Causal model of an industrial platinum flotation circuit

Abstract

Models of flotation found in literature are generally complex and reliant on manual surveys and laboratory data. Furthermore, these models do not usually indicate the causal relationships from true process independent variables to process outputs such as individual cell recovery and metal grade. These models are often found to be suitable for design but less applicable to the simulation of an industrial plant to perform offline control and optimization studies. This study explains these relationships from the individual cell inputs, level and air valves, to cell hydrodynamics and ultimately mass produced, recovery and grade. A technique is proposed to derive individual cell froth recovery by adjusting one froth stability parameter across the entire bank. The model is trained on an industrial platinum concentrator using predominantly online plant data and the model predictions are compared to real plant data over a period of a month.