Prediction of the Dynamic Soil-Pile Interaction under Coupled Vibration Using Artificial Neural Network Approach

Pile foundations resulting from dynamic excitation is a complex phenomenon involving complex pile-soil interaction. Pile foundations behave in a nonlinear fashion at large displacements because of the soil nonlinearity at high strain and pile separation. This paper attempts to study the dynamic soil-pile interaction on the nonlinear response under coupled vibration by both experimental and numerical study. This paper presents the small-scale field test results of single piles and 2 x 2 group piles subjected to different excitation intensities. Resonant frequencies and amplitudes are determined for both horizontal and rocking motion from the observed response curves. Artificial neural network (ANN) models are developed based on field test results for the prediction of dynamic behaviour of piles under coupled motion. Importance of different strategies in developing robust ANN models are explored and discussed. Different ANN models are developed using evolutionary learning algorithm and Bayesian regularization algorithm. Various statistical performance criteria and its importance to compare the developed ANN models are discussed. Different sensitivity analyses are made to identify the important input parameters for amplitude and resonance determination.