Research And Application On Constratined Multi-objective Evolutionary Algorithm

Multi-objective optimization is an important problem in the field of natural sciences and engineering. As the optimal solution of multi-objective optimization problems is different from than of single-objective optimization, so the traditional optimization algorithm can not properly solve the multi-objective optimization problems. The late 80s of last century, the international community began to research the multi-objective optimization evolutionary algorithm, and a variety of classical multi-objective evolutionary algorithm appeared. To the beginning of this century, the study on multi-objective evolutionary algorithm entered a boom period, and evolutionary algorithms have become the research focus with most practical value. Today, the multi-objective evolutionary algorithms have been widely used in natural sciences, engineering and social sciences and other areas.This article examined a large number of domestic and foreign literatures, and constrained multi-objective evolutionary algorithm is improved on the base of previous work. The improved algorithm is applied to mechanical optimal design. The main thesis reads as follows:1. The research progress and research focus on multi-objective evolutionary algorithm are briefly introduced.2. The basic concepts of the multi-objective optimization problem and the major multi-objective optimization algorithm are detailed introduced.3. A new cluster analysis method is introduced in multi-objective evolutionary algorithm, and an improved constrained multi-objective evolutionary algorithm based on group classification is put forward. The basic idea of this algorithm is:First, the population is divided into infeasible group and feasible group, in turn feasible group is divided into feasible non-Pareto group and feasible Pareto group, and then feasible Pareto group is divided into non-clustering Pareto group and clustering Pareto group by means of improved clustering method. The appropriate adaptation values are given with these four groups and a generation evolution is completed by the selection, crossover and mutation.4. The improved constrained multi-objective optimization evolutionary algorithm is applied to optimize the parameters of planetary gear transmission design, and the results superior to the traditional optimization methods and single-objective evolutionary algorithm results.