Determination of optimum design spaces for topology optimization

Due to tolerances allowed in deciding the design space in preliminary designs, two approaches are proposed to find the optimum design space systematically. One is to use Taguchi orthogonal array to do experiments for some design spaces and find optimum topologies in these design spaces. The results of the experiments are used to determine the optimum boundaries of the design space. The other approach is to use genetic algorithms (GA) to search for the optimum boundaries of the design space. To reduce the computational burden of GA, the artificial neural network (ANN) is used to replace the time-consuming processes of finding optimum topologies in all individuals (design spaces) of a generation. The GA search is thus accelerated. Two examples are used to demonstrate and test the approaches. Based on numerical results the methods are reliable, practical, and efficient. (C) 2000 Elsevier Science B.V. All rights reserved.