Partitioning around medoids as a systematic approach to generative design solution space reduction
Michael Botyarov, Erika E. Miller
Abstract
This study explores an approach to generative design solution space reduction by offering a flexible, efficient, and accessible method by leveraging clustering techniques. The studied method of generative design solution space reduction uses clustering analysis with a combination of the Gower distance matrix and partitioning around medoids in an iterative process. This iterative generative design solution space reduction method retains the originality of unique design solutions, while simultaneously reducing the quantity of design solutions presented to the user, theoretically improving cognitive function during the design process. Design originality is maintained since the clustering process groups similar designs into clusters, from which a systematic reduction of similar designs can be achieved, thereby leaving novel solutions from the design envelope. This paper presents this clustering approach in the context of an aircraft engine loading bracket with multiple nominal and continuous variables, however, this approach be transferred to other applications with similar variables. Further work can explore the relationship between reduced generative design solution spaces and human cognitive function.