Knowledge management for preliminary mechanical design

Richard Haygood Crawford, Purdue University

Abstract

Preliminary mechanical design can be viewed in terms of two exploration processes: exploration of different problem formulations (different constraints, alternate goals, variations on solution strategies, etc.) and exploration of problem solutions. There are a number of computer tools available to support the generation of single solutions to design problems, including optimization programs, mathematical libraries, geometric modelers, and sophisticated analysis packages. However, there are few tools to manage the exploration of alternative problem formulations and competing solutions. The thesis describes research in knowledge management for preliminary computer-aided design. The goal of the research is to develop a representation and control mechanism which (1) allow comparison of different sets of design specifications, (2) assist the study of trade-offs between competing design solutions, (3) support hierarchical decomposition of complex problems into interacting subproblems, and (4) model not only design problems but also their solution processes. A computer-based representation scheme which models design problems with distinct specifications as design contexts, and possible solutions to these problems as context instantiations, are the results of this research. Complex problems are modeled as networks of coupled design contexts, and networks of design instantiations form possible solutions. Instantiation of context networks is directed by an event-driven control mechanism. Iterative solution techniques, such as optimization, are supported, and traditional engineering analysis methods are naturally integrated into the representation. This representation forms the basis for a problem-solving CAD architecture to assist mechanical engineers in preliminary design.

Degree

Ph.D.

Advisors

Anderson, Purdue University.

Subject Area

Mechanical engineering|Computer science

COinS