A Multi-domain Hybrid Genetic Algorithm For The Discrete Variable Structure Optimization

Aiming at the lack of local search ability of genetic algorithm for structural optimization and the features of structural optimum design with discrete variables, Combined with the direct search methods and the improved genetic algorithm, A multi-domain hybrid genetic algorithm for the discrete variable structure optimization is proposed in this paper. The main contents include:(1) The basic principle and method steps are outlined for discrete variable structure optimization design of direct search to the full stress method, the Fibonacci method and the relative difference quotient method.(2) Biological basis of genetic algorithm, the basic operation of genetic algorithm, parameter selection and constraint conditions of processing are introduced; the characteristics of genetic algorithm and the algorithm convergence criterion are analyzed; the improved genetic algorithm is put forward.(3) A multi-domain hybrid genetic algorithm for the discrete variable structure optimization is proposed. The feasible domain of structure optimization is divided into stress constraint feasible domain, the displacement constraints feasible domain, frequency constraints feasible domain, etc. The quasi full stress method is used to search the optimal solution in the stress feasible domain; In this stress constraint optimal solution as the initial point, the relative difference quotient method or the Fibonacci method is used to find the optimal solution of meet displacement constraint; In order to find the global optimal solution, the improved genetic algorithm is carried out; The local optimal solution by direct search method as initial population genes of genetic algorithm, so that the genetic algorithm and direct search algorithms both organic unifies in together; The genetic algorithm and direct search method together, A new multi-domain hybrid genetic algorithm is constructed. Both played a direct search algorithm local search ability strong characteristic, and has played a good overall characteristics of genetic algorithm; combining the strengths of both genetic algorithm and direct search algorithm, so as to maintain the population diversity and avoid the local optimum.(4) The computer programs of the quasi full stress algorithm, Fibonacci algorithm, the relative difference quotient algorithm, the improved genetic algorithm and the multi-domain hybrid genetic algorithm are compiled; the correctness of the algorithm and the validity is validated by the examples of the function optimization and structure optimization design.