Economic model predictive control for a rougher froth flotation cell using physics-based models

Abstract

The development of an economic model predictive control (E-MPC) strategy is presented. The strategy uses a novel dynamic flotation model that incorporates the physics of the froth phase in a flotation cell. The dynamic model was previously calibrated and validated using experimental data.

Sensitivity analyses were conducted to select a suitable objective function that accounted for both process economics and control variable sensitivities. While the ultimate goal of a rougher flotation cell is to maximise the metallurgical recovery at a steady state for a specified minimum grade, it was evident that the incorporation of air recovery dynamics (which can be measured in real-time) and concentrate grade dynamics (calculated through first-principle models) led to the best results. The addition of a dynamic variable that can be easily measured online, i.e. air recovery, offers great potential to improve plant performance in existing froth flotation systems. Furthermore, a minimum concentrate grade was imposed in the E-MPC strategy. This acts as an economic constraint as it allows the metallurgical recovery to be optimised while ensuring that concentrate grade requirements are met.

The dynamic optimisation problem for the E-MPC strategy was discretised using orthogonal collocations, and was implemented in Matlab using automatic differentiation via CasADi. Two typical manipulated variables were considered: air flowrate and pulp height setpoints. Based on laboratory-scale data, the implementation of the E-MPC strategy resulted in improvements ranging from +8 to +22 % in metallurgical recovery, while maintaining the specified grade. This is therefore an encouraging control strategy to explore in larger flotation systems.