First in Urban Rail: AI-Powered Agent for Quality Hazard Identification & Control Enters Commercial Service

Recently, AECC of UCD has achieved a milestone by independently developing and launching the first AI-powered agent for quality hazard identification & control in urban rail, a solution that has now entered commercial service.

Over a year ago, a large-scale intelligent AI model was successfully applied to identify safety hazards in urban infrastructure construction. This latest achievement represents another technological product that helps to comprehensively secure  both safety and quality in the sector with AI. 

The agent features four core functions and application scenarios: visible quality defect identification, waterproofing detail quality identification, water leakage identification, and unexpected personnel gathering identification. It can capture and assess problems within seconds, automatically generate traceable registers, and provide around-the-clock services for excavation pits, tunnels, elevated structures, etc. It enables precise identification, closed-loop traceability, and zero blind spot omission. This effectively addresses the long-standing problems in rail construction, such as difficult quality control, delayed hazard detection, and human errors in supervision, ensuring the safety and quality in rail construction.

Visible Quality Defect Identification: It precisely addresses visible quality defects in tunnel construction. The system can automatically capture, accurately identify, intelligently assess, and archive common quality problems in core construction areas, such as concrete structures, waterproof-layer installation, and shield tunnel segment assembly, while providing real-time alerts. It automatically creates standardized, traceable registers that fully record defect location, type, occurrence time, and other relevant details. This effectively tackles shortcomings of manual inspection, including missed detection, false detection, and non-standard record, assisting management personnel in rapidly identifying problems and achieving efficient closed-loop handling.

Waterproofing Detail Quality Identification: Concealed works in urban rail  involve complex processes and highly confined operations, making post-completion verification and evidence collection difficult. As a result, traditional manual inspection cannot effectively maintain full-process supervision. The agent is equipped with high definition intelligent capture cameras and a quality hazard identification model. It automatically identifies the on-site construction areas and key operation elements, targeting core concealed processes like construction joints, deformation joints, waterproof-layer installation, and segment joint treatment. It provides full-process, real-time video traceability and gives real-time alerts for non-compliant operations and unexpected problems. This enables truly comprehensive construction monitoring, data traceability, quality verifiability, and closed-loop problem resolution.

Water Leakage Identification: Considering the high level of concealment, rapid development, and serious hazards of water leakage, the system performs real-time intelligent detection on key leak-prone areas such as the tunnel face, primary support, shield launching and receiving portals, horsehead-shaped door openings, and ring beam construction. Using high resolution dedicated cameras and time series recognition technology, the system realizes three types of early warnings: no water → water detected: initial water leakage identification and warning, eliminating potential hazards at an early stage; small amount of water → large amount of water: real-time alerts for water-volume changes and dynamic monitoring of evolving trends; clear water → muddy water: warning of changes in water quality, allowing early identification of high risk signs of water and mud inrushes. Following the principle of “process visibility, real time warning, and closed loop rectification”, the system provides full process monitoring of high risk leak prone areas, including the tunnel face, working face, construction joints, deformation joints, and the joints between shield segments and the portal ring beam. It can detect unusual changes in water conditions at the earliest possible moment.

Unexpected Personnel Gathering Identification: Rail construction takes place in complex environments, where work areas are scattered and the workforce is large, posing a great challenge for control. Unexpected personnel gathering is often an early sign of safety incidents. For instance, if a worker is injured, equipment fails, or a sudden hazard arises, not promptly noticing the gathering will quickly worsen the situation. Equipped with high precision intelligent visual analysis technology, the agent provides around-the-clock dynamic monitoring across all work areas, and accurately identifies unexpected personnel gathering. The system supports adjustable parameters, including the range for personnel gatherings, personnel density, and duration, to accommodate different environments like excavation pits, tunnels, and elevated structures. Once the system detects that a pre-set risk level has been reached, it immediately issues an alert and pushes notifications to remind management personnel to intervene quickly. This tackles key drawbacks of manual inspections (such as limited coverage and lack of nighttime or blind spot supervision), and reduces secondary safety incidents.