Fourier-based surrogate model of electromagnetic properties for accelerated time-domain simulation

Unbalanced magnetic pull (UMP) in electric motors is commonly analysed using finite element analysis (FEA). For parallel windings, circulating currents must be computed through time-domain integration of the electrical state equations, resulting in a large number of required evaluations of the FEA model and significant computational cost. This work proposes a Fourier coefficients-based surrogate modelling approach to accelerate the ODE-based simulations. For cases with a periodic mechanical state, the surrogate model represents the motional voltage vector and inductance matrix with Fourier coefficients identified from FEA data sampled over one period. Surrogate accuracy is assessed by comparing different numbers of sampling points, with a 500-point surrogate showing close agreement with FEA evaluations for both motional voltage and inductance. Finally, the surrogate model is applied within an ODE-based simulation, demonstrating successful computation of circulating currents with a converged solution. The method enables significant speed-up for simulations requiring very large number of FEA evaluations but is inherently limited to periodic mechanical orbits.