Abstract: Tunnel construction has a large impact and significantly disturbs the surrounding environment.To address complex geological disturbances and multi-system coupling failures,this study proposes a tunnel safety intelligent management method based on a five-dimensional digital twin architecture,overcoming traditional limitations.A five-dimensional model integrating “physical entity,virtual model,connection layer,association mapping layer,and optimization feedback layer” has been established.Combined with a BIM-FEM closed-loop mapping system and CNN algorithm,this model achieves full-chain integration from data collection to dynamic optimization.A hierarchical mesh division strategy has been proposed for different research objects,enabling multi-scale coupling simulation of soil-tunnel-building.A dynamic perception network integrating IoT sensors and 3D laser scanning has been developed,with a 5G cloud platform achieving real-time millimeter-level monitoring.By designing a safety evaluation model based on the CNN algorithm,the optimal tunnel reinforcement scheme was generated for the White Tower Hill project,reducing surface settlement from 8.221 mm to 0.038 mm and meeting the protection requirements of ancient buildings.The results show that this method can reduce construction redundancy and environmental disturbance while enabling real-time risk warnings.The research provides a “perception-simulation-decision-making” integrated intelligent governance paradigm for tunnel engineering,promoting a shift in tunnel construction safety management from experience-driven to model-driven approaches.