MagLearn - Data-driven Machine Learning Framework with Transfer and Few-shot Training for Modeling Magnetic Core Loss

Aug 5, 2026

Upload by Violet

MagLearn presents a data-driven solution for predicting magnetic core loss in response to the MagNet Challenge 2023, addressing the difficulty of modeling loss without complete physics-based models. The method uses datasets from PowerLab Princeton to map flux density waveforms (sinusoidal, rectangular, trapezoidal) to core loss outputs. An LSTM extracts features from B-waveforms, while transfer learning and few-shot training handle small, imbalanced datasets via data augmentation and alignment. Random shift/flip decouples output from phase shift, and experimental test sets confirm high prediction accuracy.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages8

Reading Time20 min

AccessFree to download

Abstract

Introduction

Zhang, L. , McKeague, T. , Cui, B. , Rasekh, N. , Wang, J. , Liu, S. , & Martinez, A. (2024) . MagLearn – Data-driven Machine Learning Framework with Transfer and Few-shot Training for Modeling Magnetic Core Loss. In 2024 26th European Conference on Power Electronics and Applications (EPE'24 ECCE Europe) Institute of Electrical and Electronics Engineers (IEEE) . [https://doi.org/10.1109/ECCEEurope62508.2024.10751860](https://doi.org/10.1109/ECCEEurope62508.2024.10751860) Peer reviewed version License (if available): CC BY Link to published version (if available): 10.1109/ECCEEurope62508.2024.10751860 Link to publication record on the Bristol Research Portal PDF-document This is the accepted author manuscript (AAM) of the article which has been made Open Access under the University of Bristol's Scholarly Works Policy. The final published version (Version of Record) can be found on the publisher's website. The copyright of any third-party content, such as images, remains with the copyright holder. University of Bristol – Bristol Research Portal General rights This document is made available in accordance with publisher policies. Please cite only the published version using the reference above. Full terms of use are available: [http://www.bristol.ac.uk/red/research-policy/pure/user-guides/brp-terms/](http://www.bristol.ac.uk/red/research-policy/pure/user-guides/brp-terms/) MagLearn – Data-driven Machine Learning Framework with Transfer and Few-shot Training for Modeling Magnetic Core Loss Lizhong Zhang 1, Tom McKeague1, Binyu Cui1, Navid Rasekh 1, Jun Wang 1, Song Liu 1, Alfonso Martinez2 1 University of Bristol, UK 2 Wurth Electronik, Germany [Jun.Wang@bristol.ac.uk](Jun.Wang@bristol.ac.uk) Abstract—In response to the MagNet Challenge 2023, this paper describes the solution developed by the University of Bristol team, awarded the 3rd Place Outstanding Performance among 24 competing teams worldwide. The core loss of magnetic components has been a challenge for engineers to model due to the lack of full physics models. Classic Steinmetz-Equation-based approaches show significant limitations under power electronicsexcitations. Data-driven approaches have emerged in recent years as a new solution to this problem as an active research area. Based on the datasets supplied by PowerLab Princeton, this work employs a machine learning framework to predict the core loss of magnetic components from a range offlux density waveforms, e.g. sinusoidal, rectangular, trapezoidal, as the input. The proposed approach builds on an LSTM neural network to extract features from the input B waveforms and predict the power loss value. Designed for the small and imbalanced datasets supplied in the competition, a machine learning pipeline is proposed in this work featuring transfer learning and few-shot training, which is realized through data augmentation and alignment. As a modification to decouple the output from the phase shift of the input waveform, a random shift/flip algorithm is applied in both pre-and post-processing blocks. The performance of the proposed approach is validated through the experimentally measured testing sets, demonstrating a high prediction accuracy. Keywords—MagNet Challenge, machine learning, neural networks, magnetic core loss, power electronics I. INTRODUCTION Nowadays, magnetic components are involved in most power electronic converters for functionality and filtering purposes. They are typically known to be the least efficient components that have a significant impact on system performance in terms of efficiency and size/weight [1], [2] . However, an accurate core loss model for magnetic components that is based on the first principle remains elusive due to their nonlinearity and other intercoupled factors, such as dc-bias and waveform shapes. Numerous research studies have been carried out to factor in the external parameters contributing to magnetic loss under nonsinusoidal excitati

FAQs

Why is magnetic core loss modeling challenging in power electronics?

Core loss modeling is difficult because first-principles physics models are incomplete and magnetic behavior is strongly nonlinear, with coupled factors such as dc-bias and waveform shape.

How does MagLearn predict core loss from input waveforms?

MagLearn predicts core loss from flux density waveforms by using an LSTM neural network to extract features from the B-waveform input and output the corresponding power loss value.

What techniques does the framework use for small and imbalanced datasets?

It introduces a machine learning pipeline combining transfer learning and few-shot training, realized through data augmentation and alignment to improve learning under limited data.

Aug 5, 2026

Upload by Violet