GIVING DESIGNERS A CHOICE OF OPTIMAL DESIGNS

摘要

This paper discusses how a single execution of a genetic algorithm can provide designers with a choice of designs representing a variety of significantly different ideas. Each of the ideas is optimum or near optimum in some sense. This was made possible through two modifications to the genetic algorithm: topology limitation and MOGA. Topology limitation forces the genetic algorithm to consider multiple topologies in each generation. Topologies represent significant differences in design as defined by topology variables. In the bridge example, topology limitation produced a final generation of designs that included both familiar (i.e. suspension, cable-stayed, etc.) and unfamiliar bridge topologies. In the frame example, the final generation included both rigid and hinge-connected topologies. Furthermore, it revealed that unbraced topologies were not feasible for the frame example. In essence, topology limitation forces the genetic algorithm to find the best designs for each of the best topologies. Even though one topology may emerge as the best according to the design criteria implemented in the algorithm, the designer may select another topology based on subjective criteria that were not implemented in the algorithm.