Condition Monitoring and Fault Diagnosis of Induction Motor Using Support Vector Machine

Abstract

This article deals with the condition monitoring and fault diagnosis of a three-phase induction motor using a support vector machine classifier. By acquiring the three line voltages and currents of the motor in real time (experimentally on a 3-HP motor in the laboratory environment), total harmonic distortion is calculated, which in turn is used for the training of the support vector machine. A laboratory prototype has been developed through which various data have been generated by conducting extensive experiments on a healthy motor as well on a motor having faulty bearing, shorting of stator turns, and broken rotor bars with varying load conditions. The performance of the projected support vector machine-based scheme has been assessed for two kernel functions on the basis of fault classification accuracy. It can be noted that the accuracy of the radial basis function kernel is higher than that of the polynomial kernel. The proposed support vector machine-based scheme gives satisfactorily results, as the fault discrimination accuracy is found to be more than 98%. Simultaneously, it also gives an accuracy of the order of 95% for different motor design specifications, which confirms robustness of the proposed scheme.