Evaluation of geotextile filter application in embankment dams with artificial neural network

In the past three decades Geosynthetics have been widely used for many different civil engineering applications. Geotextiles are flexible permeable geosynthetics used in geotechnical and hydraulic engineering structures. First use of geotextile in embankment dam was in 1959 in Cotrada Sabetta, in Italy (FEMA, 2011). But first use as filter was in Valcoros dam in France in 1970(Faure, et. Al. 1999). In this paper using geotextile as filter in embankment dams is evaluated, using 384 Geostudio SEEP W models and analyzing the results with artificial neural network in MATLAB. Input parameters are: crest of dam, height of dam, bottom width of dam, reservoir water level, permeability of core, top and bottom thickness of dam core, numbers of geotextile layers, and PP geotextile type (200, 400, 500, and 800gr⁄m^2 and the target function is the ratio of the flux with geotextile to flux for those without geotextile in the middle of dam foundation. Using ANN (Artificial Neural Network) helps us to simulate the flux. For any other embankment parameters or other geotextiles without using different permeability of geotextile in different pressures. Calculating the minimum error (MSE) for ANN with two layers and different number of neurons, the simulations are in a good agreement with the model results.