DOI: 10.3390/buildings16122442 ISSN: 2075-5309

A Double-Hardening Elastoplastic Load-Transfer Model for Assessing Load-Carrying Performance of Axially Loaded Piles

Yexun Li, Yunzhe Zhang, Haoyu Liu, Xian Wang, Song Qiu, Jian Yu, Lin Li

Accurate prediction of the load–settlement response of axially loaded piles remains challenging because the pile–soil interface undergoes progressive elastoplastic shear deformation accompanied by stress-dependent volumetric changes. Conventional one-dimensional load-transfer models are computationally efficient but usually rely on empirical or hyperbolic fitting functions, making it difficult to explicitly describe the coupled evolution of interface shear hardening, volumetric hardening, and radial effective stress. Although three-dimensional elastoplastic models provide a more rigorous mechanical representation, their high computational cost limits routine engineering application. To address this gap, this study develops a double-hardening elastoplastic load-transfer model for axially loaded piles based on a physically interpretable pile–soil interface constitutive formulation. In the proposed model, the Hardening Soil model is used to characterize interface shear hardening, while the Modified Cam-clay model is introduced to describe volumetric hardening. These two mechanisms are coupled through a stress–dilatancy relationship. According to the loading direction and the position of the current stress point relative to the shear and volumetric yield surfaces, the p′–q stress plane is divided into elastic, shear-hardening, volumetric-hardening, and coupled double-hardening regions. The corresponding incremental constitutive equations are derived and embedded into a conventional load-transfer framework. The model is validated using interface direct shear tests and field-scale static pile load tests. The predicted shear stress–displacement curves and pile-head load–settlement responses agree well with the measured data. Quantitative evaluation shows that the MAPE values are lower than 5%, the maximum relative errors are below 7.6%, and the R2 values exceed 0.96 for all validation cases.

More from our Archive