Induction motor fault diagnosis via current analysis on time domain

This study proposes a novel approach to induction motor fault diagnosis through motor current analysis. Most of the previous works employing motor current analysis use spectral methods to extract required features for detecting motor faults. The proposed method, however, utilizes time domain information for this purpose. Energy, local extrema, kurtosis and skewness parameters constitute the feature set extracted from the motor current on time domain within sliding window. In fault detection and classification experiments, six identical three-phase induction motors are used with one of them being healthy reference and the remaining five motors being deliberately broken to have different faults. The proposed time domain based features are employed in well known Bayesian classifier. Efficiency of the proposed method is examined at various motor load levels. Experimental results verify that the proposed method successfully detects and discriminates different motor faults.