Human-Computer Interaction for Part Selection in Product Design

We examine how humans can interact with a computing machine to explore good (least cost) combinations of parts in a functionally decomposed modular product. The machine uses data of parts usage from previous applications and generates the preferred combination that meets current design requirements. We use a Boolean function to represent functionality requirements and a classifier to estimate the Boolean function from incomplete previous parts usage information based on a decision tree. When no previous data are available, we propose an efficient data collection strategy. Results from simulation are presented to validate the algorithmic concept.