Research On Hybrid Flow-shop Scheduling Methods Based On Firefly Algorithm

Hybrid Flow-shop Scheduling Problem(HFSP)is a synthesis of flow-shop scheduling problem and parallel machine problem,which widely exists in real-world production scenarios,such as paper-making,pharmaceutical and semiconductor manufacturing industries.HFSP is a typical NP-hard problem,thus it is difficult to be solved by traditional mathematical programming methods.This paper focuses on single-objective HFSP,multiobjective energy-aware HFSP,and multi-objective HFSP with dynamic events.Efficient intelligent scheduling methods based on Firefly Algorithm(FA)are proposed to solve these problems.Firstly,FA is proposed to solve the single-objective HFSP.A mathematical model with the optimization objective of makespan is established;In this algorithm,considering the discrete characteristic of the problem,a job-based encoding method and a hybrid decoding methods are employed;a crossover strategy and probability-based multi-neighborhood collaboration search are proposed to update the population;For the best found solution,a neighborhood search based on job adjustments is used.The results on benchmark instances verify the effectiveness and superiority of FA for solving HFSP.Secondly,a multi-objective firefly algorithm(MOFA)is proposed to solve the multiobjective HFSP.A mathematical model with the optimization objectives of makespan and total energy consumption is established;According to the three sub-problems of energy efficiency optimization HFSP,a job-based two-layer encoding method and the selection of machine speed is designed;Based on the theory of multi-objective optimization,the Pareto dominant relationship is employed to obtain the Pareto frontier;For the Pareto solution,the energy-saving strategy is used.The results of comparing with other algorithms on random instances verify the significant advantages of MOFA.Then,an improved MOFA is proposed to solve the multi-objective dynamic HFSP.Considering machine breakdowns,a mathematical model is established with the optimization objectives of makespan,total energy consumption,and the stability of system;MOFA is improved based on the characteristics of the dynamic problem.And fast non-dominated sorting and crowded distance are employed to update populations.The results of comparing with other algorithm on random instances verify that MOFA performs well in all the aspects.Finally,the main work is summarized,and the future research directions are put forward.