Improved Differential Evolution Algorithm And Solving Flexible Job Shop Scheduling Problem

Optimization problems exist widely in the fields of scientific research and engineering and require efficient and general-purpose algorithms to solve them.Heuristic algorithms to obtain feasible solutions to problems by simulating the evolutionary mechanism of survival of the fittest in nature have become a research hotspot at home and abroad.Differential Evolution(DE)is a heuristic algorithm.Because of its simple structure and strong robustness,it is currently widely used to solve various optimization problems.Flexible job shop scheduling problem(FJSP),as a typical optimization problem,can be solved by heuristic algorithm to obtain a scheduling solution with shorter completion time on the premise of limited resources.Focusing on improving DE and solving FJSP,the main research contents of the thesis include:(1)Aiming at the problems of poor local search ability,easy premature convergence and sensitive parameter setting of the differential evolution algorithm,an adaptive differential evolution based on newton cubic interpolation(ANCIDE)was proposed.Newton’s cubic interpolation is used to provide directions for the local search of the algorithm and improve the search speed.An adaptive argumentation strategy is proposed to avoid premature convergence of the algorithm.The control parameters F and CR are self-adjusted using adaptive learning strategies to avoid artificially setting parameters and improve the generality of the algorithm.Finally,28 benchmark test functions on the standard test set CEC2013 are used to compare experiments with 8 algorithms in the literature.The results show that for most benchmark functions,the performance of ANCIDE is better than other comparison algorithms.(2)The ANCIDE algorithm proposed in this thesis is used to solve the flexible job shop scheduling problem.First,a mathematical model is established for FJSP whose optimization goal is makepan.Secondly,the chromosome adopts a double-layer coding method and proposes a vector conversion mechanism to realize the conversion between continuous variables and double discrete vectors.A decoding algorithm is proposed to decode the scheduling scheme into an active scheduling scheme.Finally,the three FJSP standard test sets are compared with five algorithms in the literature.The results show that for most scheduling examples,the stability of ANCIDE and the ability to obtain the optimal scheduling scheme are better than other comparison algorithms.(3)Simulation verification in complex engineering environment.Further considering the three complex constraints in the actual production process:buffer,set time and recovery time,a complex scheduling case of 10 "×" 10 is proposed.With Plant Simulation simulation software,a simulation model is established and simulation experiments are performed.The results show that when solving FJSP in a complex engineering environment,compared with DE,ANCIDE can shorten the completion time and increase the average machine utilization.