A mineralogy characterisation technique for copper ore in flotation pulp using deep learning machine vision with optical microscopy

Among the flotation system process variables, mineralogy is the most difficult one to measure online. Mineralogy is typically measured through methods like Mineral Liberation Analysis (MLA) and QEMSCAN but these require sample preparation in polished sections only providing results after days or shifts. Alternatively, process plants utilise X-Ray Fluorescence (XRF) or Laser Induced Breakdown Spectroscopy (LIBS) to measure elemental grades online. However, the flotation performance is dictated by surface liberation of minerals rather than elemental grade.Recently, researchers have tried using optical microscopy to characterise mineralogy for an isolated particle, but this is not scalable for measuring process streams. This study investigates a technique utilising deep learning machine vision and optical microscopy for in-pulp characterisation of mineralogy and particle size distribution for multiple minerals in a copper ore pulp. The methodology was developed on samples from a polymetallic deposit in New South Wales, Australia that contained Cu, Pb, Zn, and Fe sulfides. This technique can predict the particle size, and mineralogy for chalcopyrite, quartz, and other sulfides in-pulp within 5 min.

Automated summary

This study investigates a technique using deep learning machine vision and optical microscopy for in-pulp characterization of mineralogy and particle size distribution for multiple minerals in a copper ore pulp. The technique can predict the particle size and mineralogy for chalcopyrite, quartz, and other sulfides in-pulp within 5 minutes.