Design of fault diagnosis expert system using improved fuzzy cognitive maps and rough set techniques

Abstract

Fault in a system can cause serious problems like explosion in the nuclear power plant or gas supply system. In order to diagnose faults, experts who has superior knowledge on the system is required. But, if the experts‘ knowledge or their reasoning skills are coded into the expert system, it can be used with many benefits. Until recently many fault diagnosis algorithms were developed. However, the design of the fault diagnosis expert system was being done with classical methods using graph searching, backward chaining techniques. When the expert diagnose faults, they use redundant knowledge without consciousness some times, so rule minimization technique adapted from rough set theory was applied to remove redundant knowledge which is not necessary to diagnose faults in this paper. Additionally, Improved Fuzzy Cognitive Map(I-FCM) which is a network represents system dynamics with feedback and Fuzzy Associative Memory(FAM) which is a neural-like memory capable of memorizing causal relationships are used to diagnose faults in the system. That is, new design approach to fault diagnosis expert system was tried without classical searching or reasoning methods. This design scheme was applied to a plant called "Valve station" which delivers LNG from LNG terminals to individual users or industrial plants.