In [1] we suggested an original Discrete Element Method that offers the capability to consider non-spherical particles of arbitrary convex shape. We elaborated on the foundations of our numerical method and validated it on assorted test cases. However, the implementation was serial and impeded to examine large systems. Here we extend our method to parallel computing using a classical domain decomposition approach and inter-domain MPI communication. The code is implemented in C++ for multi-CPU architecture. Although object-oriented C++ offers high-level programming concepts that enhance the versatility required to treat multi-shape and multi-size granular systems, particular care has to be devoted to memory management on multi-core architecture to achieve reasonable computing efficiency. The parallel performance of our code Grains3D is assessed on various granular flow configurations comprising both spherical and angular particles. We show that our parallel granular solver is able to compute systems with up to a few hundreds of millions of particles. This opens up new perspectives in the study of granular material dynamics.

Large-scale numerical simulation using the Discrete Element Method (DEM) contributes to improve our understanding of granular flow dynamics involved in many industrial processes and geophysical flows. In industry, it leads to an enhanced design and an overall optimisation of the corresponding equipment and process. Most of DEM simulations in the literature have been performed using spherical particles. A limited number of studies dealt with non-spherical particles, even less with non-convex particles. Even convex bodies do not always represent the real shape of certain particles. In fact, more complex shaped particles are found in many industrial applications as, e.g., catalytic pellets in chemical reactors or crushed glass debris in recycling processes. In Grains3D-Part I [39], we addressed the problem of convex shape in granular simulations while in Grains3D-Part II [33], we suggested a simple through efficient parallel strategy to compute systems with up to a few hundreds of millions of particles. The aim of the present study is to extend even further the modeling capabilities of Grains3D towards non-convex shapes, as a tool to examine the flow dynamics of granular media made of non-convex particles. Our strategy is based on decomposing a non-convex shaped particle into a set of convex bodies, called elementary components. We call our method glued or clumped convex method, as an extension of the popular glued spheres method. Essentially, a non-convex particle is constructed as a cluster of convex particles, called elementary components. At the level of these elementary components of a glued convex particle, we employ the same contact detection strategy based on a Gilbert-Johnson-Keerthi algorithm and a linked-cell spatial sorting that accelerates the resolution of the contact, that we introduced in [39]. Our glued convex model is implemented as a new module of our code Grains3D and is therefore automatically fully parallel. We illustrate the new modeling capabilities of Grains3D in two test cases: (i) the filling of a container and (ii) the flow dynamics in a rotating drum.

In many dry granular and suspension flow configurations, particles can be highly non-spherical. It is now well established in the literature that particle shape affects the flow dynamics or the microstructure of the particles assembly in assorted ways as e.g. compacity of packed bed or heap, dilation under shear, resistance to shear, momentum transfer between translational and angular motions, ability to form arches and block the flow. In this talk, we suggest an accurate and efficient way to model collisions between particles of (almost) arbitrary shape. For that purpose, we develop a Discrete Element Method (DEM) combined with a soft particle contact model. The collision detection algorithm handles contacts between bodies of various shape and size. For nonconvex bodies, our strategy is based on decomposing a non-convex body into a set of convex ones. Therefore, our novel method can be called "glued-convex method" (in the sense clumping convex bodies together), as an extension of the popular "glued-spheres" method, and is implemented in our own granular dynamics code Grains3D. Since the whole problem is solved explicitly, our fully-MPI parallelized code Grains3D exhibits a very high scalability when dynamic load balancing is not required. In particular, simulations on up to a few thousands cores in configurations involving up to a few tens of millions of particles can readily be performed. We apply our enhanced numerical model to (i) the collapse of a granular column made of convex particles and (i) the microstructure of a heap of non-convex particles in a cylindrical reactor.

Random packed beds of cylindrical, trilobic and quadrilobic particles in cylindrical and bi-periodic containers are numerically studied using Grains 3D, a code based on the Discrete Element Method (DEM) that resolves all inelastic collisions and simulates dynamically the loading of packed beds. To mimic industrial or laboratory packing 1 procedures, particles initial position and orientation are random so that the same simulation repeated again yields a different packed bed structure and thus a different average void fraction. These "in silico" experiments aim at being able to optimize particle shape in heterogeneous catalysis, in particular with respect to the corresponding bed void fraction that is a critical parameter for pressure drop prediction. These "in silico" experiments are deterministic and accurate but with differences due to the loading procedure. In this paper, we first present our assessment of the uncertainty on average void fraction induced by (i) the initial random position and orientation of inserted particles and (ii) the insertion zone size. Next we investigate the effect of particle shape, namely cylindrical, trilobic and quadrilobic on the average void fraction as a function of particle length and diameter, and of the container type. Simple correlations are proposed that describe very well the simulations within the aforementioned uncertainty related to the packing procedure. While the beds made with cylindrical particles are markedly denser, the beds made of the trilobic and quadrilobic particles have statistically an identical void fraction.