Detection and Classification of Multiple Faults of Induction Motors by Using Artificial Neural Networks

Detection of induction motor faults is a critical issue for the maintenance of induction motors. Analysis of the stator current is a widely used method to detect the faults of induction motors. There are many of studies on the detection of faults of induction motors but few studies on the detection of multiple faults are reported. In this study, the detection and classification of short-circuit faults of stator winding, broken rotor bars and inner/outer race bearing faults of a 3 kW squarel-cage induction motor are implemented by ANN. The study was carried out in three stages. In the first stage, the induction motor was tested with single faults including 1%, 2%, 3%, 4% and 5% short-circuited stator windings, three broken rotor bars, and inner/outer race bearing faults. In the second stage, induction motor was tested with 3% and 5% short-circuit stator windings and with three broken rotor bars. In the third stage, induction motor was tested with 3% and 5% short-circuit stator windings, rotor with three broken bars and inner/outer race bearing faults. The induction motor has been tested under full load. The detection and classification of multiple faults were realized by the proposed method. The highest performance rate in the detection of multiple faults was achieved with 87% accuracy rate. The resuts shows the applicability of the proposed method.