Overview
Professor Wright is experienced in the application of artificial intelligence methods for the analysis of failed mechanical elements. After completion of two demonstration systems (FAXS I and FAXS II), a full-scale system (FALCON) has been implemented. The expert system produces as a final product a technical report which includes a diagnosis of both the mechanism of the fracture and the cause of failure. Categorization with Bayesian probabilities has proven more efficient than the commonly employed production rule approach. Additionally, the report includes a list of past case studies most similar to the current study.
A number of extensions to FALCON are now anticipated. The ability of the system to compare and to quantify the similarity between case studies has led us to believe that a method of learning without direct contact with experts is feasible. Further, adding the task of recommending design changes will require significant program development. In brief, the research brings together the disciplines of materials science, mechanical design and artificial intelligence.