In the past, dams were for drinking water supply and agricultural purposes, but today due to hydroelectric power and other purposes has been further developed. Forecast seepage from the earth fill dam embankment is a complex and dynamic issue. In this study, an artificial neural network (ANN) used to estimate the seepage of the dam embankment, "Bukan Shahid Kazemi" paid. To achieve this goal, the piezometric data set includes 864 data were used. The results that the proposed model showed that the model quite successful in finding purpose and high- speed operation. Using statistical parameters and applications show that the proposed network is well train and has high potential in estimating leakage.