Research On Teaching-learning-based Hybrid Genetic Algorithm In Flexible Job-shop Scheduling Problem

With the increasing competition in the global market and the development of information technology,the environment turns China in the direction of industry 4.0,manufacturing are growing in the meantime.As manufacturing enterprises,whether it can scheduling properly or not will have impact on production indexes,then it will affect the effficiency and effectiveness of companies.Flexible job shop scheduling problem is the extending of classic job shop scheduling problem,and it's also a NP-hard problem.Given the multi-objective of the realistic production,the research has focused on multi-objective flexible job shop scheduling problem.Both sequencing of processes and assignment of machines are the consideration for FJSP,it was solving this problem by evolutionary algorithms and swarm intelligence algorithms at present.Genetic algorithm(GA)and particle swarm optimization algorithm(PSO)are representatives of them.PSO has simple principle,nice commonality,fast computational speed and certain memory,it's applicable to searching in the large-scale space,yet its searching quality is low.Furthermore,a very different in GA is diversified mode of crossover and mutation for chromosome,it can provide solutions with higher quality and more specific at the expense of slower calculation speed,and its parameter can affect the performance easily.It combines GA and PSO on account of their different optimization mechanism and sharing mechanism,GA uses the method of the retrieval strategies of parallel search,and the principle of PSO is serial search.This passage combines two algorithms for the first time,and comes up with a new algorithm named teaching-learning-based hybrid genetic-particle swarm optimization algorithm(TL-HGAPSO)to solve the multi-objective FJSP.The algorithm includes three modules,that is,genetic algorithm module(GA),bi-memory learning module(BL)and particle swarm optimization module(PSO).Firstly,BL module introduces learning mechanism into genetic algorithm so that chromosome has self-learning characteristic.Moreover,this passage designs the initialization algorithm of minimum waiting time for assigning the machines reasonably,then it improves the operation of crossover and mutation in the traditional genetic algorithm,the.algorithm controls the process of mutation by using two thresholds,it can increase the producing speed of good solutions by the mutation mechanism of multi-parents.During the process of evolution,the offspring in GA learn the characters of good chromosome in the BL in order to improve its fitness value,meanwhile elitist selection strategy is used to update the BL module.In addition,it proposed discretization particle swarm optimization algorithm.Besides,the algorithm iterates the genetic population and particle population simultaneously and exchanges the information during the process,thus two populations complement each other with advantages,finally,it improves the efficiency of algorithm.This passage uses thirteen cases,and each of them runs twenty times,then it completes the two hundred and sixty experimental results with other algorithm.The experiment results show that our TL-HGAPSO algorithm is better than other algorithms compared to most problems in the aspects of results and calculating time,it can always find the solution at least as good as other algorithms compared,it proves the rationality and validity of TL-HGAPSO algorithm applying in FJSP problem.