Centralised economic model predictive control of froth flotation banks with experimental implementation

This work presents the development and experimental implementation of a centralised economic model predictive control (E-MPC) strategy for a three-cell flotation bank. The controller simultaneously coordinates pulp level and air flow across all cells to maximise total metallurgical recovery while enforcing a minimum concentrate grade constraint of 20%. A previously calibrated, dynamic physics-based model was utilised in the E-MPC strategy for both simulation and laboratory-scale experiments. In simulation, the E-MPC consistently outperformed conventional proportional–integral control under feed disturbances (flowrate, particle size, and feed grade), achieving recovery improvements of up to +5%. Laboratory implementation demonstrated that the E-MPC effectively adapts to changing conditions and improves recovery; however, challenges with sensor delays and the complexities of real-time optimisation were encountered. The results highlight the potential of model-based optimisation for flotation control and provide valuable insights into the practical challenges and future directions for large-scale flotation plants. • Centralised E-MPC delivered up to +5% recovery in flotation banks. • Grade specification of ≥ 20% is consistently achieved during optimisation. • Successful laboratory implementation highlights both benefits and practical challenges. • Future directions identified for scaling towards real-time plant operation.