A machine learning and particle-based approach for optimizing operating conditions in mineral processing units


A machine learning and particle-based approach for optimizing operating conditions in mineral processing units

Pereira, L.; Frenzel, M.; Khodadadzadeh, M.; Tolosana Delgado, R.; Hassanzadehmahaleh, A.; Antoniassi, J. L.; Hoang, D. H.; Heinig, T.; Gilbricht, S.; Ramos, G. S.; Gutzmer, J.

The integration of single-particle data and machine learning functionality into a particle-tracking framework has shown great potential in predicting and understanding the mineral processing behavior of individual particles in mineral beneficiation systems under fixed process conditions. So far, only a few studies have addressed the optimization of mineral processing units considering both machine properties and detailed particle datasets simultaneously. A methodology that provides the recoverability of each particle under varying process conditions is thus still missing. This contribution extends logistic regression-based particle-tracking to process optimization. In addition, the method includes a strategy for better visualizing the extensive particle datasets by grouping particles with similar process performance with minimum bias. The use of the methodology is illustrated on a magnetic separation test in which the recoverability of each particle is assessed as a function of magnetic field strength. The generalizable method presented here is suitable for optimizing and visualizing mineral separation processes at single-particle resolution.

Keywords: Predictive geometallurgy; particle tracking; automated mineralogy; process optimization

  • Contribution to proceedings
    International mineral processing congress, 18.-22.10.2020, Cape Town, South Africa
    XXX International Mineral Processing Congress 2020

Permalink: https://www.hzdr.de/publications/Publ-29869
Publ.-Id: 29869