EUVIMEDEUVIMED entdecken
Crossref

Distance-based collocation sampling for mesh-free physics-informed neural networks

Hyo Seung Lee, Sang Joon Lee

Physics of Fluids · 01.07.2025 · 37 (7)

DOI: 10.1063/5.0279565 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Physics of Fluids, online erschienen am 30.07.2025. Verfasst von Hyo Seung Lee und 1 weiteren Autor:innen. Die Arbeit wurde bisher 6-mal zitiert (Crossref) und nennt 59 Literaturangaben. Gefördert von: National Research Foundation of Korea.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Physics-informed neural networks (PINNs) have emerged as a powerful mesh-free framework for solving partial differential equations by embedding physical laws into the loss function of deep learning models. Their flexibility, particularly in data-scarce or noisy environments, has enabled a wide range of applications in computational physics and engineering. However, one of the persistent challenges in training PINNs lies in the accurate enforcement of boundary conditions, especially for domains with irregular geometries. Conventional collocation point sampling methods, including mesh-based and uniform random strategies, pose limitations in the context of PINNs. The mesh-based sampling strategy contradicts the mesh-free feature of PINNs, while the uniform sampling makes it a challenge to account for the geometric structure of the domain. A geometry-aware, distance-based sampling (DBS) strategy is proposed in this study to select collocation points using a signed distance function (SDF). The SDF provides geometric information by which the structured sampling is aligned with the domain shape, without requiring explicit meshing. SDF information is used to promote selective sampling in the region near domain boundaries based on the concept of proximity. This proposed approach facilitates the treatment of boundary conditions while fully preserving the mesh-free nature of PINNs. Furthermore, the proposed DBS strategy provides high flexibility in the selection of collocation points and supports the incorporation of diverse sampling methods.

Förderung

  • National Research Foundation of Korea

Zitieren

APA

Lee, H. S., & Lee, S. J. (2025). Distance-based collocation sampling for mesh-free physics-informed neural networks. Physics of Fluids, 37(7). https://doi.org/10.1063/5.0279565

Vancouver

Lee HS, Lee SJ. Distance-based collocation sampling for mesh-free physics-informed neural networks. Physics of Fluids. 2025;37(7). doi:10.1063/5.0279565

BibTeX
@article{Lee2025,
  author = {Lee, Hyo Seung and Lee, Sang Joon},
  title = {Distance-based collocation sampling for mesh-free physics-informed neural networks},
  journal = {Physics of Fluids},
  year = {2025},
  volume = {37},
  number = {7},
  doi = {10.1063/5.0279565}
}
Kostenlos starten →