Forecasting nonlinear dynamic response of layered soils due to earthquake motion using artificial neural network

Abstract

ABSTRACTSeismic response analysis of layered soils which is gained from earthquake motion must be considered in practical engineering simulations. In order to predict this nonlinear dynamic response, artificial neural network (ANN) models were used in this research. The white noise earthquake was used as an induced earthquake to train and test a neural network against a numerical model (also subjected to the white noise earthquake). The numerical model was previously calibrated against field data from a specific earthquake. In order to ensure the accuracy of training, the nonlinear dynamic response of the optimal ANN model was tested against the field observation data from the specific earthquake. To validate the method's ability, the optimal model was subjected to nonlinear dynamic analysis under numerous earthquakes with different ground motions. The proposed model has shown high predicting performance for estimating soil dynamic deformations using the data obtained from the numerical software under different earthquake excitations. Furthermore, the model has been calibrated by matching the results obtained from the nonlinear seismic response with the response recorded under the actual earthquakes. The results suggested that the proposed model can be used as a suitable method for predicting nonlinear dynamic response of layered soils.