Numerical Methods in the Stability Analysis of Slopes

Abstract

Slope failure is a complex problem that causes serious hazards throughout the world. Due to this there is large loss of life and property. So, it became very much essential to understand these slope failures, analyse and predict its vulnerability for proper mitigation of hazards. A number of methods are available for slope stability analysis and prediction. Most common among these is the convectional limit equilibrium and numerical methods. Now-a-days Artificial intelligence (AI) techniques are used widely for this purpose. This paper aims to provide a comparison of traditional methods of slope stability analysis i.e. limit equilibrium and finite element method with the recently developed artificial neural network (ANN). From the data available in the literature, a comparative study of factor of safety is carried out between the various methods through commercially available software Geostudio. The input parameters used in the study include- unit weight of soil (ϒ), cohesion (C), angle of internal friction (Ø), height of slope (H), angle of slope (α) and pore water ratio (ϒu), where as factor of safety (FOS) is the only output parameter. Also the relationship between various soil parameters and stability is established using regression analysis. The methods used in the LEM analysis are Ordinary method of slices, Bishop’s method, Morgenstein Price method, Janbu’s method, Spencer’s method and for FEM analysis strength reduction technique is used. It is found that there is good agreement between the conventional LEM and FEM. The factor of safety value obtained with FEM is a bit higher than that obtained with LEM. Also when comparing the results with ANN, it is found to give more accurate results.