Bonded-particle extraction and stochastic modeling of internal agglomerate structures
The discrete element method (DEM) is an effective computational technique that is used to investigate the mechanical behavior of various particle systems like, for example, agglomerates. However, for systems of perfectly spherical and non-overlapping particles, the structural input is almost always based only qualitatively on experimentally observed structures. In this paper, we consider the case of agglomerates where particles are nearly spherical and connected by bonds. A novel bonded-particle extraction (BPE) method is proposed for the automated approximation of such agglomerate structures from tomographic data sets.
This method can be effectively used in conjunction with various commercial or open-source DEM simulation systems. By BPE, sphere-like primary particles are represented each by exactly one (perfect) sphere, and the set of spheres is non-overlapping. Furthermore, the solid bridge bonds between primary particles are retained. Having derived such a simple description of complex tomographic data sets, one can perform DEM simulations with well-established models like the bonded-particle model. Moreover, it is shown that a larger data base of statistically equivalent microstructures can be generated by a stochastic modeling approach. This approach reduces the need for (time-consuming) experimental agglomerate production and characterization.
Keywords: agglomerate structure, bonded-particle model, bonded-particle extraction, segmentation, stochastic microstructure model.