Probabilistic back analysis of rock strength parameters in heavily jointed rock slopes based on Bayesian inference

The traditional probabilistic back analysis method of rock strength parameters is not always reasonable and reliable, where only a fixed slip surface is assumed or the condition of the safety factor Fs = 1 is used. Aiming at the heavily jointed rock slope, a probabilistic back analysis method of rock strength parameters using the DiffeRential Evolution Adaptive Metropolis DREAM(KZS) algorithm was established, with the Fs, the critical slip surface data and the prior distribution of rock strength parameters including the geological strength index (GSI) and the uniaxial compressive strength (σci) based on the Hoek–Brown failure criterion and Bayesian inference. The rationality and reliability of the proposed method were verified by combining a homogeneous jointed rock slope and an artificial slope in Manisa, Turkey. Then, the influences of different slip surface errors on the back-analysis results, prior mean and coefficient of variation (COV) of rock strength parameters on the posterior ones were studied. Results showed that the rock strength parameters can be effectively identified by the DREAM(KZS) algorithm, and the influence of the slip surface errors on the back-analysis results cannot be ignored. The posterior distribution was less affected by the prior distribution of rock strength parameters, and the posterior mean gradually deviated from the prior mean with the increase of the COV. This method can improve the back-analysis identification accuracy of rock strength parameters by considering the uncertainty of the slip surface, providing a basis for later remedial engineering reinforcement design of the slope.