Standardized workflow to measure mineralogical composition, liberation and 3D geometry of particles using micro x-ray CT

Mineralogical and 3D geometrical properties of particles affect their intrinsic separation behaviour. Artefacts from x-ray computed tomography (CT) images hinder the interpretation to determine the mineralogical and 3D geometrical information of the particles. Here we introduce a new workflow, a combination of a deep neural algorithm known as nnU-Net [1] and MSPaCMAn [2]. The workflow accounts for the partial volume artefacts in CT images. The 3D properties and the mineralogical composition of the particles are derived from the mask of the particles and individual particle histograms. The new workflow will unlock the ability to standardize and automate the mineral phase classification and quantification, determining liberation and calculation of 3D properties of the particles. This will pave the way to optimize the separation processes by finding the link between 3D properties, mineralogy and intrinsic separation properties at the particle level.