GPR‐MCS model of reliability analysis of key blocks and its engineering application

Abstract Reasonable rock bolting schemes for potentially unstable blocks should be proposed and optimized in order to prevent rockfall hazard during tunnel construction. However, the reliability index of key block caused by uncertainty of geometric and mechanics parameters is not easy to be obtained by numerical simulation. Since the performance function of stability analysis of supporting key block is too complex, it is also difficult to calculate the reliability by using analytical and graphic methods rapidly. In this paper, the Gaussian process regression (GPR) model was established instead of the explicit function of key block. The safety factor of supporting key block calculated by UnWedge procedure was taken as the sample output of response model. With the GPR model coupling Monte Carlo Simulation (GPR‐MCS), the failure probability of key block could be predicted rapidly. The reliability of key block with different supporting schemes were analyzed, and the optimization of support parameters for key block was also carried out. It has direct meaning to prevention and control of dangerous key block during tunnel construction.