Multiscale characterization and modeling of granular materials through a computational mechanics avatar: a case study with experiment

Abstract

Through a first-generation computational mechanics avatar that directly links advanced X-ray computed tomographic experimental techniques to a discrete computational model, we present a case study where we made the first attempt to characterize and model the grainscale response inside the shear band of a real specimen of Caicos ooids subjected to triaxial compression. The avatar has enabled, for the first time, the transition from faithful representation of grain morphologies in X-ray tomograms of granular media to a morphologically accurate discrete computational model. Grain-scale information is extracted and upscaled into a continuum finite element model through a hierarchical multiscale scheme, and the onset and evolution of a persistent shear band is modeled, showing excellent quantitative agreement with experiment in terms of both grain-scale and continuum responses in the postbifurcation regime. More importantly, consistency in results across characterization, discrete analysis and continuum response from multiscale calculations are found, achieving the first and long sought-after quantitative breakthrough in grain-scale analysis of real granular materials.

All you need is shape: Predicting shear banding in sand with LS-DEM

“…Although clumping, polyhedra, and ellipsoid methods have had modest success in replicating experimental results at the bulk scale [33,31,54], they have not been compared to experimental results at lower length scales, such as that of strain localization. NURBS-based methods accurately capture particle shape at both length scales and can accurately simululate granular behavior in shear bands, but are computationally time-expensive and have only been able to simulate unit cells of about one thousand particles at most [34]. One method of overcoming these impasses in both continuum and discrete modeling is the avatar paradigm, able to both characterize and simulate the behavior of granular assemblies through a DEM framework that characterizes particle shapes ("avatars") directly from experimental images, can simulate laboratory-scale amounts of particles (over 50,000), and captures the same 3D strain localization trends seen in experiments.…”

Level set discrete element method for three-dimensional computations with triaxial case study

“…However, once fracture criteria and fracture planes have been determined, the level set framework is convenient for modeling grain breakage as fracture planes can be represented by level set functions which then can be used to split a grain using binary operations between the level set function of the grain and the level set functions of the fracture planes, which would allow replications of exact fracture patterns that occur in experiments. Another area in which LS-DEM could be applied is in multiscale methods, such as the one developed and implemented in [1,13], or using LS-DEM to infer continuum quantities such as dilatancy and macroscopic friction angle to shed light on how these continuum properties originate from the grain scale. Essentially, the potential applications of LS-DEM fall under the same umbrella as those of classic DEM, but with its ability to capture shape, LS-DEM will hopefully enable us to arrive at a deeper, more quantitative understanding of the behavior of granular materials.…”

Probabilistic calibration of discrete element simulations using the sequential quasi-Monte Carlo filter

“…The micromechanical force and contact networks calculated with DEM are typically averaged over the repeated representative volumes to extract macroscopic variables such as stress and fabric tensors [5]. Alternatively, these packings can be embedded at the Gauss integration points of a FEM mesh to derive the local material responses in a hierarchical FEM×DEM multiscale model [15,30]. It is well known that the microstructure of a granular material (e.g., coordination number and anisotropy [28,29]) plays a key role in the macroscopic constitutive behavior.…”