Identifying the Inherent Variability of Granular Filters Through Graph Concepts

Abstract Filter design aims to trap finer particles within the pore space of coarser particles by meeting macro-scale geometric criteria (particle size distributions or constriction size). These criteria fail to account for the significant effect of filter particle positions on filter performance. The current contribution evaluates the variability in filter structure using pore networks generated from discrete element method filtration simulation data. The pore network edges are directed and weighted using the constriction gravity-bias pore area. A study of the network maximum-flow and minimum-cost asserted the ability of graph characteristics to identify the statistically different filter samples. This demonstrates the potential for graph characteristics to link particle-scale behaviour to macro-scale observations.