Scale-Decreasing Surrogate Modeling for Low-Deviation Optimization of Axial Flux Motors Considering Parameter Coupling
Lists
Only you see this
Abstract
In this article, the reduced order model (ROM) assisted iterative scale-decreasing two-tier surrogate method is proposed for multi-objective optimization (MOO) of axial flux motors (AFMs). It can consider parameter interaction and realize a low-level deviation during the optimization process. Firstly, the ROM is applied to construct an accurate dataset for surrogate training. Subsequently, the two-tier surrogate is employed to accelerate the MOO process effectively. The first-tier surrogate model is trained using data generated from the ROM, enabling the construction of a rapid optimization framework. Next, a second-tier scale-decreasing surrogate model is established. This second-tier model improves the approximation precision with a more comprehensive and refined sampling space after considering the parameter coupling effect. Therefore, based on the studied machine, the proposed method achieves: (1) Reduction of the optimal set deviation error from ∼2% to ∼0.2% compared to the conventional surrogate method, with finite element method (FEM)-based optimization as the benchmark. (2) Significantly saves the computational burden compared with the high-fidelity FEM. Ultimately, the effectiveness of the scale-decreasing two-tier surrogate method is validated through FEM and experiments. The optimal design demonstrates that, compared to the conventional method, the proposed method achieves improved optimization accuracy, which further validates its potential for advanced motor design and optimization applications.
BibTeX
@article{su2025scale,
author = {Xiangdong Su and Liyang Liu and Song Guo and Zhenxiao Yin and Yongcan Huang and Hang Zhao},
title = {{Scale-Decreasing Surrogate Modeling for Low-Deviation Optimization of Axial Flux Motors Considering Parameter Coupling}},
journal = {IEEE Transactions on Energy Conversion},
year = {2025},
doi = {10.1109/tec.2025.3620349},
}Loading related content...