William F. Kaemmerer, James R. Allard
A system for continuously providing advice about the operation of some other device or process, rather than just problem diagnoses, must not only function in real time, but also cope with dynamic problem courses. The reasoning technique underlying such a system must not assume that faults have single causes, that queries to the user will be answered and advice to the user will be followed, nor that aspects of a problem, once resolved, will not reoccur. This paper presents a reasoning technique that can be used in conjunction with an inference engine to model the state of a problem situation throughout the entire problem-handling process, from discovery to final resolution. The technique has been implemented and installed on-line in a factory control room, as part of a real time expert system for advising the operators of a manufacturing process.