A Machine Learning Approach to Automated Design Classification, Association and Retrieval

Acquisition and recall of associations between problem descriptions and solutions is a critical task of case based design systems. The organization of d esign knowledge impacts the quality of inference and support a designer may der ive from a case based system. A machine learning approach for classifying and le arning over case data may be used to create an intelligent interface between des igner requirements and available design knowledge. Ideally the learning algorith m should be able to characterize cases and the mapping relationships within them at varying abstraction levels. This paper explores two neural architectures bas ed upon the Adaptive Resonance Theory for performing classification and learning tasks in the engineering design domain. Results from applying these algorithms to classifying and learning from a set of bridge design case descriptions are pr esented.