Design as a Marked Point Process

Abstract Although AI systems which support composition using predictive text are well established there no analogous technologies for mechanical design. Motivated by the vision of a system that interactively suggests features to designer, this paper describes theory, implementation and assessment an intelligent learns from family previous designs generates inferences form spatial statistics. The formalism presented, models 3D design activity as ‘Marked Point Process’ enables probability specific being added at particular locations be calculated. Because resulting probabilities updated every time new feature is predictions will become more accurate develops. This approach allows cursor position on CAD model implicitly define focus query made statistical model. authors describe mathematics underlying amalgamates frequency occurrence in existing product family. Having theoretical foundations work, generic six step process described. then illustrated circular hole generated dataset hydraulic valves. how positions each design's extracted can homogenized through rotation scaling. Results suggest within part families (i.e. with common structure) marked point effective predicting incremental steps development designs.