A digital twin-enabled 3D real-time monitoring system of subway tunnel waterlogging for disaster-resilient infrastructure

Frequent extreme weather events exacerbate urban flooding risks and subway tunnel waterlogging, threatening infrastructure resilience. Real-time monitoring is crucial, yet conventional sensors are vulnerable and image-based methods fail in low-light conditions. This study proposes a digital twin-enabled Physics-Digitization-Intelligence Fusion System (PDI-FS) for real-time tunnel waterlogging monitoring. The system integrates multi-source data (RGB, depth, infrared) to dynamically reconstruct a 3D digital twin model of the tunnel. Waterlogging depth is determined via RANSAC-based water surface identification, and volume and flow rate are calculated from tunnel geometry. Validation experiments under illuminated and dark conditions demonstrate average monitoring accuracies of 99.12% for waterlogging depth, 95.18% for volume, and 97.29% for flow rate. Comprehensive robustness analysis demonstrates the system's resilience to realistic tunnel conditions, confirming field applicability. The proposed PDI-FS enables non-contact monitoring without reference objects or scale calibration, remains effective in complete darkness, and addresses the specific lack in subway tunnel waterlogging monitoring, thereby enhancing disaster prevention and infrastructure resilience.