In the present study, the estimation of vertical settlement of earthen dams caused by earthquakes has been investigated using an artificial neural network model and wavelet-artificial neural network combination. The accuracy of the methods used based on the crest settlement data obtained from 13 types of dams and a total of 151 earth dams with eight series of data was considered as input parameter and one output parameter. From a total of 151 earth dams, 121 samples were randomly selected and used for network training. Also, network validation was performed with 29 data sets, and the remaining sample was considered to evaluate the method's applicability. The results showed that the rbior6.8 wavelet function had the highest accuracy and best performance with a correlation coefficient of 83% and RMSE, 0.044, and the dmey wavelet function showed the lowest accuracy and the weakest performance with a correlation coefficient of 70% and 0.162, RMSE. Also, the Sym4 function with the correlation coefficient of 81% and the RMSE, of 0.55 provided the best performance after the rbio6.8 wavelet function in the second place. After Sym4, bior6.8 function with 78% accuracy, haar function with 77% accuracy, coif2 function with 73% accuracy, db4 function with 72% accuracy and finally dmey function with 70% accuracy. The proposed model with 2.75% error has a more accurate prediction than finite element methods with 19.92% and 4.06%. In total, the results of this combined model are considered suitable because the correlation coefficient varied between 70 and 83 percent.
