Classification and Detection of Demagnetization and Inter-Turn Short Circuit Faults in IPMSMs by Using Convolutional Neural Networks

This paper represents classification and detection algorithm for inter-turn short circuit faults (ISCF), demagnetization faults (DF), and hybrid faults (a combination of ISCF and DF) in an interior permanent magnet synchronous machines (IPMSMs). We confirm the characteristic of normal and various fault types by using the fast Fourier transform patterns of the phase currents in finite element analysis model. Based on the characteristic, we propose a classification and detection method using a convolutional neural networks model. The training and test data were extracted using three-phase currents obtained from the experiments. The test results verify the effectiveness of the proposed method.