Advanced hypoplastic contact modelling of field tests for tension piles in layered soils

Conventional Coulomb friction models often fail to reproduce the complete non-linear load–displacement response of soil–pile interfaces, particularly during unloading–reloading sequences. Advanced hypoplastic contact formulations address these limitations by capturing pressure dependency (barotropy), density dependency (pyknotropy), and surface roughness effects using the same constitutive parameter set as the surrounding soil continuum. This study presents the first three-dimensional back-analysis of a hypoplastic contact model benchmarked against full-scale tensile load tests on steel H-piles in a complex, 23-layer stratified soil profile. The methodology establishes a physically consistent calibration framework where the interface roughness parameter is derived directly from torsional interface shear tests on pile-specific steel, rather than empirical estimation. These same contact formulations are subsequently applied to simulate the laboratory tests, ensuring consistency between calibration and field-scale prediction. The model accurately reproduces the field response and reduces the root mean square error (RMSE) by 71% relative to the Coulomb model, capturing hysteretic loops that simplified models miss. A parametric sensitivity analysis identifies interface roughness and pile circumference as the dominant capacity drivers, while relative density and in-situ earth pressure exert a moderate influence. The results demonstrate that physically based contact formulations are essential for accurate pile-soil interaction modelling in complex profiles, enabling more reliable and economical foundation design. • First 3D hypoplastic contact model simulation of pile tests in 23-layer soil profile. • Interface roughness calibrated via 3D torsional shear analysis. • Hypoplastic contact reduces prediction error (RMSE) by 71% vs. Coulomb. • Model captures non-linear hysteresis and state-dependent stiffness. • Surface roughness and pile geometry identified as key capacity drivers.