Fault Classification and Prediction Using a Database Approach and Fuzzy Logic Pattern Recognition

Abstract

The importance, and likely influence, of data when selecting a fault classification and prediction methodology to review system condition by monitoring is discussed in this paper. A database approach using fuzzy sets to classify current operational parameters against a set of identified patterns is proposed. This approach is adopted because of its inherent expandability and adaptable nature when confronted initially with limited operational data.The methodology presented has been tested with data generated specifically to investigate fault prediction and classification and is discussed in detail. The main objective of this was to derive and test suitable fuzzy membership curves to be used in the fault classification and prediction database.Additionally, the database fault prediction and classification method has been proven using sample data gathered for a medium speed diesel engine. This paper concludes that the proposed methodology can be used as an effective fault identification tool.