EUVIMEDEUVIMED entdecken
Crossref

Physics-informed neural networks: peakon solutions for the generalized two-component rotational b-family system

Yunfei Yue, Shangwu Li, Lili Huang

Physica Scripta · 26.06.2026 · 101 (25) · 255213

DOI: 10.1088/1402-4896/ae7bce ↗

Kurz gefasst

Fachartikel in der Zeitschrift Physica Scripta, online erschienen am 25.06.2026. Verfasst von Yunfei Yue und 2 weiteren Autor:innen. Die Arbeit nennt 48 Literaturangaben. Gefördert von: Natural Science Foundation of Chongqing, National Natural Science Foundation of China, Scientific Research Foundation of Chongqing Normal University.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

In the present study, we investigate the peakon solutions for a generalized two-component shallow-water system, which could be used to describe the propagation of moderate-amplitude equatorial waves with the effect of Coriolis force due to the Earth’s rotation. Through distribution theory, we first construct the single and generalized peakon solutions for the generalized two-component rotational b-family system, and then the influence of parameters on the dynamic characteristics of the obtained solutions are analyzed in detail. Additionally, this paper employs the physics-informed neural networks (PINNs) method to perform numerical simulations of peakon solutions, enhancing their accuracy and stability by designing various loss function weighting schemes. To address gradient explosion issues during training generalized peakon solutions, we introduce normalization techniques for training datasets and refine the weight distribution scheme in the loss function. The training results demonstrate that this strategy reduces relative L 2 error to the 10 − 3 order of magnitude while improving both convergence speed and prediction accuracy. Finally, varying levels of Gaussian noise are introduced to verify the applicability and stability of PINNs for solving the peakon solutions.

Förderung

  • Natural Science Foundation of Chongqing
  • National Natural Science Foundation of China
  • Scientific Research Foundation of Chongqing Normal University
  • Science and Technology Research Program of Chongqing Municipal Education Commission

Zitieren

APA

Yue, Y., Li, S., & Huang, L. (2026). Physics-informed neural networks: peakon solutions for the generalized two-component rotational b-family system. Physica Scripta, 101(25), 255213. https://doi.org/10.1088/1402-4896/ae7bce

Vancouver

Yue Y, Li S, Huang L. Physics-informed neural networks: peakon solutions for the generalized two-component rotational b-family system. Physica Scripta. 2026;101(25):255213. doi:10.1088/1402-4896/ae7bce

BibTeX
@article{Yue2026,
  author = {Yue, Yunfei and Li, Shangwu and Huang, Lili},
  title = {Physics-informed neural networks: peakon solutions for the generalized two-component rotational b-family system},
  journal = {Physica Scripta},
  year = {2026},
  volume = {101},
  number = {25},
  pages = {255213},
  doi = {10.1088/1402-4896/ae7bce}
}
Kostenlos starten →