Application of Laser Vision AI Monitoring Technology during Construction of Water Supply Tunnels Constructed by Combined Pipe Jacking and Shield Tunnelling Methods

Abstract

Aiming at the technical limitations of conventional manual monitoring for composite pipe‑jacking‑shield water‑supply tunnels under deep‑buried,high‑water‑pressure and complex stratum conditions,including limited monitoring coverage,delayed data acquisition and insufficient capacity for pre‑emptive risk early‑warning,this paper proposes a dual‑optical coaxial laser‑video artificial intelligence(AI) monitoring technology.Using this system,we achieve millimetre‑level deformation monitoring for vault settlement and cross‑section convergence by synchronously acquiring three‑dimensional coordinates of tunnels and apparent images of linings via laser ranging and visible‑light imaging.Combined with deep‑learning‑based defect recognition techniques,this study constructs a multi‑source heterogeneous monitoring system that considers both macroscopic structural deformation and microscopic lining damage.With a coupled model integrating numerical‑quantitative identification and image‑based qualitative verification,this study establishes a full‑process intelligent early‑warning and data‑tracing mechanism that supports index comparison, heat‑map‑assisted verification and four‑level risk‑classification alert dispatching. A 58‑day continuous field monitoring campaign was carried out on the composite pipe‑jacking‑shield tunnel beneath the Sunjiang Bridge of the Jinmen Water Supply Project.A total of 1377 valid datasets of vault settlement and 1019 datasets of haunch clearance convergence were collected.Monitoring results reveal that the cumulative vertical vault deformation of each tunnel cross‑section ranges from -0.45 mm to 1.36 mm.The clearance‑convergence deformation exhibits a three‑stage evolutionary pattern: rapid development induced by excavation disturbance, decelerated growth due to stress redistribution,and stable creep of surrounding rock.The overall tunnel deformation remains controllable and the supporting system operates stably. The proposed monitoring system can accurately identify abnormal structural deformation and apparent lining defects of tunnels,deliver hierarchical, precise early‑warnings as well as full‑chain data traceability.It effectively improves dynamic construction‑safety management capacity for complex hydraulic tunnels under high water pressure,is adaptable to construction conditions of composite pipe‑jacking‑shield tunnels, and shows promising prospects for engineering promotion and application.