Computation of stresses in concrete gravity dam under seismic loading through ANN and FEM

摘要

In this study, an Artificial Neural Networks (ANN) model is designed and verified for quick estimation of the structural parameter (maximum principle stresses at toe and heel) developed on a concrete gravity dam under earthquake loading. The neural network model "neuro-modeler" is designed to predict the response parameters of dam. Five different typical geometric configurations are used in the data generation for training of neuro-modeler. Several factors as geometric configuration representative parameters are proposed and successfully used in the study. Designed "neuro-modeler" is tested for several earthquake records. The results showed an excellent capability of the model to predict the outputs with high accuracy and reduced computational time.