In probabilistic back analysis for rainfall-induced slope failure, the computational load is usually demanding due to time-consuming numerical deterministic model. In this paper, a polynomial chaos expansion (PCE)-based MCMC simulation is proposed to accelerate the probabilistic back analysis. A surrogate model based on PCE is used to substitute the coupled hydro-mechanical model of an unsaturated soil slope under rainfall infiltration. The coefficients in the PCE surrogate model are estimated using the spectral projection method with the Gauss-Hermite (GH) sparse grid. The posterior inferences of soil parameters are obtained using the Markov Chain Monte Carlo (MCMC) simulation. The results of an example show that the proposed method is computationally faster and more efficient than the traditional approach in exploring the posterior parameter distributions.
展开▼
