Application of Artificial Intelligence on the Uncertainty Quantification Behavior of Creep Rock and Reliability-Based Optimization Design of Deep Tunnel

This work aims at optimizing the support system of deep tunnel constructed in creep rock formation. For this purpose, the probabilistic inversion based on the Bayesian inference is firstly chosen to quantify the uncertainty of the time-dependent behavior of rock mass using the evolution in time of tunnel convergence data. The stochastic modeling is then conducted considering the quantified uncertainty of creep rock on the optimization design of tunnel support. For the uncertainty quantification and reliability-based optimization design processes, the artificial neural networks are chosen as the surrogates to simulate the time-dependent tunnel displacement and equivalent stress in concrete liners. The validation and efficiency of the developed procedure to derive the optimal tunnel support that verifies two failure modes, namely the support capacity criterion and the maximum tunnel convergence is highlighted through the numerical applications.