Model Predictive Control of the Flotation Process

"In modem concentrators, the operation of flotation machines is often stabilized by simple control loops. Froth level, slurry density and aeration rate are all physical parameters typically controlled through classical PID loops. Few concentrators, however make use of advanced computing technologies for controlling the grade of flotation product streams. Technical challenges such as long process delay and the long and uneven sampling interval of on-stream XRF analyzers often restrain their implementation. Model Predictive Control (MPC) is known to have better performance than PID for controlling processes with long delays (such as flotation circuits). This paper describes the implementation of a multi-input/single-output (MISO) MPC algorithm and an FIR model identification routine using an object-oriented graphical computing environment.The MPC module (controller and identification) was implemented on an industrial control system and applied to many sulfide flotation circuits in the Brunswick Mining concentrator (New Brunswick, Canada). In most cases, the manipulated variable was the collector addition. As long as it is provided with decent process models, the performance of the MPC controller is good on all the flotation circuits on which it was applied. Significant reductions of grade variance are reported and the controller responds well to set-point changes. The user-friendly, graphical nature of the computing environment led to a high degree of acceptance by both the process engineers and the flotation operators.1. INTRODUCTIONOver the past three decades, much instrumentation has been developed for flotation machines. The introduction of instruments such as pulp density gauges and cell level probes was particularly well received by the plants where pneumatic and electronic controllers were available for implementing PID control loops. In a way, these simple loops revolutionized the mineral industry. The operation of flotation circuits could then be stabilized through the use of automatic controllers."