FLUX-BASED FAULT DETECTION IN ROTORS OF INDUCTION MOTORS, USING FINITE ELEMENTS AND NEURAL NETWORK

Milad N. Azari, Hossein A. Khazaeli, and Mehdi Samami

Keywords

Broken rotor bar, finite element method, induction motor, proba-bilistic neural network, wavelet transform

Abstract

The authors have tried to introduce and analyse the effect of the broken rotor bar fault on magnetic flux density in a three-phase squirrel-cage induction motor (IM). To this end, finite element models have been applied, due to non-linearity consideration of the mentioned method to achieve a detailed study. The transient analysis and also the geometric parameter determination of the IM have been done, using 2D finite element electromagnetic field analysis. This method is capable of obtaining significant parameters, including the flux density, flux linkages, magnetic energy, torque and induced electromotive force. The diagnosis of the broken bar fault is based on the analysis of the transient stator current waveform, which is known as the most effective fault signal and by using the wavelet transform due to its feature extracting capability of the signal. It should be noted that the healthy and faulty conditions of the motor are identified, using a probabilistic neural network with specified inputs and outputs, which are trained with proper data. Simulation results approve incremental behaviour of the flux density by increasing the broken bars and the load and also the more serious effect of adjacent broken bars compared with non-adjacent ones.

Important Links:

Go Back