Building Information Modeling Driven By Spatio-Temporal Swin-U-Net: Ultra-short-term Deformation Prediction of Adjacent Municipal Pipelines During Deep Subway Station Excavation

The excavation of deep subway station foundation pits in densely built urban areas poses considerable risks to adjacent municipal pipelines, where deformation may evolve rapidly under complex and time-varying construction conditions. Existing statistical models often fail to capture nonlinear spatiotemporal deformation patterns, while numerical simulation approaches are computationally expensive and difficult to deploy for ultra-short-term, high-frequency prediction required on construction sites. Moreover, current applications of Building Information Modeling (BIM) are largely limited to static visualization and lack effective integration with real-time monitoring data.