Study On Improving Cellular Multi-objective Genetic Algorithm And Its Application In The Engineering Optimization

Generally speaking, most of the practical engineering optimizations aremulti-objective prolems. Multi-objective evolution algorithm can solve the above prolemseffectively. However, when solving more complicated multi-objective optimizations, themainstream multi-objective evolution algotithm, such as NSGA-II and SPEA2, canperform well in convergence but not in maintaining diversity. The multi-objective cellularalgotithm being better in both is springing up and gradually popular among scholars. Itperforms so bad in term of solving multi-objecive problems which ismulti-constraintdiscontinuous and high dimension, that we need to improve it.The algorithm combines the classical multi-objective cellular genetic algorithm withthe adaptive neighborhood structure technology, and produces an adaptive multi-objectivecellular genetic algorithm (AMOCell). It can dynamically change the structure of the cellneighbors along with the increase of the evolution generations. As a result; the algorithmcan keep a good balance between the exploration and exploitation. When comparing withthe other multi-objective evolutionary algorithms on solving benchmark functions,experiments showed that the new proposed algorithm had good convergence, distributionand expansibility on most problems.The improved multi-objective cellular genetic algorithm was applied in practicalengineering problems, in order to test its effect in solving practical engineering problems.Better Pareto front was obtained, applying the AMOCell to solve the optimal design of anI-beam model, optimal design of a machine tool spindle and optimal design of a reducer.