Prediction of hysteresis loop in magnetic cores using neural network and genetic algorithm

Abstract

The dynamic hysteresis loops of a range of soft magnetic toroidal wound cores made from 3% SiFe 0.27 mm thick M4, 0.1 and 0.08 mm thin gauge strip have been measured over a wide frequency range (50–1000 Hz). A dynamic hysteresis loop prediction model using neural network and genetic algorithm from measurements has been developed. Input parameters include the geometrical dimensions of wound cores, peak magnetic induction, strip thickness and magnetizing frequency. The developed neural network for the estimation of hysteresis loops has been also compared with the dynamic Preisach model and Energetic model. The results show that the neural network model trained by genetic algorithm has an acceptable prediction capability for hysteresis loops of toroidal cores.