Mechanical Properties of Granular Metamaterials – MICROGRAM

From an experimental point of view, the first challenge in studying these grains is their mass production. We set up a semi-industrial production process using the same technique that produces nearly all the plastic objects around us: injection molding. 3D printing was used as a complement to produce specific shapes, though with limitations on the surface properties of the grains. To observe these packings, grain by grain, since they are not transparent, we employed a medical imaging technique: X-ray scanning. It was then necessary to design, on the one hand, the devices that allow mechanical loadings inside these imaging systems, and on the other hand, the 3D image processing algorithms to isolate each grain, its orientation, and its contacts in space. From a numerical perspective, the challenge was no less significant, as the available tools could not process such a large number of grains with such complex geometries. We had to develop LMGC90, a granular simulation tool designed and developed in our laboratory for the past twenty years, to enable it to simulate the behavior of grains with complex geometries.

Our work, strengthened by its triple approach, has allowed us to better understand what induces cohesion in exotic granular materials. Furthermore, when this cohesion is lost, we have been able to establish a behavior law that describes their flow in various geometries. The accuracy of these analytical formulas, which describe all the observables of the system during its flow, has been tested through numerical simulations and experimental measurements. Beyond our expectations, these models have shown that they remain valid even for grains with very simple shapes. This thus concludes scientific questions that have lasted for 30 years.

These results are and will be published in six articles. Four manuscripts have already been submitted, two of which have been accepted. Among them is an article in Physical Review Research, presenting the analytical model that describes the evolution of density, stress, and velocity profiles in a rotating drum flow, regardless of the particle geometry. These results have received an enthusiastic reception from the granular materials community, notably during their presentation at the latest Gordon Conference.

The aim of this project is to predict the mechanical properties of granular metamaterials. Granular packings consist of unbound macroscopic, solid particles. They have been widely studied when composed of spherical or slightly non-spherical but convex particles. Their mechanical properties are, however, mainly defined by the local interactions between the individual particles and, thus, by the shape of the particles. Therefore, significant deviations from spherical and convex particle shapes can induce so far unexplored properties of the granular packings and suggests a new family of granular material whose mechanical properties can be tuned by tailoring the shape of the constituting particles: granular metamaterials. Recently, such granular packings of complex shaped particles have been investigated e.g. as construction material. Due to the complexity introduced by the particle shape, general constitutive relations between the shape of the individual particles and the macro-mechanical properties of the packing is not feasible. We therefore plan to approach the problem by means of extensive numerical simulations which will be validated by model experiments. The required particle based simulations will be optimized such that the resulting contact networks coincide with experiments. These experiments will be performed for defined model particles and loading geometries (shear and compression). To calibrate the simulation methods, we will capture the granular packings by means of X-ray computed tomography and subsequently obtain the structure, the topology and the contact network of the packing by segmenting the individual, complex shaped particles from the CT-data. Once reliable, the simulations will be used for the mass production of data for a wide range of particle shapes. This data will permit to train an artificial neural network that will output the mechanical properties of a granular metamaterial as a function of the shape of the constituting particles.