A fuzzy fault diagnosis scheme with application

In this paper, a new fuzzy logic diagnosis strategy is developed where the emphasis is placed upon knowledge expression and approximate reasoning. First, the fuzzy relations, between faults and symptoms, are considered as: one fault may cause several symptoms, in turn, one symptom may represent several possible faults. Second, to solve the problem that once some symptoms have been detected it is generally very difficult to attribute them to a certain fault, we employ the fuzzy relation matrix to represent these complex fault-symptom relations which are foundations of reasoning, in which the probabilities of faults are expressed as fuzzy numbers, and complex fault-symptom relations are represented with a fuzzy relation matrix whose elements are obtained by fault tree and Bayes rule. Third, upon these relations the fuzzy recognition reasoning is accomplished, which can list all faults whose possibilities of causing the occurring symptoms are greater than a certain threshold. Finally, to make the diagnostic conclusion more accurate, the fuzzy relation matrix will be appropriately revised further, based on the obtained information. The computer simulation results show that location of the malfunction is deduced by full use of the relations between faults and symptoms.