Study On The Improvement Of Cellular Multi-objective Evolutionary Algorithm And Their Application In The Engineering Optimization

Evolutionary algorithm, as a kind of group intelligence search method, has manyadvantages in solving multi-objective optimization problems. It has been a research hotspotin the field of evolutionary computation by using evolutionary algorithm to solvemulti-objective optimization problems, however, there are some problems existing in thesolving multi-objective optimization issues, such as poor convergence, poor distribution ofPareto Front and so on. Based on these problems discussed above, two improvedevolutionary algorithms will be proposed in this thesis to improve the performance of thealgorithms.The first type is cellular multi-objective genetic algorithm based on multi-strategydifferential evolution. Having analyzed the pros and cons of different differential evolutionmodels, the algorithm defines a selection operator of multi-strategy differentialco-evolution combining with cellular automata model. Aiming at the current disadvantageof crowding distance evaluation method, it introduces a crowding distance evaluationmethod based on entropy and a new replacement strategy is proposed. Compared withNSGA-II, MOCell and CellDE, this algorithm occupies a better convergence and diversity.Meanwhile, it has improved the coverage of the solution to some extent, particularly beingsuitable for solutions of high dimensional and complicated multi-objective optimizationproblems.The second one is cellular multi-objective particle swarm algorithm based onmulti-strategy differential evolution (MPSOCell). The algorithm, integrating the cellularautomata theory into particle swarm optimization algorithm, studies the communicationstructure and information transfer mechanism of the particle group on account of analysisof the principle of particle swarm optimization. In order to avoid particle flying too fastinto local convergence, it proposes a strategy for limiting the flying speed of the particleand introduces a multi-strategy differential evolution selection operator to increase thedisturbance of particles in this thesis. Experiments clarify that this kind of algorithm has abetter convergence and diversity compared with the comparison algorithm.In order to further verify the effect of the algorithm, cellular multi-objective particleswarm algorithm based on multi-strategy differential evolution is applied in the designissues of static multi-objective optimization of truss structure. According to the results ofcomparative experiments, MPSOCell compared with other algorithms, possesses a higher convergence precision and better spread of extreme point. What’s more, Pareto Frontdistributes more uniformly and solution sets can get optimal solution of single objectiveoptimization.