Reliability based analysis of spatially random cohesive multi-layered slopes -

Abstract

Slope stability is a branch of geotechnical engineering that is affected significantly by uncertainty due to spatial variability of soil properties. Traditional deterministic slope stability design methodologies account for these uncertainties in an indirect manner by adopting a target factor of safety that is generally equal to 1.5. In the last two decades, results from published reliability-based design methods for slopes indicated that even for the case of a simple homogeneous slope, a factor of safety of 1.5 may not lead to designs that have a consistent level of risk. More importantly, published results show lack of knowledge about the reliability levels that are inherent in the design of multi-layered slopes which are very common in applications involving compacted embankments over soft clays or in natural multi-layered slopes. The main objectives of this thesis are twofold. The first objective is to quantify the reliability levels that are inherent in spatially random multi-layered clayey slopes with different thicknesses and soil properties. The second objective is to generate simplified equations that would allow for conducting reliability-based design for a two-layered slope system using only the deterministic results and the input soil properties. To achieve these objectives, Monte Carlo simulations that involve clayey slopes with multiple layers are performed using the finite difference software FLAC. To allow for realistic modeling of the failure surface in the multi-layered system, FISH functions were used in FLAC to model the spatial variability using realistic anisotropic random fields that would ensure a realistic representation of the soil properties in the multi-layered slope and result in probabilities of failure that reflect the real level of risk. The models generated to predict the behavior when accounting for spatial variability in soil will lead to improved design practices for multi-layered clayey slopes.