Dynamic design system for Active-Passive combined support considering pre-tension force limited by geology in soft rock tunnels

Abstract

The design of active–passive support parameters traditionally relies on rock mass classification. However, the ultimate load of anchors in soft rock is significantly constrained by geological conditions and is not considered in rock mass classification methods. Consequently, the constant pre-tension force for the same rock mass classification poses a risk of anchor system tension failure. The paper introduces a dynamic design system for active–passive combined support to bridge this gap, considering the interplay between geological conditions and pre-tension force. The system established a predictive model for pre-tension force under geological conditions using on-site monitoring data and B-P neural networks. Considering the instantaneous loss during pre-tension force locking to determine effective pre-stress, it then utilized numerical simulation, on-site monitoring, and B-P neural networks to correlate pre-stress, geological conditions, passive support parameters, and rock deformation. Support parameters were subsequently adjusted based on predicted rock deformation in a predetermined optimization sequence. The system was taken to practical application in the Muzhailing Highway Tunnel. The results indicated that the predictive model for designed pre-tension force values achieved goodness of fit of 0.701 and mean absolute percentage error (MAPE) of 6.51%, guiding pre-tension force design. Additionally, a rock deformation prediction model, considering effective pre-stress and constructed using numerical simulation and field-measured data, attained an R-squared value of 0.947 and an MAPE of 2.76%. Practical application on three cross-sections demonstrated a high proportion of anchor holes reaching the designed pre-tension force (94%, 91%, and 97%, respectively), preventing anchoring system failure during tensioning. Predicted rock deformation values under designed support parameters closely match actual values, with percentage errors of 1.13%, 6.73%, and 2.15%, respectively. These outcomes contribute valuable insights to the design methodology of active–passive combined support in soft rock tunnels.