Influence of Sample Reconstitution on Advanced Model Calibration

Abstract

The simulation of granular soil using advanced soil models relies on the precise determination of model parameters from laboratory or field experiments, or both. Although it is well known that sample reconstitution influences the stress–strain response of granular soil, the effect of different reconstitution methods on the calibration of advanced soil models has not been previously comprehensively addressed. In the present study, a hypoplastic model was calibrated from geotechnical laboratory tests on Cuxhaven sand samples. Triaxial tests were performed on samples having been reconstituted by three different methods: moist tamping, vacuum pluviation, and horizontal vibration. The influence of different reconstitution methods on the simulation performance of the hypoplastic model was quantified through three scenarios of increasing complexity: (1) stress–strain behavior in representative element volume (REV), that is, triaxial tests; (2) in situ plate load test (PLT); and (3) laboratory cone penetration test (CPT). The latter two are typical examples of boundary value problems with different strains and stiffnesses. The adopted reconstitution methods significantly affected the REV simulation at high relative density of both peak friction angles and peak dilation angles. The reconstitution methods have a limited effect on boundary value problem simulations, being moderate for PLTs and small for CPTs. The influence of stiffness (intergranular strain) parameters on simulation results increased from REVs (no influence detected), over boundary value problems with low strain and stiffness (PLTs) to boundary value problems with high strain and stiffness (CPTs).