Prediction of Magnetic Domains Movement in an Electrical Steel Sheet During Magnetization Based on the Deep Learning of Artificial Intelligence

Abstract

In this paper, a longitudinal magneto-optical Kerr microscope is used to study the magnetic domains structure of a Goss-textured grain-oriented silicon steel sheet during the magnetization process. We observe that when observing along the easy magnetization direction, the magnetic domains magnetization process is mainly controlled by the movement of the domain walls. When magnetization deviates from the easy magnetization direction, the rotation of the magnetic domains can be observed during the process of the domains magnetization. In order to quantitatively analyze the magnetic domains, the area of the magnetic domains in the observation area is calculated. The area of the magnetic domains can further indicate the degree of magnetization of the silicon steel sheet. At the same time, under the limited magnetic domains structure images, the two neural networks of image processing technology and deep learning-ConvLSTM (Convolutional Long-Short Term Memory) and ConvGRU (Convolutional Gate Recurrent Unit) are used to predict the magnetic domains structure. Through the comparison of the results of the two methods, it can be seen that the predicted magnetic domain structure performance of ConvLSTM is better.