Abstract:
Excessive and uncontrolled seepage through an earth-fill dam is detrimental to its stability.
In this study, an empirical correlation was formulated for calculating the rate of seepage
discharge through a non-homogenous earthen dam resting on a pervious foundation. Seepage
analysis has been performed on SEEP/W program with three different values of each physical
and geometrical parameter of the embankment (upstream slope, dam’s height, downstream
slope, crest width, core’s slope, depth of pervious foundation, free board, permeability of
foundation material, and permeability ratio of core to shell material). Altogether 4374 dam
models were analyzed to obtain the seepage datasets. The comparison of seepage values
obtained from SEEP/W program with the analytical equations revealed an average percentage
error of less than 15 %. Non-linear regression analysis was then performed on the seepage
datasets using SPSS-19 software to obtain an empirical equation for seepage calculation.
Furthermore, a seepage prediction model was developed using artificial neural networks via
MATLAB software. The difference of seepage discharge determined from SEEP/W against
its amount computed from the developed empirical correlation and ANN model gave
coefficient of determination (R2) of 0.96 and 0.971 corresponding. Sensitivity analysis
indicates that a minor change in the dam’s height brings a huge change in the rate of seepage
(31.26%), while variation in the upstream slope has the least impact on the rate of seepage
(0.93%)
