J., *Vicente, *Reimann, J., Ferrero, C., Rack, A., Gan, Y.
The knowledge of the packing structures of granular beds is of importance for many technical applications where heat and mass transfer processes are influenced by non-homogenous porosity distributions. More fundamentally, the development of regular packing structures packing is of interest for the simulation of crystallisation processes.
3d tomography offers the unique possibility to investigate experimentally in detail the arrangement of individual granular particles within a container, the contacts with other particles and walls, contact angle distributions and even contact surfaces. To extract quantitative data from the tomography images, a first step of thresholding is necessary to segment the spheres from the void space. Because of the polydispersity of the spheres, no assumption based on a mean sphere diameter can be taken to help the sphere identification process. We propose a watershed segmentation method based on the maximal balls computation. A simple scanning algorithm is then used to identify the contact points between spheres.
In this paper, prismatic containers (80x80x20mm) are investigated, filled with either A) mono-sized aluminium spheres (diameter d= 2.3mm; ≈11000 spheres) or B) ceramic spherical particles with a diameter spread (0.8<d(mm)<1.2; 105000 spheres). Prior to vibration, the containers were filled though a small opening in the corner of the container. The container was inclined in such a way that the filling opening was at the highest elevation.
For A), a packing fraction of 0.68 was obtained which is very close to the maximum value for hexagonal dense packing of hex=0.69 for this geometry. The tomography images visualize impressively this regular packing order; quantitatively this is confirmed by calculated void fraction distributions, contact numbers, and radial distribution function. In contrast are the results for B): here is only ≈0.62, regular structures are restricted to wall layers of thicknesses of ≈4d, in the largest volume fraction, a rather non-ordered packing exists. An important result, however, is that distinct segregation effects are detected, that is the diameter distribution of spheres in different zones (wall layers, bulk zones) differs from the nominal distribution. This fact is of importance for technical applications where unhomogeneities are undesired.