Neural Network Modeling of Arbitrary Hysteresis Processes: Application to GO Ferromagnetic Steel

Simone Quondam Antonio,Vincenzo Bonaiuto,Alessandro Salvini,Fausto Sargeni

doi:10.3390/magnetochemistry8020018

Simone Quondam Antonio, Vincenzo Bonaiuto + Show 2 more

Open Access

https://doi.org/10.3390/magnetochemistry8020018

Copy DOI

Abstract

A computationally efficient hysteresis model, based on a standalone deep neural network, with the capability of reproducing the evolution of the magnetization under arbitrary excitations, is here presented and applied in the simulation of a commercial grain-oriented electrical steel sheet. The main novelty of the proposed approach is to embed the past history dependence, typical of hysteretic materials, in the neural net, and to illustrate an optimized training procedure. Firstly, an experimental investigation was carried out on a sample of commercial GO steel by means of an Epstein equipment, in agreement with the international standard. Then, the traditional Preisach model, identified only using three measured symmetric hysteresis loops, was exploited to generate the training set. Once the network was trained, it was validated with the reproduction of the other measured hysteresis loops and further hysteresis processes obtained by the Preisach simulations. The model implementation at a low level of abstraction shows a very high computational speed and minimal memory allocation, allowing a possible coupling with finite-element analysis (FEA).

Highlights

The optimum design of electrical machines and magnetic components requires the use of accurate and performing macroscale hysteresis models, capable of reproducing the evolution of the material magnetization and the prediction of the power losses, under real working conditions
We propose a hysteresis model, with magnetic field H (A/m) as the input and magnetic induction B (T) as the output, based on a standalone multilayer feedforward neural network, and we apply it to simulate the magnetization processes of a commercial grain-oriented (GO) electrical steel sheet
Since the training set for the neural network (NN)-based model was generated by the Preisach model (PM), with the aim of using the smallest possible amount of information on the material, let us begin the presentation of the results from the identification of the Preisach parameters

Summary

Introduction

The optimum design of electrical machines and magnetic components requires the use of accurate and performing macroscale hysteresis models, capable of reproducing the evolution of the material magnetization and the prediction of the power losses, under real working conditions. The most important effects of the harmonics and the DC bias, which negatively influence the performance of the devices, are extra power losses, vibrations, noise, and local heating. Achieving such a detailed knowledge on the material behavior, starting from the limited information usually provided by the manufacturers, is a challenging issue, almost completely mandated to the capabilities of the material model.

Methods

Findings

Discussion

Conclusion