HARDCORE: H-Field and Power Loss Estimation for Arbitrary Waveforms With Residual, Dilated Convolutional Neural Networks in Ferrite Cores

The MagNet challenge 2023 called upon competitors to develop data-driven models for the material-specific, waveform-agnostic estimation of steady-state power losses in toroidal ferrite cores. The following HARDCORE (H-field and power loss estimation for arbitrary waveforms with residual, dilated convolutional neural networks in ferrite cores) approach shows that a residual convolutional neural network with physics-informed extensions can serve this task efficiently when trained on observational data beforehand. One key element is an intermediate model layer, which first reconstructs the $BH$ curve, and then, estimates the power losses based on the curve's area rendering the proposed topology physically interpretable. In addition, emphasis was placed on expert-based feature engineering and information-rich inputs in order to enable a lean model architecture. A model is trained from scratch for each material, while the topology remains the same. A Pareto-style tradeoff between model size and estimation accuracy is demonstrated, which yields an optimum at as low as 906 parameters and down to below 8% for the average 95th percentile of the relative power loss error across diverse materials. This contribution has won the first place in the performance category of the MagNet challenge 2023, which further highlights the effectiveness of the proposed model.