地质科技通报 (Nov 2024)
Risk assessment and system development of surrounding rock instability in highway tunnel based on Bayesian network
Abstract
Objective With the rapid development of China's transportation industry, the geological conditions encountered in highway construction are becoming increasingly complex. Tunnels, owing to their ability to traverse mountainous terrain, are widely used in highway projects through challenging geological environments. However, as the number of tunnel projects has increased, the frequency of rock collapses and landslides during highway tunnel construction has also risen, resulting in significant economic losses and casualties. Therefore, accurate risk assessments have become crucial in tunnel engineering. Methods To address this, 40 cases of instability engineering were summarized and analyzed, refining 14 secondary indicators and establishing a comprehensive risk assessment index system. Risk was then assessed in terms of disaster probability and its consequences. The interpretive structural modeling (ISM) method was employed to construct a hierarchical topology diagram, and a Bayesian network model was established and refined using a causality graph method. The model was trained on 80% of the case data and validated with the remaining 20%. Based on this, the highway tunnel instability risk assessment Bayesian network evaluation system (RIAS) was independently developed, offering both engineering applicability and user-friendly functionality, and enabling accurate and rapid assessments of surrounding rock instability during highway tunnel construction. Results The system was applied to sections such as ZK5+937~ZK5+917 of the Beigushan Tunnel, predicting an 18.2% probability of tunnel instability with a "None(no risk)" magnitude of instability and a risk level of "Low Ⅰ" -consistent with the actual excavation results. Conclusion This study introduces and innovative approach by constructing a Bayesian network model tailored for highway tunnel risk assessments, overcoming the limitations of single-risk-level models and the challenge of insufficient engineering datasets. The model is successfully tested in the Beigushan Tunnel and holds significant potential for application in other highway tunnel projects, enhancingsafety and risk prediction capabilities.
Keywords
