Indexed Parallel Sphere Packing for Arbitrary Domains

Abstract

Particle packings are used to simulate granular matter, which has various uses in industry. The most outstanding characteristics of these are their density and their construction time, the density refers to the percentage of the space of the object filled with particles, this is also known as compaction or solid fraction. Particle packing seeks to be as dense as possible, work on any object, and have a low build time. Currently there are proposals that have significantly reduced the construction time of a packing and have also managed to increase the density of these, however, they have certain restrictions, such as working on a single type of object and being widely affected by the characteristics of the object. The objective of this work is to present the improvement of a parallel sphere packing for arbitrary domains. The packing to improve was directly affected in time by the number of triangles in the mesh of object. This enhancement focuses on creating a parallel data structure to reduce build time. The proposed method reduces execution time with a high number of triangles, but it takes up a significant amount of memory for the data structure. However, to obtain high densities, that is, densities between 60% and 70%, the sphere packing construction does not overwhelm the memory.