Mathematical Model and Neural Networks for the DynamicHysteresis Losses Calculation

Abstract

In this paper, the dynamic hysteresis energy losses of the FeBSiC amorphous alloy sample were determined by means of mathematical model and neural
networks. The changes in magnetic hysteresis curves as a function of frequency are introduced via the variation of the coercive magnetic field. So, a neural network (NN) has been trained to learn this function and used in the modelling of frequency-dependent hysteresis with a mathematical model. The model can calculate core losses based on the input parameters obtained from experimental measurements. This study is especially aimed at giving improved issue to avoid the empirical functions that require the identification of several parameters.