Characterization of Ore and Bulk Solid Systems by Use of Multivariate Image Analysis and Deep Learning Neural Networks

Abstract

The development of soft sensor technologies facilitates the characterization and modelling of complex systems in the mining and mineral processing industry. This thesis is aimed to investigate the state-of-the-art convolutional neural networks in the mineral processing and geometallurgy applications such as froth flotation system characterization, drill core recognition, and particle size segmentation. These results outperformed traditional multivariate image analysis methods by a significant margin.