Prediction of soil-pile-interaction in liquefiable ground conditions: from single element calibration to system response

Finite element analyses are an important tool for validating pile designs in complex seismic loading scenarios involving soil liquefaction. Numerous constitutive models are available in the literature and can be calibrated to capture select features of the soil response, typically the numbers of cycles to liquefaction. Even though a close comparison between constitutive models and experimental data can be achieved in terms of liquefaction resistance in element test conditions, their distinct stress-strain behaviour and limitations in reproducing dynamic soil properties affect the complex soil-structure interaction. The choice of the constitutive soil model and its impact on the global system response thus remains a key uncertainty. This paper provides a comparison of two bounding surface plasticity models, calibrated to the same laboratory data, and employed in a seismic soil-pile-interaction problem with otherwise identical conditions. The investigation includes an effective stress-based hydraulic conductivity model used to capture liquefaction-induced surface settlements. A centrifuge experiment of a single slender pile subjected to a seismic base excitation serves to validate the numerical analyses. The paper details which aspects of the global system response are most affected by the constitutive model choice. A minimal impact is observed on the far-field ground settlements and the predominant frequency of the system, whereas significant differences occur in excess pore pressure distributions in the near field of the pile and in the acceleration amplitudes. The prediction of lower excess pore pressures adjacent to the pile is herein linked to larger bending moments and smaller permanent displacements of the pile head.