Numerical simulation of the coupled hydro-mechanical response of the Leendert de Boerspolder dyke

Abstract

This study was conducted in the framework of the Leendert de Boerspolder stress test. It focuses on the investigation the performance of the available constitutive models in describing the coupled hydro-mechanical response of peats and organic clays as observed during the stress test. The soil models that are considered are the Mohr-Coulomb model, the Soft Soil model and the Hardening Soil model as are implemented in PLAXIS, a standard commercial finite element code.
The evaluation of the models is performed in two steps. First, the performance of the constitutive models is evaluated by simulating the laboratory tests as single soil element tests with PLAXIS SoilTest facility. Based on the comparison of the numerical results with the laboratory data it is concluded that the HS performs the best compared to the SS and the MC model. In order to achieve good fit it is found that it is necessary to drastically reduce the failure stress ratio, $Rf$ to an average value of 0.15 in contrast to what is mentioned in literature for soft soils. In terms of one-dimensional compression stress path both the HS and SS model are deemed to perform similarly. The MC model is found to reproduce poorly the laboratory tests due to the assumption of linear elasticity - perfect plasticity.
Subsequently, soil models are evaluated through a fully coupled hydro-mechanical simulation of the Leendert de Boerspolder stress test in PLAXIS 2D. The evaluation is done through the comparison of the measured to computed displacements and pore water pressure. It is found that the prediction of the HS model is ``soft'' for peat while the stiffness degradation in the organic clay results in excessive lateral displacements. Response of the SS is found to be better considering both displacements and pore water pressure. The best description of the stress test was found to be possibly by using the SS for the organic clay and the HS for peat. The performance of the MC model is quantitatively good however qualitatively is deemed to be poor. Furthermore, the influence of (a) the soil anisotropy and (b) the interface between organic clay and peat layers are pointed out as factors influencing the outcome of the simulation.
Based on this study, it is concluded that in general the Soft Soil model, at this stage, is recommended for use for both soils. The Hardening Soil model should be used for peat but with caution and mainly when the deviatoric strains are deemed to be important. In this case the calibration should be done focusing on triaxial tests, therefore compromising the oedometric response. Moreover, a high secant stiffness should be considered to describe peat. That might be justifiable due to presence of fibers which under tensioning provide additional stiffness. Moreover, results suggest that the use of the HS model for the organic clay is not justifiable. Finally, the Mohr-Coulomb model should be used only as a rough approximation.