Analysis of cascade collapse mechanism and prediction model for determining collapse height of block rock tunnel

Abstract

Excavation in underground engineering projects, such as tunnels and subterranean caverns, poses significant risks of sudden and destructive collapse. To explore the mechanisms and factors influencing collapse in stratified parallel structure without support, laboratory model tests, numerical simulations, and machine learning techniques have been employed. Five model tests have been utilized to focus on rock mass instability during tunnel construction, using the optical flow method to analyze instability characteristics and coupling effects in block crack tunnels. Model tests on jointed rock evaluates the surrounding rock behavior using the Universal Distinct Element Code, identifying six collapse modes and quantitatively analyzing factors governing collapse height through range analysis method. The ranking of these influential factors is: bedding inclination > tunnel span > joint friction Angle > tunnel buried depth > lateral pressure coefficient > joint spacing > joint cohesion > elastic modulus > Poisson’s ratio. The k-Nearest Neighbor algorithm is employed to develop a stable state prediction model of surrounding rock. One-variable nonlinear and multiple linear regression analyses are performed on influencing factors with a range greater than 1.5, leading to the establishment of a collapse height prediction model. The prediction results achieved over 88 % accuracy when validated against the five laboratory tests. This model was also applied to analyze the Ganggou tunnel collapse on the Beijing-Shanghai Expressway, confirming its effectiveness in predicting collapse heights. The research findings provide valuable insights for predicting, preventing, and managing the stability of surrounding rock during excavation in block fissure areas, offering significant engineering application value.