Research and application of reliability evaluation model for water inrush risk during tunnel construction

Based on the formation mechanism of tunnel water inrush risk, a comprehensive evaluation indicator group was established by applying TSP geological forecast, observation statistics and geological exploration report information. Furthermore, a probability evaluation model for tunnel water inrush risk is proposed by combining machine learning methods and reliability theory. During this process, the Monte-Carlo model was used to avoid the difficulty of solving multiple integrals of the joint probability density function, and trained the RFR model as its fast response surface. Comparative calculations show that the reliability method effectively solves the randomness problem of water inrush prediction indicators in tunnel engineering. Compared with deterministic calculation methods, it can provide more complete and accurate evaluation results from a probabilistic perspective. In order to provide a more complete reference for engineering applications, the key parameter values of RFR and the number of Monte-Carlo sampling were discussed and clarified. At the same time, the sensitivity characteristics of each evaluation indicator were analyzed. The research results have been successfully applied to the construction process of the tunnel group in the YA15 section of the Puyan Expressway in Fujian Province of China, and have achieved good results.