Improved Bearing Faults Detection in Induction Motor Through 2-D Convolutional Neural Network

Electric motors are essential for converting electrical energy into mechanical energy. Among the various types of electric motors, induction motors are widely used in industries due to their low price and high durability. To avoid costly maintenance and prevent industrial production line failures, condition monitoring and fast fault detection of induction motors are crucial. Among all types of faults in induction motors, bearing faults represent the largest share of faults in induction motors. In this paper, first the types of faults associated with induction motors and their occurrence rates of them have been stated, then the existing methods for bearing fault detection have been reviewed and finally our approach has been presented. In this paper, a method for detecting three types of bearing faults—ball faults, outer race faults, and cage faults—in induction motors has been introduced, using only motor current signals acquired from a testing apparatus. The outlined technique is based on a 2-D convolutional neural network (CNN), achieving high accuracy and low loss.