Automatic generative design to meet customer's preferences

Abstract

Design is a creative work that demands inspiration, disciplined crafts, and iterative rollouts. To assist such creative works, the generative design technique has been employed, in which design ideas are coded into computer programs and the programs build designs on behalf of human designers. Human designers work on programs, instead of designs themselves, to modify models and to adopt new design ideas. Although human designers do not directly develop models, they need to be kept involved in model development programs. This paper is concerned with a generative design method which does not require human designers to be deeply involved in the design process. The proposed method uses a genetic algorithm approach as an enabling mechanism. The genetic algorithm creates various shapes of models using the predefined genetic operators which are developed with the help of human designers. The method assumes that the customers express their preferences on designs and thus the preferences evaluation is incorporated into the fitness function of the genetic algorithm. The proposed method has been applied to ring design works.