Bayesian statistics study of a sustainable dissolution of cobalt-bearing minerals from Cu-Co ores

The present paper discusses the dissolution of cobalt-bearing minerals from a copper-cobalt ore using probabilistic models where a priori and a posteriori knowledge of leaching are used to predict the dissolution of cobalt-bearing minerals in a sulphuric acid medium in the presence of a reducing agent. Priorly, the dissolution of cobalt-bearing minerals depends on their mineralogy, leading to the use of FeSO₄ as a reducing agent for the trivalent (Co³⁺) form of cobalt (CoOOH). A posteriori, the dissolution of Co³⁺ is improved by the presence of ferrous ions, resulting from the dissolution of Fe-bearing minerals, including Fe from Co(Fe)OOH. The results showed that the predictive-oriented probabilistic graphic models based on the Bayesian approach, in combination with the design of the experiment data, made it possible to model the leaching of cobalt-bearing minerals. The results from the design of the experiment using the experimental tree methodology associated with the optimization of the multiple responses in a multiple input for a multiple output set-up derived the following optimized parameters: 60°C for the temperature (T), 850 rpm for the agitation, 40% for the solid percentage, 1.5 for the pH, and 4 g/L for the concentration of the Fe²⁺ ion. The cobalt dissolution yield obtained was 89.95%. The analysis of the dependence between the random variables only (P(Fe²⁺|T), P(pH|T), and P(Fe²⁺|pH)) and the dependence between the random variables and the responses (P(Co-yield|pH, Eh)) allowed the construction of two Bayesian networks, respectively, with and without posterior knowledge. For the Bayesian network with posteriori knowledge, the {5–2} structure was found to be the most appropriate arrangement. The model predicted a cobalt yield value, and the experimental value indicated a correlation coefficient (R²) of 0.861.