DEM modelling of mini-triaxial test based on one-to-one mapping of sand particles

This paper presents a discrete-element method simulation of mini-triaxial tests on a sand with realistically shaped grains. It compares the results with physical experiments at multiple length scales, including the macroscopic sample length scale and the particle scale. A series of image-processing techniques were utilised to binarise, segment and label the raw data in images obtained from the mini-triaxial test. The images were obtained using an X-ray synchrotron radiation scanner. A spherical harmonic analysis was used to filter the image data and to reconstruct the natural particle morphology. Two parameters, these being the radius ratio of the smallest to largest sphere $ρ$ and a characteristic distance $ϕ∈[0∘,180°]$ within the multisphere clump method, were chosen to represent the realistic particle morphology, balancing accuracy against computational cost. A one-to-one discrete-element model, where every particle in the physical experiment has its own numerical twin, was constructed. The discrete-element model was contained by a numerically generated flexible membrane allowing free deformation of the specimen under a prescribed confining stress, as in a physical triaxial test. Finally, attention was given to particle scale properties and their influences on the mechanical response of the discrete-element model. For a given strain rate it was found that shear modulus and friction coefficient affect the initial stiffness, the peak load and the dilation significantly. This study, and the simulation results within it, demonstrate that the proposed modelling approach is capable of reproducing macroscopic (e.g. stiffness, deviatoric stress response and volumetric response) and particle-level (e.g. displacement, rotation and branch vector orientation) behaviours that are very similar to what occurs within physical experiments, validating the effectiveness of the proposed one-to-one mapping technique.