Dynamic magnetic hysteresis modelling using Elman recurrent neural network

Issue title: Proceedings of the tenth International Symposium on Applied Electromagnetics and Mechanics ISEM-Tokyo

Article type: Research Article

Authors: Saghafifar, Mahmoud^a | Kundu, Anjan^a | Nafalski, Andrew^{a; *}

Affiliations: [a] University of South Australia, School of Electrical and Information Engineering, Mawson Lakes, 5095, South Australia

Correspondence: [*] Corresponding author: Andrew Nafalski, Tel.: +61 8 8302 3932; Fax: +61 8 8302 3384; E-mail: nafalski@unisa.edu.au

Abstract: Dynamic magnetic hysteresis modelling is of crucial importance in determining the electromagnetic behaviour of magnetic cores. Among the different models of magnetic hysteresis, Preisach model is one of the more commonly applied. The dynamic Preisach model is a generalisation of the static Preisach model, and is obtained by adding the rate of change of the output variable, which makes the numerical implementation more complex. Since dynamic hysteresis models require mapping of time dependent sequences, a feature available in some neural networks, in this paper the application of recurrent neural networks to model dynamic hysteresis was investigated. An Elman neural network was selected to test the concept. The network was trained with a set of measured dynamic data and was tested with another set of experimental data, showing acceptable accuracy.

DOI: 10.3233/JAE-2002-323

Journal: International Journal of Applied Electromagnetics and Mechanics, vol. 13, no. 1-4, pp. 209-214, 2002