Application of Probabilistic Methods to Slope Stability - A Comparative Study Using Three Computational Tools: PEM, SLOPE/W, and a Custom Python Script for PLAXIS

Abstract

Slope stability calculations are predominantly performed using deterministic meth ods, where a single value is assigned to each soil parameter based on data from a limited number of boreholes. Since soil properties vary naturally, and assumptions must be made, uncertainties arise. Traditionally, these uncertainties are handled through partial factors and dimensioning must be based on minimum factors of safety defined in Eurocode. An alternative is to introduce a probabilistic approach, which yields a probability of failure. This can provide a more comprehensive picture of the problem and has the potential to enable more optimized designs. In this thesis, three probabilistic methods are applied: the Point Estimate Method, Monte Carlo and Latin Hypercube simulations. The software programs SLOPE/W and PLAXIS have been used to apply these methods through limit equilibrium anal ysis and finite element modeling. Probability distributions are assigned to friction angle, unit weight, and effective cohesion, with the geometry derived from a slope in Nolhaga by the riverbank of Säveån. Since PLAXIS lacks built-in probabilistic functionality, a custom Python script was developed to automate the process. The results show that all three methods produce comparable mean factors of safety, while differences in the estimated probability of failure are more pronounced. The choice of distribution has limited effect on the mean factor of safety but can influence the probability of failure by orders of magnitude. Spatial variability is found to have a significant impact on the results. Although the water level is identified as a key driver of slope stability, further development is required to introduce it as a varying input parameter, enabling it to be analyzed alongside the other parameters and improve the robustness of the results. The thesis concludes that probabilistic analysis offers valuable insight beyond deterministic methods, but requires careful consideration of input assumptions and the limitations of available tools.