Microcracks Fault Detection of Permanent Magnet Based on Magnetic Surface Information Characteristic Matrix

A new method for permanent magnet defect detection based on magnetic surface information features is proposed to solve the problem of double secondary air core permanent magnet synchronous linear motor permanent magnet microcrack detection. First, the equivalent magnetic charge method is used to establish an analytical model of the secondary air gap magnetic field. Second, a method for constructing a magnetic surface information characteristic matrix (MSICM) is proposed, which introduces the correlation coefficient and optimizes the density of the MSICM according to the expression requirements of microcracks of different sizes. Singular value decomposition is used to process the MSICM and obtain the fault characteristics. Finally, according to the use of thrust fluctuations caused by microcracks as the basis of classification, a sample library with shape features as the characteristic quantity is established, and a microcrack classification model is generated by using a whale optimization algorithm multiscale kernel convolutional neural network; this model is compared with other classification algorithms. Simulation and prototype experiments show that the rate of permanent magnet microcrack recognition by the proposed method exceeds 98%.