Abstract
To address the limitations of conventional feature extraction methods in capturing fault information from operational current signals, the paper proposes a novel fault diagnosis method for permanent magnet synchronous motor (PMSM). The approach integrates markov transition field (MTF) image fusion, a newton raphson based optimization(NRBO), and a stochastic configuration network (SCN). Firstly, a high-fidelity simulation model of the PMSM is developed, and the three-phase current signals are extracted as the effective fault indicators. Secondly, the three-phase current signals are transformed into two-dimensional MTF images, which are then fused into RGB images. Thirdly, extract the color, shape and texture features of MTF fused images separately to construct feature vectors of fault information. Finally, to enhance classification accuracy, the SCN model is optimized using the NRBO algorithm, improving its generalization and fault identification capabilities. The extracted fault feature vectors are then input into the NRBO-SCN model for fault identification. By collecting actual PMSM fault diagnosis data for verification, compared with other fault classification methods, the proposed method in this paper has superior fault classification performance and can accurately identify the fault types of the PMSM eccentricity and demagnetization.