Simulation of soil stress in earth dams using artificial intelligence models and determination of effective features

The general purpose of this paper is to select effective features and model soil stress in earth dams at the time of construction with the neural network using an optimization algorithm and then compare the results of the artificial neural network model with usual ANFIS and GEP methods. Five features including fill level, dam construction time, reservoir level (dewatering), dewatering speed and embankment speed were selected as hybrid model inputs. By performing hybrid algorithm and sensitivity analysis and feature selection method, fill level and dam construction time, the most effective features were in modeling the total stress in selected cells, because the dual composition including fill level and construction time in TPC25.1, TPC25.3 and TPC25.4 cells, The error values (MSE) of 1.523, 2.747 and 0.750 were the most effective features in these cells, respectively. In TPC25.2 cell, the selection of three features including fill level, construction time and dewatering level according to the error (MSE) value of 5.245, has the greatest effect in modeling the total soil stress in this cell. Comparison between ANN model with ANFIS and GEP showed that although the difference in the accuracy of the models is very small, it can be said that all three models had acceptable answers. The results also show that the higher the dispersion of the model input data, the more the ANFIS model has the ability to simulate than the two models ANN and GEP,