The Application of Genetic Algorithms to Conceptual Design

Abstract

Design is not a simple hierarchical process where the designer is presented with a set of requirements and works steadily through a decomposition strategy moving from abstract concepts to the final concrete product. The design problem is ill-defined and changes as the designer explores it through solutions and partial solutions. Designers use a ‘solution focused’ strategy. Feasible solutions are posed to probe the ‘instability of the problem’ and the ‘limitations of the way the problem is framed’. It is common for architects, in professional practice, to first simplify the problem so they can generate a rough solution. This solution is then used to develop understanding of the problem which leads to a gradual refinement of that solution⁹. Neither is design simply a matter of iteration around an essentially hierarchical process. Information gained during the design process can prompt the designer to transfer the design effort to higher levels, or to a location remote in the hierarchy. Design is essentially a heterarchical, possibly chaotic, process. The heterarchical nature of design is even more apparent in team design and enshrined in the philosophy of concurrent design. Given the perceived nature of the design process, adaptive computing techniques with their property of emergent behaviour present an attractive paradigm for conceptual design. An overview of the evolutionary design capabilities is presented in Appendix A for those who are unfamiliar with the techniques.