Since it is difficult to control tunnel construction in weak and broken surrounding rock, the large deformation of tunnel surrounding rock has always been a great danger to construction safety. In this study, the characteristics of a tunnel collapse deformation and its causal factors are analysed as an example of a large deformation accident in a tunnel with granite residual soil in the surrounding rock. First, a support vector regression algorithm optimized by the genetic annealing algorithm (GASA-SVR) is used to construct a back-analysis model of the surrounding rock mechanical parameters. Then, based on the field measured data, the surrounding rock mechanical parameters of the tunnel are inverted. Finally, the finite element method is used to simulate different surrounding rock deformation control schemes, and the optimal deformation control measures are obtained. The results show that the simulated annealing algorithm (SA) enhances the local search capability of the genetic algorithm (GA) so that the hyperparameter optimization solution of the support vector regression (SVR) is the global optimal solution. The parameters obtained from the inversion of the GASA-SVR model are simulated in PLAXIS3D within 4.0% of the measured values in the field. With the application of optimized support measures at the site, the deformation of the surrounding rock has been effectively controlled.
Read full abstract