Abstract:We present a general framework for modeling power magnetic materials characteristics using deep neural networks. Magnetic materials represented by multidimensional characteristics (that mimic measurements) are used to train the neural autoencoder model in an unsupervised manner. The encoder is trying to predict the material parameters of a theoretical model, which is then used in a decoder part. The decoder, using the predicted parameters, reconstructs the input characteristics. The neural model is trained to capture a synthetically generated set of characteristics that can cover a broad range of material behaviors, leading to a model that can generalize on the underlying physics rather than just optimize the model parameters for a single measurement. After setting up the model, we prove its usefulness in the complex problem of modeling magnetic materials in the frequency and current (out-of-linear range) domains simultaneously, for which we use measured characteristics obtained for frequency up to $10$ MHz and H-field up to saturation.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the challenge regarding the modeling of the characteristics of electrical magnetic materials. Specifically, the author proposes a general framework using deep neural networks (DNNs) to simulate the characteristics of electrical magnetic materials. These problems include: 1. **Limitations of Existing Models**: Currently, the first - principle models for electrical magnetic materials are not fully satisfactory yet, and the existing models usually need to optimize parameters through methods such as curve fitting, which has limitations when dealing with complex material characteristics. 2. **Modeling of Multidimensional Characteristics**: The behavior of electrical magnetic materials in different frequency and current ranges is very complex, especially in the nonlinear region (such as the saturation state). Traditional modeling methods are difficult to capture these multidimensional characteristics simultaneously. 3. **Data - Driven Modeling**: With the development of big data and deep learning, data - driven methods have become a new trend and can model complex physical phenomena more effectively. This paper aims to use deep neural networks to learn the characteristics of magnetic materials from a large amount of synthetic data and be able to generalize to unseen data. 4. **Accuracy of Parameter Prediction**: How to accurately predict the parameters of theoretical models so that the models can not only fit a single measurement data but also generalize to the material characteristics under different conditions. To solve these problems, the author proposes a deep neural network model based on autoencoder, which can learn the multidimensional characteristics of magnetic materials in an unsupervised manner and can predict the parameters of theoretical models. In this way, the model can better capture the physical characteristics of materials, rather than just optimizing the parameters of a single measurement data.

Machine-learned models for magnetic materials

How MagNet: Machine Learning Framework for Modeling Power Magnetic Material Characteristics

PINN Based Data Driven Magnetics Loss Modeling

Deep learning model for enhanced power loss prediction in the frequency domain for magnetic materials

Machine Learning in Magnetic Materials

Machine learning based classification of vector field configurations

Image-based prediction and optimization of hysteresis properties of nanocrystalline permanent magnets using deep learning

Magnetic Hysteresis Modeling with Neural Operators

Description of collective magnetization processes with machine learning models

Neural Network Architectures and Magnetic Hysteresis: Overview and Comparisons

Design of a Novel Explainable Adversarial Autoencoder Model for the Electromagnetic Analysis of Functional Materials Based on Physics-Informed Learning

Physics-informed machine learning combining experiment and simulation for the design of neodymium-iron-boron permanent magnets with reduced critical-elements content

Advancing magnetic material discovery through machine learning: Unveiling new manganese-based materials

MagNet: machine learning enhanced three-dimensional magnetic reconstruction

Multifrequency nonlinear model of magnetic material with artificial intelligence optimization

Magnetic Hamiltonian Parameter Estimation Using Deep Learning Techniques

Feature Engineering Mediated Machine Learning Prediction of High-Frequency Property of Soft Magnetic Thin Films

NeuralMAG: Fast and Generalizable Micromagnetic Simulation with Deep Neural Nets

Physics-Informed Sparse Neural Network for Permanent Magnet Eddy Current Device Modeling and Analysis

Equivariant Neural Network Force Fields for Magnetic Materials

Nonlinear modeling of magnetic materials for circuit simulations