Risk assessment of tunnel excavation beneath ancient buildings based on dynamic Bayesian networks for double-line tunnels

Urban tunnel construction projects frequently involve large-scale excavation activities that can disturb the geological strata, posing potential hazards to ground stability and the safety of surrounding underground and surface structures. This concern is particularly pronounced when tunnels pass beneath clusters of ancient buildings, as these historical structures often possess unique architectural features and varying structural conditions that render them highly sensitive to ground movements. Accurately assessing and predicting the effects of tunnel excavation on such buildings is therefore critical for ensuring both construction safety and the preservation of cultural heritage. This study primarily aimed to develop a comprehensive, dynamic risk assessment framework that captures the interactions between tunnel excavation, soil deformation, and the structural behavior of ancient buildings. Using the Lanzhou Baita Mountain Double-Line Tunnel project as a representative engineering case in which the tunnel passes directly beneath a major historical building cluster, this research seeks to provide a scientifically grounded methodology for evaluating construction-induced risks and guiding effective mitigation strategies.

This study proposes a novel risk assessment approach that combines dynamic Bayesian networks (DBNs) with finite element analysis (FEA) to model the complex, time-dependent interactions between tunnels, soil, and building structures. A detailed finite element model was first developed to simulate the stress distribution, deformation, and displacement patterns induced by tunnel excavation at multiple construction stages. These simulations provided quantitative data on ground movements and structural responses, which were then incorporated into a DBN framework. The network models the probabilistic relationships among key variables, including tunnel construction parameters, soil mechanical properties, building structural characteristics, and historical deformation patterns. The DBN allows for real-time updating of risk probabilities as new monitoring data become available, thus enabling dynamic prediction of risk evolution throughout the tunnel construction process. This approach also facilitates the quantification of the relative contributions of various risk factors at different construction stages, thereby identifying critical phases during which ancient buildings are most vulnerable. Model validation was conducted by comparing Bayesian network predictions with finite element simulation results to evaluate predictive accuracy and reliability.

The results demonstrate that after integrating finite element deformation data into the DBN model, the risk levels of overlying ancient buildings during tunnel excavation are predominantly classified as Grade Ⅱ. The analysis identifies building structural characteristics and tunnel-related excavation factors as the primary contributors to the observed risk. Prediction error rates of 5.8% for the left-line tunnel and 10% for the right-line tunnel confirm the model’s reliability and practical applicability. The model also provides a dynamic visualization of risk evolution over time, highlighting the stages during which the ancient buildings are most susceptible to damage. Based on these findings, targeted mitigation measures are proposed, including staged structural monitoring, reinforcement or optimization of supporting structures, and real-time adjustment of excavation parameters. These measures help ensure that risk levels remain effectively controlled while maintaining construction efficiency.

The integration of DBNs with FEA provides a robust and reliable methodology for dynamically assessing the risk of tunnel construction impacts on ancient buildings. The proposed framework effectively identifies critical risk factors, quantifies the evolving risk levels during construction, and supports proactive intervention strategies. By enabling continuous monitoring and predictive assessment, this method enhances safety management in urban tunneling projects while safeguarding historically significant structures. The findings provide a scientifically validated approach for decision-making in complex urban construction projects involving heritage conservation, offering theoretical insights and practical guidance for engineers, project managers, and policymakers.