Journal of
Magnetics

Open Access
Quarterly

  • pISSN : 1226-1750
  • eISSN : 2233-6656
Editor's Choice
Residual-Correction Multi-Fidelity Surrogate Model Assisted Optimization of an Axial Flux Permanent Magnet Motor Using Q-3D and 3D FEM In-Yeong Hwang, Dong-Kuk Lim * This study proposes a residual-correction (RC) multi-fidelity surrogate model consisting of two surrogate models. First, a quasi-three-dimensional finite element method (Q-3D) surrogate model is trained using many Q-3D samples to capture the overall performance trends. Second, a residual surrogate model is trained using small number of three-dimensional finite element method (3D FEM) samples to learn the difference between the Q-3D surrogate model predictions and 3D FEM results. The Q-3D and residual surrogate models are combined to accurately predict the 3D FEM results. The RC model is coupled with a genetic algorithm for optimization, and the optimized design is evaluated using 3D FEM. The average torque increases from 551.83 Nm for the initial model to 553.63 Nm for the optimized model. Meanwhile, the torque ripple decreases from 9.65 % to 6.66 %.

Current Issue - Vol.31 No.3

Journal of
Magnetics

Open Access
Quarterly

  • pISSN : 1226-1750
  • eISSN : 2233-6656

Highlight

Editorial Office

  • Tel. +82-2-3452-7363
  • Fax. +82-2-3452-7366
  • E-mail. office@magnetics.or.kr