Motor Fault Diagnosis of a Brushless DC Motor Using Fast Kurtogram on Convolutional Neural Network

Abstract

DC motors are widely applied as reliable industrial machines. However, it may fail due to some defects, unfitting operations, or mechanical wear. Motor maintenance is necessary. To achieve this, detection of a probable problem such as a broken motor part before progressive problems occur. Detection of faults from a motor is the new trend to classify broken components of a motor. In this study, 2DCNN is classifying BLDC motor faults is used and determining its performance. By integrating the Fast Kurtogram algorithm as feature extraction, healthy and faulty signals can be converted into an image for the 2DCNN fault diagnostic algorithm. The fault-finding model was developed, and it classified healthy motor faults such as bearing, winding, and rotor faults with an overall accuracy of 83 percent. The superior performance of the 2DCNN model is evident compared to 1DCNN.