Research On Multiobjective Evolutionary Algorithms And Their Applications In Production Scheduling Problems

Multiobjective optimization is a hot topic in academic and industrial areas.A lot of optimization problems,both in scientific studies and engineering applications,need to be considered multiple criterion(objectives),and all the objectives are in conflict.Difference from the single objective optimization problem,there is no single optimal solution that can make all the objectives up to the optimum.For the multiobjective optimization problem,there are many,even infinite optimal solutions,named Pareto optimal solution set.Therefore,it is more difficult for solving the multiobjective optimization problem than that of single optimization problem.Evolutionary algorithm,an intelligent optimization algorithm based on population,has shown good performance in solving the multiobjective optimization problems.It uses one population to search the whole solution space,and can obtain multiple nondominated solutions in a single run,which is very suitable for solving the multiobjective optimization problem.Recently,it has attracted widespread attention from academic and industrial fields.The aim of multiobjective optimization algorithm is to achieve one nondominated solutions that are close to the Pareto front and uniformly distributed on the Pareto front as well as possible.This paper considers the algorithm research and application research on multiobjective evolutionary algorithm.Firstly,an adaptive multiple population strategy is proposed as the core idea of the new type of multiobjective evolutionary algorithm.Secondly,based on the achievement of the algorithm research,the multiobjective scheduling is considered in the application research.The main research work can be summarized as follows:(1)The multiobjective genetic algorithm based on adaptive multipopulation strategy is proposed in order to achieve a set of nondominated solution quickly and accurately.Based on the distribution of population in the objective space and decision space at each iteration,the multiple subpopulations are constructed to solve all subproblems simultaneously.In order to examine the performance of the proposed algorithm,the classical multiobjective optimization algorithm are chosen as the compared algorithms,and the simulation experimental results show that the proposed algorithm can achieve a set of nondominated solutions more quickly and accurately.(2)In order to get a nondominated solution set with higher quality and better distribution,an improved adaptive multipopulation strategy is designed,and the hybrid multiobjective optimization algorithm based on multiple optimization strategy integration is proposed to solve the continuous multiobjective optimization problem.In order to examine the performance of the proposed algorithm,the classical and the similar multiobjective evolutionary algorithms are chosen as the compared algorithm,the simulation experiments on some benchmark test function show that the proposed algorithm can get one nondominated solution set with higher quality and better distribution.(3)The single machine scheduling problem with deterioration effect is proposed,and the 0-1 mathematical model with the objective of minimizing the makespan and the total tardiness is built.The hybrid multiobjective evolutionary optimization algorithm based on multipopulation strategy is designed to solve this problem.The local search method is proposed based on the characteristic of this problem.In order to examine the performance of the proposed algorithm,the simulation experiments on the stochastic instances are carried out.The experimental results show that the proposed algorithm has better performance in solving this problem.(4)The general flow shop scheduling problem is considered,and the scheduling model is built to minimize the makespan and the total tardiness simultaneously.The multiobjective genetic algorithm based on multipopulation strategy is designed to solve it.In order to examine the performance of the proposed algorithm,the standard method is used to generate a set of instances,and the algorithms which solve the similar problems are chosen as the compared algorithms.The experimental results show the proposed algorithm can get the nondominated solution set with higher quality and better distribution.(5)The multiobjective flow shop scheduling problem with deterioration effect is proposed,and the multiobjective optimization model with minimizing the makespan and the total tardiness is built.The multiobjective genetic algorithm based on the multipopulation strategy is designed.In order to examine the performance of the proposed algorithm,the classical algorithms are chosen as the compared algorithms,and the simulation experiments on some stochastic instances show that the proposed algorithm has better performance in solving this problem.(6)The multiobjective hybrid parallel machine scheduling problem is proposed,and the multiobjective optimization model with minimizing the total flow time and the total tardy number of jobs is built.The nondominated sorting genetic algorithm II based on the characteristic of the proposed problem is designed to solve it.The simulation experiments on stochastic instances are carried out.Compared to the classical multiobjective evolutionary algorithms,the experimental results show that the proposed algorithm has better performance in solving this problem.