A Tunnel Squeezing Prediction Model Based on the Hierarchical Belief Rule Base

Tunnel squeezing is a time-dependent significant deformation problem often occurring in weak rock masses and areas with high horizontal in situ stress. This phenomenon can cause construction delays, increased budgets, tunnel collapse, and additional problems. Therefore, accurately predicting tunnel squeezing is crucial. The surrounding rock of the tunnel is uneven in hardness and softness, and changes frequently due to the influence of geological structure, thereby rendering the prediction of extrusion deformation highly uncertain. The belief rule base (BRB) is a rule-based modeling approach capable of handling uncertain information. However, the performance of the BRB model is not only affected by the combinatorial rule explosion caused by too many input attributes but also by the limitations of expert knowledge. First, to solve the problem of the combinatorial rule explosion, a novel tunnel squeezing prediction model using a hierarchical BRB structure based on the Random Forest (RF) attribute selection method (H-RF-BRB) is proposed. Second, to avoid the limitations of expert knowledge, parameters of the tunnel squeezing prediction model are determined by combining the expert knowledge and the information gain ratio (IGR). The model effectively integrates qualitative knowledge with quantitative information, addressing the issue of limitations in expert knowledge. Additionally, it overcomes the challenge of limited datasets due to the difficulties in collecting tunnel squeezing samples, which enhances the accuracy of the model’s predictions. Finally, the model’s effectiveness and superiority are validated through five-fold cross-validation and several comparative experiments.