Prediction of the Single Pile Seismic Deflection by Using FEM and ANN

Abstract

A failure prediction assessment is generally required to enhance infrastructural seismic design. This study used Artificial Neural Networks (ANN) and the Finite Element Method (FEM) to predict the deflection of the concrete and timber piles. Two models were simulated in accordance with the analytical method, location, and type of seismic load applied to the pile. The removal of noise from the seismic data from the Savitzky–Golay filter was applied to improve the quality of the numerical simulation results. To assess the lateral deflection of the single pile during seismic excitation, soil-pile seismic resistance prediction is carried out. Finally, the accuracy level of the obtained non-linear pile's deflection is detected. The results show that timber and concrete piles exhibit different lateral deflection mechanisms. Removing noise from the seismic data improved the prediction of the p–y curve’s quality. The deflection mechanism of the pile model depends on the soil-pile interaction and pile material. The pile type controls the allowable lateral deflection and the rigidity of the pile impact on the deflection seismic response of the pile.

Keywords: FEM, ANN, Savitzky–Golay method, Pile deflection, Lateral displacement