Identifying Significant Control Factors of Particle Swarm Optimization Algorithms in Solving Permutation Flowshop Scheduling Problems

Abstract

This paper aims to identify significant control factors of particle swarm optimization (PSO) algorithms in solving permutation flowshop scheduling problems. Control factors of PSO algorithms considered herein include inertial weight, acceleration coefficients, breeding operation, and the amount of particles. The full factorial design method is applied to plan a set of experiments. Each experiment, denoting a specific version of PSO algorithm, is used to solve the test problems, Carlier problems. The searching ability of PSO algorithms is defined by the ratio of the number of times that the optimal makespan is searched to the total number of searching times. To identify significant factors, the analysis of variance (ANOVA) method is used to analyze the results of experiments. According to the results of ANOVA, adopting time-varying acceleration coefficients, breeding operation, and a low amount of particles can advance significantly the searching ability of PSO algorithms. Adopting a high amount of particles can increase significantly the robustness of PSO algorithms. Any two-factor interaction is not significant. Inertia weight is not a significant factor, so any effort to modify inertia weight is unnecessary.