A Methodology for Shrinkage Predictions in Powder Pressed Parts Using a KBS

Summary

Information gathering, evaluation, formulation of decision logic, and optimisation are critical to design. In practice, the characteristics of a design task are such that experience is, at times, fundamental to its early completion. Some of this knowledge is representable as heuristics and hence amenable to being addressed using Artificial Intelligence techniques. Previous works, using finite elements and experimental techniques, were unsuccessful in alleviating the shrinkage variation problem the company was facing. This paper details the results of a successful and novel approach, based on a rule-based Expert System, in predicting shrinkages in both symmetrical and asymmetrical powder pressed components commonly used in deflection units. The methodology developed is based on strong and weak theories implemented in a learning Expert System called SPES that resulted from the research project.