An application of artificial neural networks and PCA for stator fault diagnosis in inverter-fed induction motors

This paper presents a method for fast classification of stator short-circuit faults in a inverter-fed induction motor operating at steady-state, under a wide range of speed and load conditions. Induction motors are widely used in several industrial applications due to its robustness, low cost and high reliability. Early detection and proper fault diagnosis reduce the maintenance cost and also increase process effectiveness. The amplitude of the stator current signal, in the time domain is presented as input to a multilayer perceptron (MLP) network for the classification of stator faults. After a proper discretization of the current signal, the technique of principal components analysis (PCA) is applied allowing a reduction of the classifier complexity. Results obtained from 750 experimental tests are provided and compared to validate this study. The obtained results indicate that this approach can be employed to classify stator short-circuit faults in inverter-fed induction motors.