Case-based reasoning applied to fault diagnosis on steam turbines

Abstract

Heuristic based expert systems have difficulty in providing comprehensive solutions to their end users. At the rule level, the knowledge is implicit and therefore can be difficult to justify. Such systems are accurate and efficient, but do not always provide comprehensive solutions. The use of case studies was therefore investigated alongside fault models, both of which are capable of providing solutions acceptable to end users. The research then lead to the design of a second generation expert system, where different knowledge sources (i.e. heuristics, models and cases) can support each others, or can be combined to produce a diagnosis which will still be accurate, but will gain in performance and comprehension. This paper discusses such ideas where case based reasoning can be at the core of the system.