A Method of Weak Fault Detection Based on Sparse Representation for PMSM

摘要

With the development of modern industry, permanent magnet synchronous motor(PMSM) fault detection has become a hot issue, for the reason that the early detection could avoid the catastrophic effects caused by the further deterioration. In this paper, PMSM weak fault detection method based on construction T^{2 statistic is proposed. Firstly, K-singular value decomposition (K-SVD) algorithm is used to sparse represent the original signal of PMSM. In this step, the reconstructed signal could be acquired by sparse coefficient matrix and the corresponding dictionary. Then, the T2 statistic construction method is established based on the residual of training and testing signal to analyze the statistical characteristics of the residuals. Finally, by judging the relationship of T2 statistic and the threshold set by the original signal, the detection result could be obtained. The simulation results show that construction T2 statistic can effectively detect the PMSM weak fault.}