An Implementation of Model Predictive Control on a SAG Mill

- Organization:
- Canadian Institute of Mining, Metallurgy and Petroleum
- Pages:
- 18
- File Size:
- 1562 KB
- Publication Date:
- Jan 1, 2019
Abstract
"Many modern concentrators with VSD driven SAG mills still rely on the control room operator to manually adjust the speed in reaction to increasing bearing pressure. All too often, the problem with this strategy is by the time the operator reacts with a sudden speed increase, the SAG Mill is likely already on its way to loading up. To assist the new operators, the SAG Mill at B2Gold’s Fekola Mine in Mali needed a robust control that could handle load changes and unmeasured disturbances with minimal operator intervention. With modern Distributed Control Systems (DCS) powerful enough to run advanced control, they often have Model Predictive Control (MPC) blocks built in or available as an “add-on” license, and thus MPC can be readily implemented within the modern concentrator with a relatively low added capital cost. This paper discusses an implementation of two single output MPC controllers to control the SAG Mill at Fekola. This was implemented using the existing signals and advanced control blocks available to the installed DCS.INTRODUCTION Many modern concentrators with VSD driven SAG mills still rely on the control room operator to manually adjust the speed in reaction to increasing bearing pressure. All too often, the problem with this strategy is that by the time the operator reacts with a sudden speed increase, the SAG Mill is likely already on its way to loading up. The next step the operator may take may be to cut back the feed rate, or production until the SAG is once again stable. With this sudden reaction comes an abrupt change in SAG rotational speed and power consumption that sends a surge of material down the circuit as the mill suddenly ejects more pulp. Furthermore, choosing a higher rotational speed can lead to higher than normal liner wear as the cataract overshoots the toe of the charge and impacts the liners. After start-up at B2Gold’s Fekola Gold Mine in 2017, the SAG mill was being controlled manually. The SAG rotational speed was adjusted manually by the operator in step changes depending on whether the SAG bearing pressure was increasing, decreasing or stable. During this time of manual control, the SAG fresh feed rate was also adjusted to a tonnage setpoint with a PID controller. Generally, the SAG fresh feed rate setpoint was kept at a target value, but then the feed rate was throttled back if the SAG mill or grinding circuit showed signs of overloading. The extra challenges were that the operators had limited experience and were concurrently learning how the plant operated. Also, they were also occupied with managing situations in the rest of the plant. Often their attention was drawn away from the SAG mill and during that time, the SAG Mill would load up before they could react."
Citation
APA:
(2019) An Implementation of Model Predictive Control on a SAG MillMLA: An Implementation of Model Predictive Control on a SAG Mill. Canadian Institute of Mining, Metallurgy and Petroleum, 2019.