Research On Multi-objective Flexible Job Shop Scheduling Based On Evolutionary Calculation

With the globalization of the market,enterprises are facing increasingly fierce competition.In order to meet the individual needs of customers,there is an urgent need for production scheduling systems that can quickly and reliably achieve customized production of small batches,short cycles and high quality.Higher requirements are placed on the flexibility of the manufacturing system.How to provide efficient scheduling schemes for increasingly complex flexible manufacturing systems(FMS)is the number one problem in modern production management,which requires in-depth research on flexible job shop scheduling(FJSP).The multi-objective flexible job shop scheduling problem(MO-FJSP)comprehensively considers the characteristics of flexibility in the production process and the diversification of management decision-making demands.It has more general research value and the research progress is relatively lagging,so this thesis regards it as the main research topic.Evolutionary algorithms are currently the most commonly used and effective methods to treat combinatorial optimization problems,with the advantages of strong versatility and high robustness.In this paper,we choose two most representative evolutionary algorithms to deal with MO-FJSP.The innovation of this article is mainly reflected in the improvement and specific design of the algorithm.This article studies MO-FJSP on the basis of existing research results.The main results are as follows:Analyze the basic concepts of MO-FJSP,clarify the internal relationship among the workpiece,operation and machine,then establish the mathematical model of MO-FJSP after making the necessary assumptions;The traditional multi-objective evolutionary algorithms have difficulties in providing enough selecting pressure when solve MO-FJSP.The latest third-generation non-dominated sorting genetic algorithm(NSGA-III)uses a reference point-based environment selection mechanism,which can effectively overcome this problem.In this paper,NSGA-III is used as the main method,and the specific encoding and decoding schemes and evolution operators are designed according to the characteristics of the scheduling solution,and heuristic rules are used to initialize the population.Finally,it is tested on internationally accepted benchmarks to prove the efficiency of NSGA-III in dealing with MO-FJSP through comparison with several advanced algorithms;The cuckoo search(CS)has excellent global search capabilities,but is limited to solving single-objective problems.Considering that NSGA-III is very suitable for multi-objective optimization problems,this article improves standard CS with reference to NSGA-III.In addition,a self-learning and self-adapting neighborhood search algorithm is put forward to enhance its local search capabilities.In the specific implementation,because the positionupdating formula of the cuckoo algorithm is not suitable for directly operating on integer coding,this paper solves the discretization of cuckoo algorithm by seeking and establishing the mappings between integer coding and real coding.Finally,the superiority of the improved algorithm compared to NSGA-III is verified by simulation of benchmarks.