EUVIMEDEUVIMED entdecken
Systematische Übersicht

AI-Driven Field Reconstruction of Structural Responses: A Systematic Review

Jian-Sheng Fan, Hao-Yang Zhao, Chen Wang

Civil Engineering Sciences · 01.01.2026 · 2

DOI: 10.34133/cesci.0019 ↗

Kurz gefasst

Systematische Übersicht in der Zeitschrift Civil Engineering Sciences, online erschienen am 10.06.2026. Verfasst von Jian-Sheng Fan und 2 weiteren Autor:innen. Die Arbeit nennt 87 Literaturangaben. Gefördert von: National Natural Science Foundation of China.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The advancement of smart infrastructure and high-fidelity digital twins in civil engineering relies critically on real-time, high-resolution measurements of structural and environmental fields. However, such measurements are often limited by sparse sensor deployments. Artificial intelligence (AI)-based full-field reconstruction has thus emerged as a pivotal technology for inferring dense physical fields from limited observations, enabling real-time analysis of complex engineering systems. This review systematically surveys AI-driven reconstruction methods, organizing the literature into 2 principal paradigms: data-driven models, which directly learn sparse-to-dense mappings using architectures such as graph neural networks and transformers, and theory-data-fused models, which integrate physical knowledge into learning frameworks, including physics-informed neural networks, AI-assisted reduced-order models, and neural operators. Quantitative bibliometric analysis and qualitative comparison highlight a fundamental trade-off: Data-driven approaches offer high flexibility and excel at capturing complex nonlinear behavior when ample, high-quality data are available, whereas theory-data-fused approaches ensure greater physical consistency and data efficiency, making them particularly suitable for safety-critical applications with sparse or noisy measurements. The current methodological landscape is defined by problem-driven model selection and a growing emphasis on generalization, robustness, and interpretability, supported by advances in transfer learning, operator learning, and physics-regularized generative modeling. Future developments are expected to converge toward hybrid frameworks that merge the expressiveness of data-driven models with the reliability of physics-based constraints, and toward the integration of such models as core computational engines within large-scale digital twins and operational civil infrastructure systems.

Förderung

  • National Natural Science Foundation of China

Zitieren

APA

Fan, J., Zhao, H., & Wang, C. (2026). AI-Driven Field Reconstruction of Structural Responses: A Systematic Review. Civil Engineering Sciences, 2. https://doi.org/10.34133/cesci.0019

Vancouver

Fan J, Zhao H, Wang C. AI-Driven Field Reconstruction of Structural Responses: A Systematic Review. Civil Engineering Sciences. 2026;2. doi:10.34133/cesci.0019

BibTeX
@article{Fan2026,
  author = {Fan, Jian-Sheng and Zhao, Hao-Yang and Wang, Chen},
  title = {AI-Driven Field Reconstruction of Structural Responses: A Systematic Review},
  journal = {Civil Engineering Sciences},
  year = {2026},
  volume = {2},
  doi = {10.34133/cesci.0019}
}
Kostenlos starten →