Probabilistic Analysis of Velocity Distribution under Earth Embankments for Piping Investigation

Abstract

Piping has been attributed as a major cause of dam and embankment failures. Current national prediction models to predict piping use the hydraulic gradient between the upstream and downstream water levels as an indicator. They are deterministic and take no account of preferential pathways. However, piping usually initiates from a discrete downstream location and recent research indicates that piping happens when the water velocity reaches a critical value. The local velocity close to the downstream ground surface is strongly related to the inherent heterogeneity of the soil. This paper takes account of the spatial variability of hydraulic conductivity. Based on stochastic seepage analysis, local velocities in and under an embankment are computed. The results show that, when the coefficient of variation of hydraulic conductivity is small, the locations of the maximum local velocity from 500 realizations aggregate into a small area which is near the downstream toe of the embankment. In contrast, increasing the coefficient of variation scatters the possible locations of the maximum local velocity. Two typical situations are identified: one situation where the maximum local velocity is close to the downstream ground surface, and the other where the maximum local velocity is far from the downstream ground surface and located near the center of the foundation. In the first situation it is easier to reach the critical value to initiate piping because the maximum local velocity is near the ground surface, but it is also relatively easy to protect against. In contrast, in the second situation it is easier to form a passage for piping development once piping is initiated. This is due to the velocity towards the center of the foundation increasing.