Fully Stochastic Seepage Analysis through Spatially Randomly Distributed Embankment Dams

Spatial random field applications of soil properties integrated with finite element modeling tools are substantial in civil engineering to delineate stochastic characteristics of system responses that reflect real site conditions due to uncertainty. This research combines a random finite element model of a homogeneous hypothetical embankment dam’s hydraulic conductivity in the x and y directions and a stochastic upstream reservoir head on the upstream boundary. Statistical parameters of random hydraulic conductivities (mean, coefficient of variation, and horizontal–vertical scale of fluctuation) and probability distribution characteristics of stochastic boundary conditions are derived from available literature. The Karhunen–Loève expansion method is utilized in random field generation. Seepage fluxes and hydraulic gradients through the embankment dam are computed by Monte Carlo simulations. Sensitivity analyses are conducted to interpret the influences of correlation distances and coefficient of variation (COV) on the seepage flow through the embankment dam. Fully stochastic analyses of seepage also provide ranges of exceedance probability to evaluate piping risk conditions rather than assessing a single probability value. Results demonstrate that the effects of autocorrelation have minor importance when compared to COVs in both directions. It is of particular importance to compare fully stochastic seepage analysis for various scenarios and deterministic analysis to perform reliability-based design for decision-making processes.