Combined Preisach–Mayergoyz-neural-network vector hysteresis model for electrical steel sheets

Abstract

Three versions of a vector hysteresis model for electrical steel sheets are presented, based on the function approximation capabilities of feed-forward neural networks and the memory mechanism of vector hysteresis proposed by Mayergoyz. The first model handles arbitrary vector magnetization patterns, but requires a very extended data set for the training of the neural network. The second model is suitable for convex induction loci and allows a reduction of the required training set. The third model handles the features of the considered magnetization pattern in an alternative way and relaxes the convexity requirement. The choice of the specific model, its parameters, and the network training set depends on the types of magnetization patterns concerned. Arbitrary high accuracy can be reached by extending the complexity of the model and/or the size of the training set. Experimental results for the third model are presented and show the good accuracy of the approach. Standard neural network algorithms are used.