Evolution of Optimum Structural Shapes Using Genetic Algorithm

The optimum structural shape design of skeletal structures has traditionally been solved by using conceptual designs, often based on ground structures, in which the generated designs resemble the conceptual designs. If the design could be approached without the use of conceptual designs or ground structures, there is a potential for generating new and innovative designs, especially when more complex design problems are attempted. A methodology using such a design approach, made possible by the use of the genetic algorithm, is proposed in this paper. The proposed methodology uses the genetic algorithm to evolve optimum structural shape designs, which are free to assume any geometry and topology. The sizing, configurational, and topological aspects of the design are simultaneously addressed. Discrete member sizes are considered. The methodology is capable of addressing single and multiple loadings and plane or space structures. A new string representation scheme, generalized penalty function, and fitness function are introduced. The proposed methodology is applied to two illustrative examples involving single-span plane trusses, and the results are described.