Hybrid Finite Element Physics-Informed Neural Network for 3-D Electromagnetic Simulation of Microwave Devices
IEEE Transactions on Microwave Theory and Techniques · 2026-08 · 74 (8) · 7041-7053
DOI: 10.1109/tmtt.2026.3706906 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Microwave Theory and Techniques, 2026-08. Verfasst von Qing Nan Fan und 8 weiteren Autor:innen. Die Arbeit nennt 35 Literaturangaben. Gefördert von: National Natural Science Foundation of China.
Abstract (Originalsprache)
To address the critical bottlenecks of low computational efficiency and excessive memory consumption in the electromagnetic (EM) simulation of 3-D complex microwave devices with the conventional finite element method (FEM), a hybrid finite element physics-informed neural network (HFE-PINN) is proposed in this article for efficient and high-precision 3-D EM simulation. A cascaded framework combining finite element discretization and physics-informed neural networks (PINNs) is employed. The Helmholtz equation is discretized into a sparse linear system by the finite element discretization module. A distributed subnetwork structure is utilized in the PINN module, with the midpoint coordinates of edges as inputs and the predicted solution of the finite element matrix equation as outputs. The physics-constrained loss function is minimized via the L-BFGS optimization algorithm, so that the predicted solution is driven to converge quickly to the true solution of the EM governing equation. Furthermore, the convexity of the HFE-PINN loss function is rigorously proved theoretically, which ensures that the global optimal solution can be obtained by the L-BFGS algorithm. Numerical examples demonstrate that the proposed method achieves high-precision EM simulation of arbitrarily complex microwave devices without labeled data, and outperforms the conventional FEM in computational efficiency and memory usage.
Förderung
- National Natural Science Foundation of China
Zitieren
APA
Fan, Q. N., Jia, Z. L., Wu, B., Liu, Y., Li, X., Feng, F., Ren, Z., Zhang, J., & Zhang, H. H. (2026). Hybrid Finite Element Physics-Informed Neural Network for 3-D Electromagnetic Simulation of Microwave Devices. IEEE Transactions on Microwave Theory and Techniques, 74(8), 7041-7053. https://doi.org/10.1109/tmtt.2026.3706906
Vancouver
Fan QN, Jia ZL, Wu B, Liu Y, Li X, Feng F, et al. Hybrid Finite Element Physics-Informed Neural Network for 3-D Electromagnetic Simulation of Microwave Devices. IEEE Transactions on Microwave Theory and Techniques. 2026;74(8):7041-7053. doi:10.1109/tmtt.2026.3706906
BibTeX
@article{Fan2026,
author = {Fan, Qing Nan and Jia, Zheng Lang and Wu, Bian and Liu, Ying and Li, Xiaolong and Feng, Feng and Ren, Zeyang and Zhang, Jincheng and Zhang, Huan Huan},
title = {Hybrid Finite Element Physics-Informed Neural Network for 3-D Electromagnetic Simulation of Microwave Devices},
journal = {IEEE Transactions on Microwave Theory and Techniques},
year = {2026},
volume = {74},
number = {8},
pages = {7041--7053},
doi = {10.1109/tmtt.2026.3706906}
}
EUVIMED