Particle swarm optimization with variable neighborhood search for multiobjective flexible job shop scheduling problem

Abstract

The simulation on benchmarks is a very simple and efficient method to evaluate the performance of the algorithm for solving flexible job shop scheduling model. Due to the assignment and scheduling decisions, flexible job shop scheduling problem (FJSP) becomes extremely hard to solve for production management. A discrete multi-objective particle swarm optimization (PSO) and simulated annealing (SA) algorithm with variable neighborhood search is developed for FJSP with three criteria: the makespan, the total workload and the critical machine workload. Firstly, a discrete PSO is designed and then SA algorithm performs variable neighborhood search integrating two neighborhoods on public critical block to enhance the search ability. Finally, the selection strategy of the personal-best individual and global-best individual from the external archive is developed in multi-objective optimization. Through the experimental simulation on matlab, the tests on Kacem instances, Brdata instances and BCdata instances show that the modified discrete multi-objective PSO algorithm is a promising and valid method for optimizing FJSP with three criteria.