An expert system approach to diagnosis of product quality deviations

The diagnostic task of relating observed product quality data to operating parameters involves mapping from magnitude and directional changes in product quality attributes to explanatory changes in operating parameters. Working knowledge is typically in the form of individual parameter-product quality relationships. Thus, the predominant diagnostic task is one of assembling an overall hypothesis about changes in operating parameters from relationships which offer pieces of explanatory information. This kind of inferencing is referred to generically as Abductive Inference, where, given observed data, an explanation is found which best accounts for the data. A computational structure which captures general aspects of this type of reasoning is hypothesis assembly in which interacting hypothesis parts are combined to form an overall explanatory hypothesis.