Answers are grounded in this paper and cite their sources.
The service environment of long span bridges is inherently complex, with localized wind fields exhibiting pronounced spatio-temporal variability and strong fluid-structure interactions. In practical bridge health monitoring, sparse deployment of wind sensors necessitates estimation of spatially distributed wind fields from limited observations. Therefore, this paper proposes a physics-informed deep learning framework to estimate wind sequences at dense virtual points. By establishing a fusion model termed Spatio-Temporal estimation with Attention-based Refinement, the proposed framework integrates Convolutional Encoder-Decoder Networks as the backbone module and Geometry-Enriched Attention Networks as the refinement module. Validation on both numerical simulation and real monitoring data demonstrates that the framework achieves accurate wind field estimation and missing data imputation. Compared with the ablation frameworks, the attention-based refinement module and physics-informed constraints reduce estimation errors by 68.0% and 16.6%, respectively. These results demonstrate the effectiveness of dense virtual sensing for bridge wind field estimation under sparse sensor conditions.
No takes yet. Share an insight, caveat, or question.
Liu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: