Neural/fuzzy systems for incipient fault detection in induction motors

Abstract

Industrial motors are subject to incipient faults which, if undetected, can lead to motor failure. The necessity of incipient fault detection can be justified by safety and economical reasons. The technology of artificial neural networks has been successfully used to solve the motor incipient fault detection problem. The artificial neural network, however, does not provide any heuristic knowledge of the fault detection procedure. This paper will introduce a hybrid neural network/fuzzy logic system that not only provides better performance on detecting motor faults, but also allows heuristic interpretation of the network fault detection process. The system will be applied to bearing faults in single phase induction motors. The paper will discuss how to extract heuristic information from the system to gain further insight into the motor fault detection procedure. >