Bearing Fault Diagnosis in Permanent Magnet Synchronous Motor Using Deep Neural Network

The diagnosis of bearing faults poses an essential problem in the monitoring of rotating machinery and has gained increasing attention. Rolling bearings are prone to various problems under demanding operating conditions that can lead to complete machine failure. This is reflected in the machine performance. If these faults are left unattended, they result in high levels of friction, which further causes immense heat in the machinery, higher than the machine could probably handle, eventually leading to equipment failure. This highlights the necessity of monitoring the machinery's condition to anticipate any potential issues arising from bearings. This study proposed and validated the use of deep learning techniques to diagnose and detect bearing faults in permanent magnet synchronous motors. In this study, a Visual Geometry Group (VGG) neural network was employed to diagnose the inner race fault and outer race fault against healthy data with the help of images obtained through wavelet analysis and kurtogram. The current and vibration signals from the permanent magnet synchronous motor have been taken for analysis. The Paderborn University bearing dataset is used to validate the training model's accuracy.