Automated performance-based design space simplification for parametric structural designNathan Brown and Caitlin Mueller, Proceedings of the International Association for Shell and Spatial Structures (IASS) Symposium, 2017
As computation has advanced, more designers are becoming familiar with parametric and performance-based design space exploration, techniques that can provide feedback and guidance even in early-stage design. However, two downsides of such techniques are the time and expertise required for problem setup, and the potential of the large volume of generated data to become overwhelming and difficult to absorb. Researchers must find ways to organize performance-based information and simplify exploration so that the design process is more manageable, while ensuring that performance feedback leads to better outcomes. This paper proposes two new applications of traditional optimization methods that can help simplify early-stage architectural or structural parametric design. The first involves analyzing the design variables considered in the problem, ranking their importance, and determining which ones should be eliminated or emphasized during exploration. The second method clusters designs into families and enables designers to cycle through these families during exploration. Two structural design case studies are presented to illustrate the possibilities created by variable analysis and clustering in conceptual, performance-based design.