Abstract:

Loess tunnels are prone to large deformation and collapse. Accordingly, to assess the stability of loess tunnels accurately and efficiently in construction, five physical and mechanical parameters of loess, such as water content, dry density, void ratio, cohesion, and internal friction angle, are selected as stability predictors under the same construction condition when considering the coexistence of randomness and ambiguity in stability assessment. Furthermore, based on principal component analysis, the weights of predictors are obtained, showing that water content, cohesion, and internal friction angle affect the stability of loess significantly. Then, the characteristics of the cloud model are obtained through the grading standard of predictors. Finally, based on the weights and characteristics of the cloud model, the stability evaluation model of a loess tunnel during construction based on principal component analysis and normal cloud theory is constructed using the forward cloud generator. The proposed evaluation model was applied to the loess tunnel of the Yan′an section of the Menghua railway. The prediction accuracy rate is as high as 90%, which is better than that of the Criticcloud model, thereby validating the feasibility and effectiveness of the proposed model.

Key words: loess tunnel, stability assessment, principal component analysis, cloud model