Using deep artificial neural networks for diagnosing bearing defects in induction motors with specific indicators

Abstract

Bearings are the most common type of defect in induction motors in the industrial world. This study aims to develop a comprehensive approach for monitoring and diagnosing bearing faults in these motors. However, two motors were dedicated to collecting a very large database using vibration sensors, one healthy and the other with a bearing defect. Sixteen temporal vibration indicators, including six that are specific to bearings, were calculated from the vibration signals, which represent the different operating states of the two motors. Based on simultaneous monitoring of these 16 vibration indicators, our Artificial intelligence (AI) system based on deep neural network (DNN) has proven its performance for early detection of rolling defects in induction motors, with very high correlation rates and very low error. This study provided a real approach for the use of remote monitoring of the state of induction motors in industry, with vibration indicators, based on DNN.