Real time application of artificial neural networks for incipient fault detection of induction machines

Abstract

This paper describes several artificial neural network architectures for real time application in incipient fault detection of induction machines. The artificial neural networks perform the fault detection in real time, based on direct measurements from the motor, and no rigorous mathematical model of the motor is needed. Different approaches used to develop a reliable fault detector are presented and compared in this paper. The designed networks vary in complexity and accuracy. A high-order fault detector neural network is discussed first. Then noise considerations are included in more complex fault detector models, since noise is an important factor in the design and analysis of real time fault detector neural networks. Simulation results show that with appropriate designs, artificial neural networks perform satisfactorily in real time incipient fault detection of induction machines.