-dependent setups and preventive maintenance activities in order to minimize the total completion time (makespan). In a group scheduling problem, scheduling of groups and the jobs within each group are determined. As the considered problem is strongly NP-hard, we propose two meta-heuristics based on simulated annealing (SA) and genetic algorithm (GA) to solve it. Matrix solution representation is a key feature of GA-based algorithm that makes possible the representation of groups and the jobs within groups simultaneously. In addition, the SA-based algorithm is equipped with a local search procedure to enhance the quality of its solution. In order to set parameters and better achieve the performances of the algorithms, we exploit Taguchi robust parameter design method. The performance of the proposed algorithms is evaluated on a variety of test problems, namely small- and large-sized problems. Makespan and elapsed central processing unit (CPU, or processing) time to obtain it are considered as two response variables representing effectiveness and efficiency of the algorithms, respectively. The obtained results show that there is statistically significant difference between performances of the proposed algorithms. GA-based algorithm shows better performance on two response variables for both problem sizes (except regarding elapsed CPU time to obtain the best solution for large-sized problems) with a p  value of 0 and outperforms SA-based algorithm.