DIAGNOSING THE OCCURRENCE OF FAULT IN MULTIPHASE INDUCTION MOTOR THROUGH FEED FORWARD NEURAL NETWORK

Abstract

The significance of induction motors has remarkably increased in the modern era since it plays a predominant role in both the industrial and domestic applications. The present study emphasizes the implementation of advanced approaches in detecting the occurrence of fault in a multiphase induction motor. The existence of faults like bearing fault, rotor fault, stator fault and winding fault in motor has to be rectified since it affects the performance of the motor. Hence, a signal processing approach is presented in this work to diagnose the specified faults, in which the Least Mean Square (LMS) filter is employed to remove the noises in the signal. After the removal of noises, the signal is segmented into multiple samples by using the Hilbert Transform and the features are extricated from the segmented signal with the assistance of Scale Invariant Feature Transform (SIFT). Eventually, the extracted features are classified by implementing the Feed forward Neural Network (FFNN) classifier, through which the faults in the speed of the motor is accurately diagnosed. In consideration with the failure model, a total of 100 samples are tested to assess the accuracy of FFNN classifier. The simulation results have proved that the proposed FFNN provides maximum accuracy in detecting the occurrence of fault in a multiphase induction motor.