PROBABILISTIC BACK ANALYSIS OF RESIDUAL LATERITIC SOIL CUTTING OF KONKAN REGION

Abstract

Landslides are one of the most impactful natural disasters, posing a significant threat to both life and property. Rainfall-induced landslides stand out as the most prevalent form of slope failures globally. It has been proposed that analyzing the stability of a slope necessitates a thorough and comprehensive study of saturated-unsaturated transient seepage analysis, considering the major effects of rainfall infiltration. This is because the traditional methods of stability analysis typically assume a saturated condition, however, the actual state of a number of hill slopes is partially saturated condition. Matric suction plays a crucial role in stability analysis of unsaturated soil slopes by augmenting the negative pore water pressure, consequently enhancing the effective stress levels. This report presents the deterministic and probabilistic back analysis of lateritic soil cutting at MIDC-Zarap in Ratnagiri district, Maharashtra. The deterministic back analysis is performed using uncoupled and coupled flow deformation analysis in GeoStudio 2D. The probabilistic back analysis is performed using Bayesian analysis. The soil is collected from the cutting in Ratnagiri district of Maharashtra for the basic soil classification tests. The input parameters for the soil water characteristics curve are obtained from the grain size distribution data. The shear strength parameters are estimated by performing back analysis on old cutting geometry that was failed due to rainfall in August 2019. The numerical analysis is then performed for old and new cutting geometry under the effect of rainfall infiltration using both uncoupled and coupled flow deformation analysis. The parametric study is conducted for the new geometry for varying rainfall intensity and duration. The effect of variation in water table and depth of competent strata below the slope is also studied for new geometry using UFD analysis. The sensitivity analysis is conducted on old geometry and sensitivity indices are calculated for material properties cohesion, friction angle and unit weight of soil. The normal random fields of cohesion, friction angle and unit-weight are generated using Karhunen-Loéve Expansion method in Optum G2 software. The prior and posterior probability density functions of random variables are then established using Bayesian method.