MagNet: Equation-based Baseline Models

Introduction

This repository contains the slides and code related to the following webinar:

• 2023 MagNet Challenge Webinar: Equation-based Baseline Models
• IEEE PELS Webinar - May 12 2023
• Thomas Guillod - Dartmouth College

This webinar focuses on equation-based loss models for soft-magnetic materials:

• Several models are presented (SE, iGSE, ISE, iGCC, and Stenglein equation).
• The model performances are evaluated for different frequencies, waveshapes, and temperatures.
• The advantages and drawbacks of equation-based models and machine learning models are discussed.
• A MATLAB implementation of the iGSE and iGCC is discussed in detail and the pitfalls are highlighted.

Main Files

• slides.pdf - Slides of the webinar
• paper.pdf - APEC paper introducing the iGCC
• run_igse.m - Parametrize and evaluate the iGSE model
• run_igcc.m - Parametrize and evaluate the iGCC model

Dataset

• For the software implementation, the EPCOS/TDK N87 ferrite material is considered.
• The material is measured at ambient temperature (25C) without DC bias.
• For parametrizing the models, the following dataset is used:
  • 346 symmetric triangular waveforms (50% duty cycle)
  • Dataset contained in N87_25C_fit.mat
• For evaluating the models, the following dataset is used:
  • 2446 asymmetric triangular waveforms (10% to 90% duty cycle)
  • Dataset contained in N87_25C_eval.mat
• Both datasets are extracted from the following repository:
  • Guillod, T. and Lee, J. S. and Li, H. and Wang, S. and Chen, M. and Sullivan, C. R.
  • Calculation of Ferrite Core Losses with Arbitrary Waveforms using the Composite Waveform Hypothesis: Reproducibility Dataset
  • Zenodo Repository, 2022
  • 10.5281/zenodo.7368936

Warnings

Warning This implementation is provided for pedagogical purposes:

• The goal of this code is to highlight the typical workflow of equation-based loss models.
• The implementation is not meant to be comprehensive and/or accurate.

Warning In order to limit the complexity of the code, several assumptions are made:

• Single material measured at ambient temperature
• Only triangular signals are considered
• No DC bias and relaxation effects
• Simple model parametrization
• Reduced dataset size

Compatibility

• Tested with MATLAB R2021a and R2023a.
• The optimization_toolbox is required.
• The signal_toolbox is required.
• The statistics_toolbox is required.

Author

Thomas Guillod - GitHub Profile

License

This project is licensed under the MIT License, see LICENSE.md.