Research On Dynamic Scheduling Problem Of Flexible Job-shop With Hybrid Genetic Algorithm

The concept of "Industry 4.0" and "Made in China 2025" will lead the manufacturing mode to "Multi-products,Small-batches,Custom-made" in the future.There is no doubt that the discrete production model will affect the production scheduling of manufacturing enterprises.The production manufacturing system in practical is partially flexible,the number of working processes depends on the job,and each working process can be processed by different machines.In addition,the production manufacturing system is an open space,the original scheduling scheme may be influenced by all sorts of things that occurs suddenly,and these interferences may make it infeasible.In order to improve the production efficiency of manufacturing enterprises,how to solve the Dynamic Scheduling Problem of Flexible Job-shop become a burning issue.Firstly,this paper analyzes the dynamic scheduling problem of flexible job-shop.Based on the characteristics of the problem,it builds a model and makes some assumptions which more closer to the dynamic environment of the real production.Secondly,to solve the dynamic scheduling problem of flexible job-shop,this paper designs an operation framework.The framework includes two elements:dynamic scheduling strategy and optimization algorithm.A cycle and event driven strategy with rolling window technology is selected as the dynamic scheduling strategy.A hybrid genetic algorithm based on machine learning is proposed in this paper.Then the framework is applied to solve some problems.In particular,the hybrid algorithm has been improved in the following aspects:For the sake of the quality of the initial population,the machine learning theory is used to create a part of initial population;The knowledge base is built as the strategy for retaining good populations,to make up for the deficiency of traditional genetic algorithm,which is easy to lose the optimal solution;Applying classification algorithms of machine learning to choice operator,it helps to keep all kinds of excellent individuals to next generation proportionately;To establish competition mechanism,the Metropolis criterion from the simulated annealing algorithm is combined with crossover operators and mutation operators,it can not only keep the diversity of population but also accelerate the speed of convergence;The mean of the environment is introduced to mutation rate,which decay with the convergence of the algorithm gradually,make it possible to converge to global optimum more quickly.Moreover,in order to verify the practicability of the hybrid algorithm and to satisfy the general requirements of production workshop,a website is implemented,which called flexible job-shop scheduling platform.In this paper,development process of this platform is described in detail.The simulation results show that the hybrid genetic algorithm has good performance on solving flexible job-shop problem and dynamic scheduling problem of flexible job-shop.The prototype system also confirmed that the hybrid genetic algorithm are able to suggest a better solution under the three types of dynamic events in the actual production effectively.