A fuzzy diagnosis system for automatic classification of acoustic burst events in nuclear power plants

A fuzzy diagnosis system for automatic classification of acoustic burst events of the loose parts monitoring system (LPMS) in nuclear power plants has been developed. Useful features of accelerometer signals which are characteristic for different types of acoustic events are extracted as input of the diagnosis system. Membership functions of the fuzzy variables and the fuzzy rules are generated using statistical data analysis. For the fuzzy distance measure an inference strategy which combines the min-max- and the average-operator is used. The performance of the fuzzy classification/diagnosis system is comparable to a neural network system trained with the same data. Copyright (C) 1998 IFAC.