Intelligent design problem solving using case-based and constraint-based techniques

Date of Completion

January 1995


Artificial Intelligence|Computer Science




The importance of engineering design to the final product's quality, reliability, manufacturability, and ultimately to company profitability has sparked lively research in design automation. In design, past experience is often used to guide and inspire new design solutions. Thus, Case Based Reasoning (CBR) has received much attention as a viable formulation for design problem solving, as it allows prior designs to be stored in the case base to help solve new problems.^ The case based reasoning process of adaptation, where an old design case is changed to fit the needs of the new problem situation, is often considered to be the most difficult component of a case based reasoning system, since adaptation strategies often only apply to particular domains, resulting in an ad hoc adaptation methodology dependent on domain knowledge. Adaptation is particularly difficult in complex problem domains such as design, where it is often not enough to simply adapt one old case to solve the new problem, but rather many existing cases contribute their knowledge about solving the new problem, and thus these many cases must be combined into a valid solution.^ This research investigates a methodology which formalizes this case combination process using constraint satisfaction techniques, in order to make the process systematic and allow its application across a general class of design problems. As the constraint satisfaction problem (CSP) itself is a difficult problem to solve, another goal of this research is to improve CSP performance by providing it guidance from the case base. The described methodology has been implemented into a system for adaptation, COMPOSER, and tested on assembly sequence generation and configuration design problems. ^