Reconstruction capability of the kernel-based stereological inversion applied to particle separation techniques

The paper tackles the problem of characterizing the particulates that derive from the comminution of ores or two-component materials in terms of fraction of particles within a narrow volumetric composition. It reports some refinements to the procedures of regularized-inversion for the solution of the kernel-based stereological model applied to the conversion of areal grade distributions (readable from polished sections by image analysis) into distributions of volumetric grades. These solutions are further exploited to infer the efficiency of separation/concentration systems treating particulates, an important information rarely obtained due to the complex experimental work required with traditional heavy liquid fractionating. The use of synthetic kernels is also discussed. Finally, one practical application with computer generated data demonstrates the resolution of the efficiency coefficients of a separation process. These are recovered solving a stereologically posed mass balance around the separator with an original regularizing strategy that allows including a priori information.