Reliability – based analysis of embankment dams

Embankment dams are widely used as water-retaining structures through the world. Two principal structural failure modes of embankment dams are internal erosion and slope instability. The objective of this thesis is to investigate the feasibility of applying probabilistic methods in embankment dam assessment. Compared to deterministic methods, probabilistic methods considers the variability in material properties and the results are more assessable. Meanwhile, the derived sensitivity factors of the variables can be used to find the most influential one.Through a literature review, it is found that the difficulties in defining a limit state function for the full process of internal erosion for embankment dams hold back the application of reliability-based methods. However, reliability-based methods have been recognized as suitable for slope stability assessment.In this thesis, both deterministic and probabilistic calculations have been performed for a homogenous embankment dam. The deterministic calculation is carried out in Slope/W to first find out the most critical slip surface. The failure probability of this slip surface is estimated by both Monte Carlo simulations in Slope/W and FORM analysis in Comrel. Then, the 34 most critical slip surfaces from Slope/W are chosen to calculate the system reliability with simple bounds and applied integration. The results from the deterministic and probabilistic calculations showed that the studied dam has a higher risk of failure than allowed. In the system reliability calculation, simple bounds gives a wide range, especially the upper bounds. It was found that classification of different slip surfaces into different groups with respect to geometry and material was useful to simplify the calculations. Not much improvement in accuracy of the system probability of failure is found when increasing the numbers (from 1 to 5) of the most critical slip surfaces from each group. Even the combination of 1 slip surface from each group only gives an error of 6.5%. Categorizing slip surfaces before performing a system reliability analysis is a good way for simplification of the calculations. Due to this, the criteria used for categorization of the slip surfaces are of interest. In this work, it was found that the slip geometry in combination with the materials included in the slip surface constitute a possible way forward to do this.