EUVIMEDEUVIMED entdecken
Crossref

Physics-informed neural network modeling of a bar under pure axial tension: an efficient framework

Loan Bùi Thị, Việt Trần Bảo, Quân Nguyễn Hoàng, Thủy Phạm Thị Thanh, Mai Bùi Thị Thanh

Transport and Communications Science Journal · 14.07.2025 · 76 (8) · 1108-1122

DOI: 10.47869/tcsj.76.8.6 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Transport and Communications Science Journal, online erschienen am 14.07.2025. Verfasst von Loan Bùi Thị und 4 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

In structural mechanics, solving partial differential equations (PDEs) is traditionally performed using numerical methods such as the Finite Element Method (FEM). However, when applied to optimization problems, FEM often encounters limitations due to its high computational cost and difficulties in handling continuously varying geometries. Recently, Physics-Informed Neural Networks (PINNs) have emerged as a promising alternative, enabling the solution of PDEs without mesh generation by directly embedding physical laws into the loss function of deep learning models. This study initially develops a PINN model for the uniaxial tension bar problem, with the aim of evaluating the effectiveness of three commonly used training techniques: non-dimensionalization of input and output variables; the use of hard versus soft boundary conditions; and the choice between PDE-based and energy-based loss functions. Results show that non-dimensionalization improves accuracy by balancing gradients across physical quantities; hard boundary conditions lead to faster and more stable convergence; and PDE-based loss functions enforce local physical constraints more effectively than energy-based ones. This work represents only an initial study on a simple problem; nevertheless, the obtained results demonstrate the potential of PINNs in structural mechanics and provide a foundation for further development and extension to more complex problems

Zitieren

APA

Thị, L. B., Bảo, V. T., Hoàng, Q. N., Thanh, T. P. T., & Thanh, M. B. T. (2025). Physics-informed neural network modeling of a bar under pure axial tension: an efficient framework. Transport and Communications Science Journal, 76(8), 1108-1122. https://doi.org/10.47869/tcsj.76.8.6

Vancouver

Thị LB, Bảo VT, Hoàng QN, Thanh TPT, Thanh MBT. Physics-informed neural network modeling of a bar under pure axial tension: an efficient framework. Transport and Communications Science Journal. 2025;76(8):1108-1122. doi:10.47869/tcsj.76.8.6

BibTeX
@article{Th2025,
  author = {Thị, Loan Bùi and Bảo, Việt Trần and Hoàng, Quân Nguyễn and Thanh, Thủy Phạm Thị and Thanh, Mai Bùi Thị},
  title = {Physics-informed neural network modeling of a bar under pure axial tension: an efficient framework},
  journal = {Transport and Communications Science Journal},
  year = {2025},
  volume = {76},
  number = {8},
  pages = {1108--1122},
  doi = {10.47869/tcsj.76.8.6}
}
Kostenlos starten →