Accounting for Hysteresis and Eddy Currents in Finite Element Simulations of Ferromagnetic Laminated Cores using a Recurrent Neural Network

arXiv cs.AI · 6 hours ago

Researchers developed a recurrent neural network trained to approximate the behavior of laminated ferromagnetic cores during finite element simulations, enabling accurate electromagnetic modeling while reducing computational cost to about twice that of simplified simulations. The surrogate model was trained on artificially generated magnetic field sequences and integrated into two-dimensional magnetodynamic finite element simulations using magnetic vector potential formulation. This approach provides engineers with a practical tool for designing electrical machines without requiring the prohibitively expensive full simulations that resolve fields at every point and iteration.