A diagnostic system based upon knowledge and experience.

This paper proposes a system, a meta-reasoner, to diagnose other systems. This system, called DSKE, performs its diagnosis based upon the knowledge of the structure of the system under diagnosis and based upon previous diagnostic experiences with that system. Since DSKE is based upon first order logic, it has strong deductive powers. Given adequate information, it can diagnose problems quite impressively. Contrary to probabilistic and fuzzy logic approaches, our system is not based upon data which may or may not be available, and which may or may not be reliable. Further, our system makes use of available knowledge. Contrary to model based diagnosis systems, our system can handle uncertainty and incomplete knowledge. The algorithm we provide yields a minimal diagnosis which covers all the abnormal behaviors detected. This paper provides a detailed description of DSKE. The inner workings of the system are carefully and comprehensively analyzed. Examples illustrate the advantages of our system, and demonstrate its use. [ABSTRACT FROM AUTHOR]