EUVIMEDEUVIMED entdecken
Crossref

Multifidelity Space-Mapping-Based Approach for Accelerated Multiphysics Optimization of Microwave Devices

Xiao Yang, Zimeng Wang, Haitian Hu, Wei Zhang, Feng Feng, Qi-Jun Zhang

IEEE Transactions on Microwave Theory and Techniques · 2025-12 · 73 (12) · 9902-9919

DOI: 10.1109/tmtt.2025.3627935 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Microwave Theory and Techniques, 2025-12. Verfasst von Xiao Yang und 5 weiteren Autor:innen. Die Arbeit wurde bisher 5-mal zitiert (Crossref) und nennt 59 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Fund of State Key Laboratory of Information Photonics and Optical Communications (IPOC) [Beijing University of Posts and Telecommunications (BUPT)].

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This article proposes an advanced multifidelity space-mapping-based approach to accelerate the multiphysics optimization of microwave devices. The proposed method develops a novel two-layer space mapping (SM) framework that strategically integrates two complementary mappings. The first-layer mapping bridges high-fidelity multiphysics analysis (multiphysics fine model) with single-physics electromagnetic (EM) analysis (EM fine model), while the second-layer mapping further leverages this EM fine model to a computationally inexpensive EM coarse model constructed on a coarse mesh (EM coarse model). To effectively capture the complex nonlinear relationships between different fidelity models, artificial neural networks (ANNs) are employed to construct both mappings, enabling efficient fidelity transitions and preserving surrogate model accuracy. A trust-region optimization algorithm, tailored for this multifidelity SM framework, is developed to guarantee stability and convergence efficiency during the optimization process. This algorithm updates the optimization step size based on the local credibility of the surrogate model, effectively guiding the design toward a robust convergence with limited high-fidelity data. By utilizing the two-layer neural network mapping and the custom optimization strategy, our method systematically shifts computationally intensive optimization tasks to the lowest-fidelity model. This significantly reduces the reliance on expensive multiphysics simulations. Validation through three diverse design examples, including waveguide filters and a tunable antenna, demonstrates that the proposed method excels in accelerating multiphysics optimization. The framework effectively reduces the reliance on expensive multiphysics simulations and shortens the total optimization time compared to conventional surrogate modeling approaches, providing a powerful tool for the efficient design of complex microwave devices.

Förderung

  • National Natural Science Foundation of China
  • Fund of State Key Laboratory of Information Photonics and Optical Communications (IPOC) [Beijing University of Posts and Telecommunications (BUPT)]

Zitieren

APA

Yang, X., Wang, Z., Hu, H., Zhang, W., Feng, F., & Zhang, Q. (2025). Multifidelity Space-Mapping-Based Approach for Accelerated Multiphysics Optimization of Microwave Devices. IEEE Transactions on Microwave Theory and Techniques, 73(12), 9902-9919. https://doi.org/10.1109/tmtt.2025.3627935

Vancouver

Yang X, Wang Z, Hu H, Zhang W, Feng F, Zhang Q. Multifidelity Space-Mapping-Based Approach for Accelerated Multiphysics Optimization of Microwave Devices. IEEE Transactions on Microwave Theory and Techniques. 2025;73(12):9902-9919. doi:10.1109/tmtt.2025.3627935

BibTeX
@article{Yang2025,
  author = {Yang, Xiao and Wang, Zimeng and Hu, Haitian and Zhang, Wei and Feng, Feng and Zhang, Qi-Jun},
  title = {Multifidelity Space-Mapping-Based Approach for Accelerated Multiphysics Optimization of Microwave Devices},
  journal = {IEEE Transactions on Microwave Theory and Techniques},
  year = {2025},
  volume = {73},
  number = {12},
  pages = {9902--9919},
  doi = {10.1109/tmtt.2025.3627935}
}
Kostenlos starten →