Multidimensional characterization of separation processes – Part 1: Introducing kernel methods and entropy in the context of mineral processing using SEM-based image analysis

An alternative method for the particle tracking approach for scanning electron microscopy-based image analysis is introduced, using kernel density estimates instead of discrete bins. This allows for information that is more robust. Uncertainties of the data are assessed using the bootstrap resampling method. The presented methodology enables the calculation of multidimensional partition curves, which can be used for a detailed analysis of separation processes. It has been found that the statistical entropy is a helpful tool to evaluate the separation efficiency of these partition maps. The methodology was applied to a density separation process of a cassiteritebearing skarn ore from the Hämmerlein deposit in the Erzgebirge region in Germany, which serves as a case study. A Sepro™ Falcon concentrator was utilized for the density separation.