Investigation of a Practical Method of Structural Optimization by Genetic Algorithms

Abstract

A new hybrid genetic algorithm (GA) and a parallel genetic algorithm are proposed in this paper in order to improve the convergence of structural optimization. The fully stressed design, FSD, method and a GA were incorporated into the hybrid GA. In the parallel GA, the “island model” was applied to reduce the computation time. Stress analyses were carried out with a commercial FEM code to obtain the optimum solution with high reliability even in a complex model. As a result of the axi-symmetric shell model, the optimum solution by using the proposed hybrid GA was 50% lighter than that by using simple GA. Furthermore, the computation time using the parallel GA with 10 processors was reduced by up to 1/7 of the time using one processor. These methods were then used to optimize the thickness of a three-dimensional shell model of a CRT. As a result of the application, the weight of the optimum model was 25% lighter than that of the reference model.