Prediction of pile-separation length under vertical vibration using ANN

Abstract

This paper attempts to study the nonlinear soil-pile interaction under vertical vibration by both experimental and theoretical study. The field test results of single and group piles subjected to different excitation intensities are presented. The measured response is compared by the continuum approach of Novak with nonlinear solution. The soil properties of boundary zone and separation length at pile-soil interface used in this numerical methodology are fine-tunned by trial and error in order to match the experimental results. Artificial neural network (ANN) models are developed based on field test results and the pile separation length considered in the analysis. Different ANN models are developed using evolutionary learning algorithm and Bayesian regularization algorithm. Various statistical performance criteria are used to compare the developed ANN models. A sensitivity analysis is also made showing the effects of input.