Collection of different data-driven algorithms for modeling the B-H characteristics of magnetic materials
This repo contains some of the work performed at CompEM lab - McGill University. Based on the following paper: IEEE paper link
- pdf available (contact authors - arbaaz.khan@mail.mcgill.ca)
- A private copy available on ResearchGate
Cesay S, Teng P, Wang R, Yue H, Khan A, Lowther D. Generalizable DNN based multi-material Hysteresis Modelling. In 2022 IEEE 20th Biennial Conference on Electromagnetic Field Computation (CEFC) 2022 Oct 24 (pp. 1-2). IEEE.
Summary: Over the last twenty years, the reliance on magnetic materials has surged significantly. Each type of magnetic material exhibits unique behaviors, making it challenging to develop a universal model. Existing models and simulation tools can represent these behaviors, but they often require days or even a week to compute within an analysis system, which is highly inefficient. A well-designed machine learning (ML) model can drastically reduce computation time with minimal error. The hysteretic behavior of the material is crucial for determining performance parameters like the efficiency of an electrical machine, and the chosen representation can affect the performance of a simulation tool. This work aims to investigate various deep learning (DL) network architectures to provide a generalized representation for hysteresis problems and reduce the computational effort needed for finite element-based simulations compared to current methods.
The citation for the work:
@inproceedings{cesay2022generalizable,
title={Generalizable DNN based multi-material Hysteresis Modelling},
author={Cesay, Saikou and Teng, Paul and Wang, Ruoli and Yue, Haupeng and Khan, Arbaaz and Lowther, David},
booktitle={2022 IEEE 20th Biennial Conference on Electromagnetic Field Computation (CEFC)},
pages={1--2},
year={2022},
organization={IEEE}
}