On the application of natural algorithms to structural design optimization

Abstract

Powerful computational methods based on metaphors of ‘urvival of the fittest’ (the genetic algorithm) and human brain activity (the neural network) have made significant progress in engineering where there are needs for search and learning mechanisms. The principal subject of the paper is the ‘genetic algorithm’, a ‘population-based’ method of searching large combinatorial (design) spaces to find the optimum combination of design variables. Attention needs to be given to the form and organization of the algorithm if it is to be applied to large-scale problems. Consideration is given to the development of a space condensation heuristic which progressively reduces the size of the multidimensional space being searched thus leading to a more economical application of the algorithm. The approach to adaptivity of controls and the type of penalty function used for the design constraints are explained. Some results from a study of optimum design of a multistorey frame are included by way of illustration.