Optimising Control of Sag Mills at Anglogold’s Mponeng Mine

"Anglogold’s Mponeng Mine is situated in South Africa’s North West Province approximately 80km from Johannesburg. The reefs being mined in this area are among the deepest deep level mining operations in the world. These shafts often exceed several thousand meters making these operations complex and expensive.The gold plant is fed with run-of-mine ore which is stored in mill silos. Feedrate to the mills is controlled by a variable speed belt. Three parallel SAG mills (4.85 x 9m) treat 80-100tph, operating in closed circuit. Hydrocyclones classify the mill discharge, returning the oversize to the mill feed. Cyclone feed density and pressure are measured and are used with typical control loops to determine water addition rates to the mill. Control loops at PLC/SCADA level were historically used with limited success in the control of the milling circuits. Mill control for an installation of this type used mill mass (measured by load cells) and feedrate to control throughput. An online particle size analyser (PSI) was installed to assist grind control in tandem with pressure/density control on the cyclones. Although the accuracy of the measured grind was reliable, long term consistency of the PSI could not be guaranteed. Economic pressure from the business sector requires consistent mill throughput at optimum grind conditions. Fluctuations in production targets can manifest due to inconsistent control and the lack of real-time information from the process.This sensitivity to process movements may vary between shifts and also with changes in ore competency.Whilst tools are available to improve control, few have the sensitivity to process movements to provide real-time information for optimal grind control. For this reason an Advanced Control solution was sought. Advanced Control is often described as the supervision of basic PLC/SCADA control. Advanced control systems have to seamlessly compliment the existing control strategy and bridge the gap that exists between basic control and optimised control. A solution was found in the Metso Minerals CISA’s OCS (Optimising Control System). This French designed system makes changes to process setpoints to achieve peak performance and is distinguished from others by its soft sensor capabilities. The soft sensor is an embedded module in the OCS package that can be configured to estimate values normally obtained from field instruments. Advanced control using a Fuzzy Expert System assisted by Statistical tools (such as moving averages and a Soft Sensor using Phenomenological Models) would serve to bridge that gap. Metso Minerals CISA’s OCS is a modular system with various capabilities.It was successfully used to provide on-time process information via Soft Sensors. Circuit stability was achieved using Rule Based Expert Control in complete synergy with Fuzzy Logic. Two rule-based strategies were developed for this circuit. The first uses the online product size analyser as a primary control variable. The second uses a calculated kWh/t value and a modelled mill discharge density as primary control variables. The latter had more availability than the particle analyser and is currently in use."