R as an environment for data mining of process mineralogy data: A case study of an industrial rougher flotation bank

Through a series of in-house routines of R, an open source programming language for statistical computing, statistical analysis is applied to automated process mineralogy data in order to describe the performance of an industrial scheelite rougher flotation bank. These routines allow: 1) exploring all particles properties over residence time, not only particle size or surface liberation but also mineral association and a wealth of other particle properties, 2) to free the user from the limitations of the menu-driven built-in mineralogy software or spreadsheets, for calculation, data plotting or predictive model fitting, in particular for the parallel analysis of several streams; and 3) a more flexible manipulation of the data, both class and particle wise, for instance allowing for data mining across streams.

In an illustration case study, these functions are used to show the separation efficiency shift over residence time and over particle size; to indicate which associated minerals have a greater influence on the flotation of scheelite; to determine which gangue minerals are more impacted by entrainment; and finally to link said entrainment to particle shape. In general, the Helmholtz Institute Freiberg for Resource Technology intends to use such a programming platform on automated mineralogy data as a routine to understand processes better, as a potential diagnostic tool for process troubleshooting, and also for predictive model building within the frame of geometallurgy.