Mineralogy of fine particles in slurry from multispectral imaging

Quantitative mineralogy of fine particles in slurry from multispectral imaging
Sophie Leroy, Godefroid Dislaire, Pierre Barnabé and Eric Pirard

Mineral processing especially in its final stages relies heavily on the differential behavior of particles in pulps. In order to monitor hydrocyclones and flotation cells in almost real time, it is important to develop at line particle characterization providing information on particle distribution but also mineralogy. A sampling device has been elaborated based on a flow cell with variable wall spacing linked to a high pressure peristaltic pump. This setup allows for dispersion and dilution of the mineral slurry into the cell. Depending on the ore or gangue minerals to be controlled, particles can be imaged either in diffuse reflectance or in transmittance mode.
A simple multispectral imaging module has been designed to acquire images at eight different wavelengths. The design is based on a series of dichroic filters thereby avoiding any moving part and enabling a very fast acquisition of multispectral images. Imaging artifacts due, among others, to specular reflectance from the glass window are minimized.
Multispectral classification is used to outline particles appearing in the field of view and to qualify their main mineral component. In particular, areas reflecting a meaningful and discriminative spectrum are identified and compared to a multispectral database. The database is continuously enriched through testing of pure mineral particles under similar pulp conditions (dilution, grain size,…).
Results of final classification are compared to the modal analysis obtained from polished blocks and conventional reflected light microscopy. The technology developed in this work sets the basis for at line monitoring of ore slurries with reasonably simple mineralogy. Extension of the spectral range is being considered for future developments.