Pile response in submerged lateral spreads: Common pitfalls of numerical and physical modeling techniques

A three dimensional dynamic numerical methodology is developed and used to back-analyze experimental data on the seismic response of single piles in laterally spreading slopes. The aim of the paper is not to seek successful a-priori (Type A) predictions, but to explore the potential of currently available numerical techniques, and also to get feedback on modeling issues and assumptions which are not yet resolved in the international literature. It is illustrated that accurate simulation of the physical pile-soil interaction mechanisms is not a routine task, as it requires the incorporation of advanced numerical features, such as an effective stress constitutive soil model that can capture cyclic response and shear-induced dilation, interface elements to simulate the flow of liquefied ground around the pile and proper calibration of soil permeability to model excess pore pressure dissipation during shaking. In addition, the "conventional tied node" formulation, commonly used to simulate lateral boundary conditions during shaking, has to be modified in order to take into account the effects of the hydrostatic pore pressure surplus that is created at the down slope free field boundary of submerged slopes. A comparative analysis with the two different lateral boundary formulations reveals that "conventional tied nodes", which also reflect the kinematic conditions imposed by laminar box containers in centrifuge and shaking table experiments, may underestimate seismic demands along the upper part of the pile foundation.