The development of the discrete element method (DEM) in computational geomechanics, civil engineering, and mining has facilitated advanced modeling and understanding of the macroscopic behavior of granular materials in terms of micro and meso scale mechanisms. In particular, this approach enabled modeling of particle shapes and interactions using non-uniform rational B-splines (NURBS) and now Level Sets (LS), that allowed the inclusion of arbitrary grain shapes with a relatively low computational cost. As of now, it is evident that the inclusion of arbitrary shapes in DEM simulations is of key importance to more accurately capture and reproduce the grain-scale physics that furnishes the bulk behavior of granular materials.
Following recent technological advances in 3DX-ray Computed Tomography scanning (3DXRCT), utilization of such novel properties lent by Level Set DEM schemes have enabled researchers to capture real grain shapes, such as Hostun sand with very high detail. Unfortunately, costly equipment and specialized personnel are required to perform the 3D X-ray CT scans. In addition, the size of the scanned samples is small compared to the amount of the material required to reproduce macroscopic/engineering behavior.
Recently, Prof. Alex X. Jerves and his research assistant David A. Medina at Escuela Superior Politécnica del Litoral (EPOL) and Universidad San Francisco de Quito (USFQ) generated new “cloned” grains, that could accurately capture and reproduce the morphological features displayed by any given sample of real granular materials. They hoped that such a development could be used as part of DEM based engineering-scale simulations and help in overcoming sample size constrains inherent to 3DXRCT scanning processes. Their work is currently published in the research journal, Granular Matter.
To begin with, the two scientists computed the multivariable probability density functions from the parents’ morphological parameters (morphological DNA). Next, an improved, now parallelized and better tuned version of the geometric stochastic cloning algorithm, was employed. They then extracted the morphological DNA from the new “cloned” grains and compared to the one obtained from the parent sample. The research pair then subjected the clones and the parents to triaxial compression tests using a level set discrete element scheme (3DLS-DEM), and then, compared in terms of their mechanical response. Lastly, the error of the “clones” in the morphology and mechanical behavior was analyzed.
The authors observed that the clones 2.0 morphologically matched much more accurately to their parents than the ones generated using the original version of the algorithm. This was revealed following a comparison between the distributions of clones 1.0 and 2.0 morphological DNA. Generally, the improved approach was seen to produce “cloned” grains that more accurately approached the morphological features displayed by their parents.
In summary, the study presented an improved version of a computational algorithm that enabled “cloning” of an arbitrary number of grains based on grain morphologies of a real sample of digitalized grains (avatars). In general, the researchers highlighted that the yielded grains not only satisfied their parents morphological features in a more accurate way than the original version, but they also showed a similar mechanical behavior once “clones” and parents were included into a triaxial compression test.
David A. Medina, Alex X. Jerves. A geometry-based algorithm for cloning real grains 2.0. Granular Matter (2019) 21:2Go To Granular Matter