Neural network-DFT based model for magnetostrictive hysteresis

Abstract

The magnetic behavior of materials is highly sensitive to the applied stress, which is subject of this paper. In magnetostrictive materials, the relationship B(H) is strongly related to the applied mechanical stresses. The purpose of this paper is to present a simple method to identify this relation for any value of stress σ. The proposed approach is based on the knowledge of the magnetic flux density spectrum for different stress σ. So, a feed-forward neural network is trained to learn the relation between the rank of harmonics, stress and the amplitude of each harmonic. To construct the hysteresis loop for a new value of the stress σ, the neural network predicts the spectrum of the magnetic flux density B, which is transformed back to the time domain.