This article presents the findings of a comprehensive assessment of the predictive capabilities and limitations of advanced geotechnical numerical tools utilizing two sophisticated constitutive models for sands: the hardening soil model with small strains and hypoplasticity with intergranular strain. The evaluation is based on simulations of laboratory and centrifuge tests under monotonic and cyclic loading conditions. Initially, these models were calibrated and assessed using an experimental database on Fontainebleau sand. This database encompasses a range of laboratory results, including isotropic compression, drained monotonic triaxial, and undrained cyclic triaxial tests with varying initial conditions. The models, in general, provided good representation for monotonic experiments while some discrepancies were observed in undrained cyclic experiments. Subsequently, the calibrated models were employed to replicate a series of centrifuge tests involving a pile embedded in the same sand. The pile was subjected to various episodes of monotonic and cyclic lateral loading. In general, the models accurately replicated the experimental observations from tests conducted under monotonic loading conditions. Some small discrepancies were found in pile tests subjected to cyclic loading, these were however minor when compared to issues in predicting cyclic element tests at undrained conditions.

Landslide dams consist of unconsolidated heterogeneous material and lack engineering measures to drain water and control pore water pressure. They may be porous and seepage through them could potentially lead to piping failure. In this research, the internal processes within a long-existing landslide dam are assessed under transient seepage force. The implemented approach includes a 3D finite element numerical simulation executing fully coupled flow-deformation and consolidation methods based on hydraulic data measurements and geotechnical laboratory tests. The nonlinear constitutive model 'Hardening Soil' is applied to accurately calculate the stressinduced pore water pressure, effective stress, deformation, and flow. Further, the possibility of slope failure due to seepage force is investigated through the strength reduction method. The results highlight the dependency of the seepage flow on the corresponding variation of the relative permeability and saturation in the soil mediums under different rates of seepage force. Small rates of seepage force, however, impose deformation at the dam's crown. High effective stress is obtained at negative small rates of seepage force where the long duration of fluctuation is modeled. In the drawdown simulation, there is a reverse relation between effective stress and the rate of the seepage force. Through the modeling process and based on the measured data, two seepage paths are detected within the landslide dam, while their activation depends on the lake level. The modeling approach and the required data analysis are suggested for utilization in further studies regarding the seepage process understanding at the long-existing landslide dams and their hazard assessments in addition to the common geomorphological approaches.

The original elastoplastic Hardening Soil model is formulated actually partly under hexagonal pyramidal MohrCoulomb failure criterion, and can be only used in specific stress paths. It must be completely generalized under Mohr-Coulomb criterion before its usage in engineering practice. A set of generalized constitutive equations under this criterion, including shear and volumetric yield surfaces and hardening laws, is proposed for Hardening Soil model in principal stress space. On the other hand, a Mohr-Coulumb type yield surface in principal stress space comprises six corners and an apex that make singularity for the normal integration approach of constitutive equations. With respect to the isotropic nature of the material, a technique for processing these singularities by means of Koiter's rule, along with a transforming approach between both stress spaces for both stress tensor and consistent stiffness matrix based on spectral decomposition method, is introduced to provide such an approach for developing generalized Hardening Soil model in finite element analysis code ABAQUS. The implemented model is verified in comparison with the results after the original simulations of oedometer and triaxial tests by means of this model, for volumetric and shear hardenings respectively. Results from the simulation of oedometer test show similar shape of primary loading curve to the original one, while maximum vertical strain is a little overestimated for about 0.5% probably due to the selection of relationships for cap parameters. In simulation of triaxial test, the stress-strain and dilation curves are both in very good agreement with the original curves as well as test data.