Fault Diagnosis of the Asynchronous Machines Through Magnetic Signature Analysis Using Finite-Element Method and Neural Networks

Abstract

This paper presents a method for the identification of winding failures in induction motors. The types of failures include unbalanced currents flowing into the motor and short-circuit of the winding. The radiated magnetic field of a typical induction motor was studied while various types of failures applied to the machine. The implementation was performed by applying different types of unbalanced currents flow into the machine. The fields were obtained from both numerical finite-element simulations as well as from experimental setups. The turn to terminal and turn to turn short-circuit of the motor's winding were studied. The frequency response of the 3-D finite-element (3DFE) model of the motor was implemented up to high-order frequencies. The numerical results were compared with the measurement results. The fields with unbalanced currents and short-circuit conditions were identified by studying the harmonic orders of the radiated magnetic fields. This was also implemented using artificial neural networks (ANN). The results show that the signature study of the experimental as well as the simulation models can be utilized for failure identification in electric motors with a high level of accuracy.