Diagnosis of rotor demagnetization and eccentricity faults for IPMSM based on deep CNN and image recognition

Abstract Permanent magnet synchronous motors (PMSMs) have many advantages attributed to the structure of permanent magnet, which connects influences each type fault. However, diagnostic methods are not capable monitoring multitype faults with one index or method, due closed-loop control non-stationarity. There a few based on convolutional neural networks (CNNs) and vibration analysis; however, they limited by installed sensor locations affected various factors driven system. This paper proposes method diagnose eccentricity demagnetization in an interior PMSM (IPMSM) image recognition deep CNN motor stator current analysis. To extract fault features from IPMSM’s CNN, gray transformation algorithm that uses autocorrelation matrix is proposed enhance feature representations currents, currents data processed recursively timely. The testing accuracy 98.74% both designed model pyramidal Resnet-9 Resnet-15 indicates immune speeds loads.