Design of a Probabilistic FEM Analysis Using the Sellmeijer Model for Backward Erosion Piping

Backward erosion piping (BEP) is an important failure mechanism for clay dikes founded on sandy soil or dikes resting on an impermeable blanket above a sandy layer. To compute a failure probability for this mechanism, probabilistic calculations are required. Until now these are computed in the Netherlands using the analytical Sellmeijer model. This model is based on a strongly simplified subsurface. Therefore, it is desirable to perform probabilistic calculations with models that can more accurately depict the subsurface features such as with a Finite Element Methods (FEM). In this paper, the FEM BEP model DgFlow is used, which couples the Sellmeijer erosion model to a FEM model for groundwater flow, allowing to take complex features, such as heterogeneity and anisotropy, into account when assessing dikes. This paper shows the first attempt to perform probabilistic BEP calculations using DgFlow. Different limit state functions and probabilistic methods are tested to find a method that provides accurate, fast, and reliable results. It is found that FORM computations result in sufficiently quick (time) and accurate (compared to MC) results. This method is applied to more realistic conditions to test its robustness, and hence, allows for a wider application of probabilistic BEP FEM calculations.

Keywords: Dike, Backward erosion piping, DgFlow, D-Seepage, Probabilistic models