Core Loss Modeling of Fe-Based Amorphous and Nanocrystalline Magnetic Materials for Developing High-Performance EV Motors and Electromagnetic Devices

Abstract

Abstract

Fe-based amorphous and nanocrystalline alloys are attractive candidate core materials for high-performance electric vehicle (EV) motors and other electromagnetic devices because their low core loss, high saturation flux density, and high electrical resistivity can support high-efficiency, high-power-density electromagnetic designs. Their application, however, requires reliable loss prediction under the complex excitation found in electrical machines and other electromagnetic devices, including wide frequency variation, non-sinusoidal waveforms, temperature changes, local vectorial or rotational magnetization, and other coupled operating effects. Classical approaches, including the Steinmetz equation family, Bertotti-type loss separation, the Jiles–Atherton model, and finite element methods, remain important but do not provide a unified treatment of these coupled effects. Data-driven models improve nonlinear prediction but are limited by the scarcity of material-specific datasets and poor physical interpretability. Physics-informed neural networks (PINNs) can embed governing magnetic-loss relationships as training constraints, while Kolmogorov–Arnold networks (KANs) offer potential for interpretable symbolic regression. This review compares these modeling generations from the perspective of EV motor and electromagnetic device design, identifies the main gaps in magnetic material-relevant datasets and multidimensional magnetization modeling, and discusses physics-informed KAN as a proposed direction that requires material-specific validation before application to core loss prediction and design optimization.