A Hybrid Evolutionary Algorithm For Solving Many-objective Flexible Job Shop Scheduling Problem

Many-objective flexible job-shop scheduling problem(MaOFJSP)desires to optimize four or more conflicting objectives simultaneously in real-world industries,and its study can make it possible to provide practical guidance for the real-world industries.For the difficulty in solving MaOFJSP is greatly increased with the increase in number of objectives,it is of great academic significance to improve the search efficiency and optimization ability of the algorithm.In this context,this thesis is aimed to investigate the MaOFJSP considering a number of economic indicators and green indicators comprehensively.First,the many-objective flexible job shop scheduling model is established by adding the energy consumption objective to the existing multi-objective flexible job shop scheduling model.The energy consumption indicator is designed based on the analyzing of the power consumption during the machining process,and then the many-objective flexible job shop scheduling model is constructed to minimize the maximum completion time,bottleneck machine load,total machine load and total power consumption.Second,to address traditional multi-objective optimization algorithm’s lacking selecting pressure in the high-dimensional space,a hybrid evolutionary algorithm based on NSGA-III is proposed to solve the MaOFJSP.The classic NSGA-III is adapted to make it more suitable for solving the MaOFJSP.The improved algorithm mainly includes the following key operations:(1)The objectives space of MaOFJSP is decomposed into multiple sub-objectives spaces by using uniformly distributed reference points,and then the algorithm is guided to search in the multiple sub-objectives spaces parallelly.(2)The Niche-Preservation Operation based on reference points is used to analyze the distribution of individuals in the population more systematically,and then it guides the algorithm to select the individuals near the reference points,which can improve the global search ability and maintain population diversity effectively.(3)The external elite archive based on reference points is introduced to preserve the excellent individuals that may be deleted.Parent populations are selected with equal probability from both the current population and the elite archive.The introduction of elite individuals can effectively guide the search to the optimal solution and maintain the diversity of the population.In addition,an improved hybrid population initialization mechanism is adopted to generate a high-quality initial population through the random combination of multiple scheduling rules,improving the speed of searching for scheduling solutions.Finally,the hybrid evolutionary algorithm based on NSGA-III is implemented by using MATLAB programming language,and the performances of the proposed algorithm are compared with other two algorithms.Experiments are carried out on 13 sets of extended benchmark instances including machine processing power data and no-load power data.Each instance runs 10 times,generating 390 sets of experimental results.The results are compared with NSGA-III and NSGA-II algorithm regarding to the quality metric,distribution metric,diversity metric and the number of solutions.The experimental results show that the hybrid evolutionary algorithm based on NSGA-III can obtain better Pareto solution than NSGA-III and NSGA-II algorithm.It verifies the feasibility and effectiveness of the proposed algorithm for solving MaOFJSP.