Optimal Design of Mechanical Components with Genetic Algorithm

Abstract

Our aim is to propose design-making tools, which can be introduced in Computer Aided Design software. Thank to the analysis of the design process of a known mechanical system, we can formulate the mechanical design problem as an optimization problem, called problem of optimal design. The latter contains non-linear equations, inequality constraints and mixed variables that are continuous and discrete. There are also interdependent discrete parameters whose values may be taken in normalized tables and which directly depend on the choice of one of the discrete variables. Some problems of optimal design have been solved with a method using the augmented Lagrange multipliers, combined with a branch and bound algorithm [5]. In order to calculate the functions gradients, interpolation functions have been used to bind the discrete parameters to the discrete variable, on which they depend. Good results have been obtained for many design problems. However, for some complex structures, we can not obtain a simple analytical expression relating the different parameters to a discrete variable. The functions gradients can not be calculated so that classical methods can not be used. In this paper, we propose to use genetic algorithms (GAs) to solve these difficult problems of optimal design. The genetic algorithm is a recently emerged heuristic optimization technique, based on concepts from natural genetic and guided by the model of Darwin.