Risk Assessment of Tunnel Water Inrush Based on Hybrid Optimisation Algorithms: A Case Study in Southwest China

Tunnel water inrush risk assessment poses a complex multi-attribute decision-making challenge, garnering significant global attention as a critical yet unresolved issue in geotechnical engineering. While the Attribute Interval Recognition Theory (AIRT) has been widely adopted for geological hazard prediction, its application in tunnel water inrush scenarios often faces limitations in precise index attribute quantification and risk probability estimation. To address these constraints, this study proposes an improved integrated model combining AIRT, the Analytic Hierarchy Process (AHP), the Inverse Entropy Method (IEM), Game Theory (GT), the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS), and Monte Carlo random simulation technology. In this framework, AHP and IEM are used to determine subjective and objective weights, GT is used to optimise the combined weights, TOPSIS is introduced to determine the index attribute measure coefficients, and Monte Carlo simulation is employed to estimate risk probabilities. The novelty of the model lies in improving both risk attribute identification and probabilistic assessment under interval-valued index conditions. The model was validated using 20 representative tunnel water inrush cases, demonstrating robust performance in capturing dynamic risk scenarios. When the confidence level was set to 0.65, all 20 cases were correctly classified, achieving an accuracy of 100%. Subsequently, the model was applied to the inlet section of the Nafeng Tunnel. The results show that the assessed section is classified as Class II, with risk probabilities of 0.515 and 1.000 for the D0 + 220.6–D0 + 285 m and D0 + 285–D0 + 340.6 m sections, respectively. These findings indicate that the proposed model can provide a quantitative basis for tunnel water inrush risk assessment and prevention.