The solution of many design problems involves two steps: the designer (1) creates a configuration by making component choices and (2) selects values for the parameters associated with the components in that configuration. For example, in automobile design a configuration decision may be to use disk brakes and a single turbocharger. Consequently, the parametric values to be chosen include disk radius and turbo inlet area. Mathematical models used to represent such problems and to evaluate chosen alternatives are often large, nonlinear, and involve both discrete and continuous variables. Because no single design algorithm will usually suffice in solving such problems, currently available computer tools are typically limited to a small range of problems or to parts of large problems. We believe a computer environment that allows flexible access to a diverse set of tools can help designers rapidly generate high quality solutions to a broad range of problems. In this paper, we test this belief on a design problem taken from a commercial auto manufacturer. We propose a framework for dealing with the general class of problems, and we describe the implementation of a novel design system that integrates math programming with knowledge-based and graph theoretic tools.
Solving Parametric Design Problems Requiring Configuration Choices
