Detection of Permanent Magnet Damage of PMSM Drive Based on Direct Analysis of the Stator Phase Currents Using Convolutional Neural Network

Permanent magnet synchronous motors (PMSMs) due to their numerous advantages, such as simple and compact design, easy production and high power-to-weight ratio, high power factor in the range of constant torque, low inertia, and more precise control compared to other electric motors, are increasingly used in various applications, such as electric vehicles, wind power, home, and industrial appliances. However, PMSMs, such as other electrical machines, may subject to various types of damage during the operation. Due to their use in devices of a critical nature (e.g., transport applications), the detection of these damages at their initial stage constitutes an extremely important issue. In this article, we investigate the possibility of the permanent magnets (PMs) faults detection of PMSM using a convolutional neural network (CNN) based on the raw stator current data. The article aims to show the possibility of detecting the PM demagnetization using the information obtained directly from the measured current signal as well as indicate the influence of simultaneous incipient interturn short circuits and the operating condition of the drive on the accuracy of the developed diagnostic system. The results of the experimental research carried out on a specially designed PMSM show the impressive capability of the developed CNN-based diagnostic system to precisely detect the initial phase of PMs damage.