Probabilistic Back Analysis of Failed Slopes using Bayesian Techniques

摘要

Slope stability problems are generally associated with uncertainties in soil properties, pore pressure regimes, failure modes and geometries, in addition to model uncertainties from the analysis methods. For failed slopes, information about the observed failure geometry could be used to refine or update the uncertainty in the shear strength and pore pressure parameters. This paper adopts a simplified probabilistic back-analysis approach based on Bayesian techniques that is applied within the framework of a limit equilibrium analysis for a case history involving the stability of a proposed highway stretch in Algeria. The objectives are to study the effects of (1) the quality of prior shear strength and pore pressure data, (2) the correlation between shear strength parameters, (3) the incorporation of the pore pressure regime in the back analysis, and (4) effect of the failure surface location and geometry on the results of the probabilistic back analysis procedure. Results indicate that the updating process generally results in mean parameters that are likely to exist at the initiation of failure and in reduced uncertainties in these parameters. Results also show that the uncertainty in the geometry and location of the failure surface impacts the posterior data and should be accounted for in designs of any future remedial measures.