The widening of an expressway presents an unique engineering challenge: converting the existing four-lane twin-bore tunnel into a single-bore, four-lane structure through loose rock formations. This challenge is currently being addressed at the Yanfeng Tunnel section on the north Taizhou segment of Zhejiang's Yongtaiwen Expressway, a key coastal corridor. The "two-in-one" reconstruction of this twin-arch tunnel has officially commenced, and leading the venture is a robot dog named "Xiao An".
The original Yanfeng Tunnel spans 605 meters. The expansion involves a maximum excavation cross-section of 257.6 square meters through unstable Class IV/V rock mass. The poor self-stabilizing capacity of the surrounding rock presents significant technical challenges and safety risks. Furthermore, traditional methods for inspecting the tunnel crown for over-excavation and stability are inefficient and hazardous, forming a critical bottleneck for the project's safe progress.
The project team has deployed the intelligent robot dog "Xiao An" to conduct inspections in high-risk areas,replacing human workers. Equipped with an intelligent analysis system powered by the DeepSeek large model, "Xiao An" can collect data in real-time and use AI algorithms to accurately identify loose rock masses and other potential hazards in the tunnel crown. It features multi-modal fusion capabilities: high-precision video serves as keen "eyes", while radar point cloud modeling provides "X-ray vision", enabling it to quickly generate a millimeter-accurate 3D model of the tunnel. This model allows for precise calculation of over-excavation volume and facilitates the advanced prediction of collapse risks, thereby evolving the tunnel inspection from "post-incident discovery" to "pre-warning".
The use of "Xiao'an" is helping build smart construction capabilities in the Yongtaiwen Expressway's Taizhou North Section. From the drone-based "aerial traffic control" over the Wu'ao hub to the "smart sentinel" in inspection robots inside the new tunnel, the project is steadily establishing a new safety management system characterized by human-machine collaboration and intelligent governance. Moving forward, the project team will promptly summarize the lessons learned, aiming to establish standardized and replicable models for border application.