Generating rules in expert systems for nuclear power plant diagnostics

Construction of a suitable rule base currently represents the most difficult, arduous, and time-consuming task involved in expert system development. In this paper a technique is discussed for automatically codifying the information in existing fault tress in the form of rules and an inference algorithm is presented which is suitable for diagnosing component failures from fault-tree-constructed rules. The algorithm uses a matrix representation of the rule base and fuzzy logic as the inferencing mechanism. An example of the diagnosis of failures in an electric power system using the system's fault tree is presented. An approach for generating rules form experience is also discussed.