Predictive modeling of static and seismic stability of small homogeneous earth dams using artificial neural network

Abstract

The previous results of artificial neural network (ANN) prediction models in natural slope stability are encouraging, which gives logical hope for the practical application of these models for earth dams. In this context, this study presents an ANN model which allows the user to get four factors of safety (FS(i)) of small earth dams under long-term stability condition with static or earthquake loading immediately. Safety factors (FS(i)) for the model are calculated by carrying out two computations each for two cases, that is, earth dam subjected to full reservoir steady-state seepage condition with and without earthquake (FS(F + EQ) and FS(F)) and earth dam empty reservoir with and without earthquake (FS(Em + EQ) and FS(Em)). A database of 1372 different inputs and 4 outputs was built through strength reduction finite element method (SR-FEM). The used ANN is a feed-forward back-propagation neural network (FBNN) with three layers. The most appropriate FBNN architecture was found 11–21–4, as this gave the best FS(i) prediction with the lowest error. The relative importance of the inputs parameters is studied using both Garson’s algorithm and connection weight approach. Moreover, for further verification, the developed model has been used for prediction FS(i) of new earth dam datasets. The predicted results have been compared with the obtained ones from different limit equilibrium (LE) slope stability computations. The comparison had confirmed a very satisfactory capability of the ANN model to predict the FS(i).