Isotropic vector hysteresis modeling with feed-forward neural networks

Abstract

A new vector hysteresis model is presented, based on the function approximation capabilities of feed-forward neural networks. Two-dimensional circular and elliptical magnetization of laminated SiFe steel sheets can be successfully handled by the model. A feed-forward neural network with four inputs, derived at each time step from the time-dependent magnetic induction vector, yields an accurate prediction of the magnetic field strength vector. Measurement results for a steel sheet sample are used to train and test the neural network. The model accuracy is good and can be easily adapted to the requirements of the application by extending or reducing the network training set and thus the required amount of measurement data. Besides, the presented technique is fast, requires no large data set, and applies standard neural network algorithms. Future extension of the model to other magnetization patterns is possible.