Probabilistic Back Analysis of Failed Lateritic Soil Cutting Using Bayesian Approach

In this paper, the back analysis is performed on the failed lateritic soil cutting using a probabilistic Bayesian approach, updating uncertain spatially varying parameters based on observed slope failure behaviour due to rainfall. This study demonstrates that coupled-flow deformation (CFD) analysis and Bayesian approach, combined with random field theory, are effective for back-analysing slope parameters. The prior probability density functions (PDFs) of cohesion (c'), friction (ϕ') and unit weight (³) are obtained by generating normal random fields for the prescribed coefficient of variation and anisotropic scale of fluctuation using the Karhunen-Loéve expansion method. The factor of saftey obtained from CFD analysis and prior PDFs are then used to generate the posterior PDFs of random variables using the Bayesian approach. It is observedthat the posterior PDFs are narrower than the prior ones with reduced mean for c' and ϕ', whereas the posterior PDFs for ³ have a higher value of the mean, overall indicating a reduction in shear strength of soil leading to instability of slopes. This method enhances the understanding of geotechnical strength parameters through probabilistic inverse analysis. The updated information on strength and unit weight parameters from this approach can be utilized for improved slope stability assessments, addressing the significant issue of uncertainty in soil parameters and offering a robust tool for geotechnical engineering applications.

Keywords: Laterite soil cutting, Bayesian method, Random fields, Anisotropic scale of fluctuation.