Mechanical design synthesis from sparse feature-based input

We are researching a new paradigm for CAD which aims to support the early stages of mechanical design well enough that designers are motivated to actually use the workstation as a conceptual design tool. At the heart of our approach is shape synthesis, the computer generation of part designs. The need for such automation arises from the fact that any mechanical part is defined by two kinds of geometry: features that are critical to its function (application features), and the material that merely fleshes out the rest of the part (bulk shape). Application features are most often associated with contact surfaces of the part, for example, a bore for a bearing or a mounting surface for a motor. They are the high-level entities in terms of which the designer reasons about the design. Bulk shape must obey certain constraints, such as noninterference with other parts, minimum allowable thickness of the part, etc., but is somewhat arbitrary. We are developing a system wherein the designer inputs the application features, along with topological constraints, degrees of freedom, and boundary volumes, then the bulk shapes of the parts are synthesized automatically. Overall economy is enhanced by reducing the amount of input necessary from the designer, by providing for more complete exploration of the design space, and by enhancing manufacturability and assemblability of the component parts. This paper presents the functional requirements of such a system, and discusses preliminary results.