| dc.description.abstract | Cobalt is an essential metal for the transfer to a low carbon economy, being a component in lithium-ion batteries and wind turbines. More than half of the global production and the world’s resources are from the Democratic Republic of Congo (DRC), where cobalt is extracted as a by-product of copper. Generally, the recovery of cobalt is lower compared to the recovery of copper. Geometallurgy aims to integrate ore deposit characteristics such as metallurgical response into the resource model. In this thesis, a geometallurgical workflow was developed to gain understanding into the mineralogical factors that influence the flotation performance of cobalt bearing minerals. This was done in collaboration with an industrial partner, the world’s largest cobalt producer in 2019, which had the intention to develop a pro-active flotation strategy for processing copper-cobalt sulphide ore. At this time, copper-cobalt sulphide ore is not yet processed by the industrial partner. The geometallurgical approach used in this thesis existed out of a combination of chemical and mineralogical feed material and flotation product characterisation, combined with measured flotation performance. As the first step of the development of the flotation strategy, a screening was done into the best performing collector. A mixed oxide sulphide ore was homogenised and characterised. Copper was hosted for equal amounts in sulphide phases, such as bornite, chalcopyrite and chalcocite, and oxide oxides phases, such as malachite and to a lesser extent chrysocolla. Three cobalt-bearing oxide minerals were observed in the mixed ore: heterogenite, kolwezite and cupro-asbolane, while carrollite was the only cobalt-bearing sulphide. About 75% of the cobalt could be found in oxide phases and 25% in the single cobalt sulphide mineral. A two-stage rougher-scavenger flotation process was used in which sulphides were extracted from the ore first. In the second stage, oxides were activated using controlled potential sulphidisation followed by their recovery. The flotation tests were performed with a range of collectors: xanthate, phosphorodithioate, dithiophosphate, thiocarbamate, and a blend. A dithiophosphate collector had the best performance, recovering 94% of carrollite recovery, over 90% of the copper sulphides and 70% of copper oxide minerals. Recovery of the cobalt oxides was unsuccessful, recovering half of the kolwezite and only 20% of the heterogenite and cupro-asbolane. Using the best performing dithiophosphate collector, the required collector dosage for a rougher-scavenger flotation process was optimized using a copper-cobalt sulphide ore sample. Using the optimized dosage, which was found to be 30 grammes per ton of feed, the changes in mineralogical characteristics through one complete rougher-scavenger cycle were traced. Each collected concentrate and final process tailings were analysed for quantitative mineralogy using QEMSCAN analysis. Characterization showed that the ore was sulphide rich, with the sulphide mineral content of the feed, consisting of carrollite, bornite, chalcopyrite and chalcocite, totalling over 40% by weight. The QEMSCAN analysis showed that carrollite had the best particulate properties for flotation: particle size distribution in the correct range, well liberated particles and not associated with unfavourable minerals. However, the flotation experiment showed that the copper sulphide minerals preferably attached to the collector, compared to carrollite. This resulted in higher final recovery and kinetics for bornite and chalcopyrite, compared to carrollite. In both the rougher and scavenger stage of the flotation process, carrollite particles with favourable characteristics were recovered later in the process compared to bornite and chalcopyrite particles. Using the best performing dithiophosphate collector and the optimum dosage of 30 grammes per ton of feed, the influence of changes in mineralogical properties of the feed samples on flotation performance was studied. This was done using fixed condition flotation tests, with a range of core copper-cobalt sulphide samples from different locations within the same deposit. Measured equilibrium recoveries, rate constants and final concentrates were correlated to feed mineralogical properties as determined by QEMSCAN analysis using Multiple Linear Regression. Using Leave One Out Cross Validation the predictability of the relationships developed using Multiple Linear Regression between flotation performance and feed mineralogy properties were evaluated. This showed that the final copper and cobalt grades were predicted with an R2 of 0.80 and 0.93 and Root Mean Square Error of Cross Validation (RMSECV) of 4.41% and 1.34%. Recovery of cobalt and copper with time could be predicted with an R2 of 0.94 for both metals and an overall test error of 4.70% and 5.14%. Overall, the developed geometallurgical models showed that understanding changes in process feed mineralogy is key to understanding and eventually optimizing copper and cobalt flotation performance. | en_GB |
