Stress-Based Crossover Operator for Structural Topology Optimization

Abstract

In this paper, we propose a stress-based crossover (SX) operator to solve the checkerboard-like material distributation and disconnected topology that is common for simple genetic algorithm (SGA) to structural topology optimization problems (STOPs). A penalty function is defined to evaluate the fitness of each individual. A number of constrained problems are adopted to experiment the effectiveness of SX for STOPs. Comparison of 2-point crossover (2X) with SX indicates that SX can markedly suppress the checkerboard-like material distribution phenomena. Comparison of evolutionary structural optimization (ESO) and SX demonstrates the global search ability and flexibility of SX. Experiments of a Michell-type problem verifies the effectiveness of SX for STOPs. For a multi-loaded problem, SX searches out alternate solutions on the same parameters that shows the global search ability of GA.