Risk assessment of slopes at railway tunnel portals in cold regions under freeze–thaw action

Abstract

In the cold regions of Northeast China, slopes at railway tunnel portals are subjected to long-term freeze–thaw cycles, which progressively degrade the soil structure and seriously threaten the infrastructure and operational safety of railway tunnels. However, existing evaluation methods cannot adequately integrate the mesoscopic freeze–thaw damage mechanism with macroscopic multi-factor coupling effects, and a systematic risk assessment framework is lacking. This study focuses on the slope at the Limin Tunnel portal. The discrete element method is used to simulate the mesoscopic damage process of the slope soil under freeze–thaw action. By combining global navigation satellite system field measurements, a four-dimensional evaluation index system is constructed, covering freeze–thaw driving factors, soil resistance properties, mesoscopic damage evolution, and macroscopic deformation response. Furthermore, the analytic network process (ANP) is introduced to establish a dependency and feedback network among the indicators, which overcomes the limitation of conventional methods that treat indicators as independent and ignore feedback mechanisms. Finally, risk assessment is performed using a coupled set pair analysis and variable fuzzy set model. The results show that shear-dominated damage develops during freeze–thaw action, with shear fractures accounting for 90.43% of the total fractures. The force chain strength exhibits a characteristic evolution: an initial increase, followed by fracture-induced degradation during rapid freezing, and eventual disintegration in the thawing stage. The overall risk level of the slope is Grade II (moderately low risk), with a grade characteristic value of approximately 2.44, while the assessment also shows an evident tendency toward Grade III. Sensitivity and uncertainty analyses indicate that the final risk classification remains stable under ANP weight perturbations of up to ±30%, whereas expert-score perturbations exert a greater influence, with the protective structure identified as the dominant source of uncertainty. A comparative AHP–FCE analysis yields the same Grade II classification, further supporting the stability of the overall risk assessment. The discrete element freeze–thaw damage model and the coupled ANP–SPA–VFS evaluation framework developed in this study overcome the disconnection between freeze–thaw mechanisms and macroscopic assessment and the difficulty in quantifying nonlinear multi-factor coupling. The proposed framework demonstrates a feasible approach for integrating mesoscopic freeze–thaw damage with multi-factor risk assessment at the Limin Tunnel portal slope and may provide a methodological reference for similar cold-region slopes.