Adaptation of a neural/fuzzy fault detection system

Artificial neural networks and fuzzy logic have previously been used for systems decision- and adaptation. Neural networks have provided a robust means of making systems decisions for nonlinear applications, while fuzzy logic has proven capable of properly classifying “gray area” decisions. This paper introduces a hybrid neural/fuzzy system along with its adaptation algorithm which will have the advantages of both the aforementioned techniques. The neural/fuzzy system is applied to detecting bearing faults in single phase induction motors as an illustration. The system not only gives highly satisfactory fault detection results, but also provides valuable expert knowledge that may otherwise be unknown or incorrect