On estimating the feasible solution space of design

Abstract

The feasible solution space for a mechanical engineering design is determined by the variables and constraints involved in the design. Although constraint satisfaction programs can be used to find individual solutions in this space, their search strategies become very inefficient if the search space is much larger than the solution space. The user then cannot be sure whether a prolonged unsuccessful search is due to this inherent inefficiency, or because no solution space actually exists. In addition, such programs provide no information about the amount by which a satisfactory design solution can be adjusted without violating any constraints. This paper presents the implementation of a computer program which utilises a domain propagation algorithm to estimate the feasible design space. The theoretical descriptions of domain propagation and the algorithm are also discussed. Case studies show that this program enhances the efficiency and robustness of the constraint-based design process.