EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Residual Physics-Informed Neural Networks for Seismic Tomography with Multi-Source Prior Constraints

Yanjin Xiang, Ziang Song, Zhiliang Wang, Guojie Song

Mathematics · 21.04.2026 · 14 (8) · 1392

DOI: 10.3390/math14081392 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Mathematics, online erschienen am 21.04.2026. Verfasst von Yanjin Xiang und 3 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 45 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Special Project of the Central Government to Guide Local Science and Technology Development, Nanchong City University Science and Technology Strategic Cooperation Project, the scientific research starting project of SWPU.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Seismic traveltime tomography is essential for constructing subsurface velocity models that underpin high-resolution imaging and inversion. Traditional ray- and eikonal-based methods are sensitive to the starting model and lack a unified, physically consistent framework to integrate seismic data with high-confidence prior information. PINN-based approaches offer flexible, grid-free inversion but often suffer from training instability and limited use of prior constraints. We propose a Multi-Source Prior-Guided Residual Physics-Informed Neural Network (MSP-ResPINNs) to address these limitations through two key technical advancements. First, MSP-ResPINNs integrates a Residual Network (ResNet) architecture with Sinusoidal Representation (SIREN) activation to replace standard MLPs, ensuring the robust capture of high-frequency velocity gradients. Second, the framework implements a unified loss function that rigorously enforces multi-source constraints, including well-logs and geological horizons. Numerical experiments demonstrate that MSP-ResPINNs accurately reconstructs sharp velocity contrasts and complex geological features compared with conventional PINN-based approaches. Among the tested variants, the multiplicative factorization consistently provides the most stable and physically consistent results, outperforming the additive factorization.

Förderung

  • Special Project of the Central Government to Guide Local Science and Technology Development
  • Nanchong City University Science and Technology Strategic Cooperation Project
  • the scientific research starting project of SWPU
  • Southwest Petroleum University Graduate Research Innovation Fund

Zitieren

APA

Xiang, Y., Song, Z., Wang, Z., & Song, G. (2026). Residual Physics-Informed Neural Networks for Seismic Tomography with Multi-Source Prior Constraints. Mathematics, 14(8), 1392. https://doi.org/10.3390/math14081392

Vancouver

Xiang Y, Song Z, Wang Z, Song G. Residual Physics-Informed Neural Networks for Seismic Tomography with Multi-Source Prior Constraints. Mathematics. 2026;14(8):1392. doi:10.3390/math14081392

BibTeX
@article{Xiang2026,
  author = {Xiang, Yanjin and Song, Ziang and Wang, Zhiliang and Song, Guojie},
  title = {Residual Physics-Informed Neural Networks for Seismic Tomography with Multi-Source Prior Constraints},
  journal = {Mathematics},
  year = {2026},
  volume = {14},
  number = {8},
  pages = {1392},
  doi = {10.3390/math14081392}
}
Kostenlos starten →