ABSTRACT
Accurate prediction of tunnel settlement during the operational phase is critical for developing effective preventive maintenance strategies, thereby prolonging the tunnel’s service life. However, existing prediction models often suffer from low accuracy or overfitting due to limited settlement data, neglect of spatial correlations between acquisition points, and insufficient integration of multi-source heterogeneous sensing data. To address these challenges, this study proposes a novel prediction model based on multi-source spatiotemporal data fusion. The model systematically analyzes spatiotemporal patterns, settlement time series, structural deformations, surrounding environmental factors, and spatial attributes. An attention-based feature fusion mechanism is employed to prioritize and integrate key influencing factors, thereby enhancing prediction accuracy. The model’s effectiveness was successfully validated using data from the Shanghai Yangtze River Tunnel project.
