The paper describes an approach to the problem of magnetic hysteresis identification based on the use of neural networks. The problem for many of the models currently being proposed for representing hysteresis in analytical systems is the identification of the model, i.e. the reliable determination of the appropriate parameters for a particular material. Some work has been published in the use of neural networks in this area and this paper seeks to extend this approach. In addition, the networks being considered are structured in such a manner that the information acquired by the network may be extracted at the end of the learning process - thus providing explicit values of the parameters, if needed. The basic methodologies being considered are based on the Preisach models and neural networks. Two techniques structured around the use of radial basis functions and a CMAC architecture are examined. The CMAC paradigm can allow incremental training and thus the retraining of the network for a new hysteresis curve can take place relatively quickly. Once the parameters have been derived, the model can be used within a conventional finite element analysis system.
%0 Journal Article
%1 Saliah1997neural
%A Saliah-Hassane,
%A Lowther, David A.
%D 1997
%E Elsiever,
%J Physica B Condensed Matter
%K Curve Magnetization Material Micromagnetism Modelling Networks Neural Preisach model
%N 4
%P 318-323
%T The use of neural networks in magnetic hysteresis identification
%U http://www.sciencedirect.com/science?_ob=ArticleURL&_udi=B6TVH-3SPM8PK-8&_user=10&_rdoc=1&_fmt=&_orig=search&_sort=d&_docanchor=&view=c&_searchStrId=1098099382&_rerunOrigin=scholar.google&_acct=C000050221&_version=1&_urlVersion=0&_userid=10&md5=0fb0b654aa44b29f9a15fd9e00a9b451
%V 233
%X The paper describes an approach to the problem of magnetic hysteresis identification based on the use of neural networks. The problem for many of the models currently being proposed for representing hysteresis in analytical systems is the identification of the model, i.e. the reliable determination of the appropriate parameters for a particular material. Some work has been published in the use of neural networks in this area and this paper seeks to extend this approach. In addition, the networks being considered are structured in such a manner that the information acquired by the network may be extracted at the end of the learning process - thus providing explicit values of the parameters, if needed. The basic methodologies being considered are based on the Preisach models and neural networks. Two techniques structured around the use of radial basis functions and a CMAC architecture are examined. The CMAC paradigm can allow incremental training and thus the retraining of the network for a new hysteresis curve can take place relatively quickly. Once the parameters have been derived, the model can be used within a conventional finite element analysis system.
@article{Saliah1997neural,
abstract = {The paper describes an approach to the problem of magnetic hysteresis identification based on the use of neural networks. The problem for many of the models currently being proposed for representing hysteresis in analytical systems is the identification of the model, i.e. the reliable determination of the appropriate parameters for a particular material. Some work has been published in the use of neural networks in this area and this paper seeks to extend this approach. In addition, the networks being considered are structured in such a manner that the information acquired by the network may be extracted at the end of the learning process - thus providing explicit values of the parameters, if needed. The basic methodologies being considered are based on the Preisach models and neural networks. Two techniques structured around the use of radial basis functions and a CMAC architecture are examined. The CMAC paradigm can allow incremental training and thus the retraining of the network for a new hysteresis curve can take place relatively quickly. Once the parameters have been derived, the model can be used within a conventional finite element analysis system.
},
added-at = {2009-11-18T05:44:46.000+0100},
author = {Saliah-Hassane and Lowther, David A.},
biburl = {https://www.bibsonomy.org/bibtex/225972ff277fe588f7b79fc2078dd49a9/saliah},
editor = {Elsiever},
interhash = {e16659d0c36f6314e6dfe64d5adad4ef},
intrahash = {25972ff277fe588f7b79fc2078dd49a9},
journal = {Physica B Condensed Matter},
keywords = {Curve Magnetization Material Micromagnetism Modelling Networks Neural Preisach model},
number = 4,
pages = {318-323},
timestamp = {2009-11-18T05:44:47.000+0100},
title = {The use of neural networks in magnetic hysteresis identification },
url = {http://www.sciencedirect.com/science?_ob=ArticleURL&_udi=B6TVH-3SPM8PK-8&_user=10&_rdoc=1&_fmt=&_orig=search&_sort=d&_docanchor=&view=c&_searchStrId=1098099382&_rerunOrigin=scholar.google&_acct=C000050221&_version=1&_urlVersion=0&_userid=10&md5=0fb0b654aa44b29f9a15fd9e00a9b451},
volume = 233,
year = 1997
}