PRODS: A shell for industrial diagnosis

摘要

PRODS (PROspective Diagnosis Expert), a shell for building expert systems based on prospective diagnosis, provides the domain expert with a knowledge acquisition interface and a suitable inference engine. Two kinds of knowledge contribute to modeling the diagnostic process. Failure knowledge expresses cause-effect relations between faults as well as the relations between the result of tests and the ‘degree of belief’ about the presence of a fault. Strategic knowledge contains criteria to guide the selection of the appropriate action. A criterion refers to a particular viewpoint (such as the cost of a test or the estimated occurrence of a fault) and its application orders or prunes the set of the actions to take with respect to this viewpoint. The tool for knowledge acquisition provides the expert with a user-friendly interface in order to acquire both kinds of knowledge. Prospective diagnosis evolves acquiring new data to increase the knowledge about the system while the strategic criteria guide in this choice. The inference engine proceeds, alternating data acquisition and focusing, until the most probable cause of the malfunction is found and a repair is executed.