Optimizing control for a two-stage grinding circuit, O. Haavisto, M. Olofsson, J. Martikainen, S. Kauvosaari, and M. Kosonen

The target in grinding circuit operation is usually to maximise the throughput and energy efficiency while keeping the product particle size on an optimal level for downstream operations. Furthermore, feed ore variations in size and hardness need to be compensated in order to stabilise the process. This is a challenging task for any circuit and requires good instrumentation, combined with careful control design. This study investigates the control of a two-stage iron ore grinding circuit with an AG and pebble mill at LKAB Kiruna KA3 concentrator in northern Sweden. A new control approach based on multivariable model predictive control (MPC) is designed, implemented and evaluated for the circuit. Additionally, advanced instrumentation including a volumetric charge measurement for grinding mills is utilised. The MPC control approach starts with step tests that are used to model the effect of manipulated variables (MVs) like feed rate, water addition and primary mill speed on controlled variables (CVs) such as final product particle size, power consumption and volumetric charge of the mills. In the control implementation phase, the models are then utilissed together with current process measurements and known operating limits to predict optimal future changes for the manipulated variables in order to reach the desired process output values. It is shown that the new control approach improves the stability of the circuit and reduces energy consumption. Keywords: Model predictive control, grinding, optimisation