The application and future development of adaptive control to the froth flotation process

Abstract

Control studies were carried out in the rougher circuit at Bougainville Copper Limited, and in the lead cleaner circuit at Mount Isa Mines Limited .At BCL a simple rougher control system was designed and implemented. A standard approach was applied, using fixed parameter feedback (PI), and feed forward controllers. The control objective was to increase the rougher copper concentrate grade, while maintaining the rougher-scavenger recovery. Since recovery is determined largely by the ore type, the above objective is both logical and achievable.The rougher control system performed successfully over an extended period of time, increasing the average concentrate grade by 1 %Cu, without sacrificing the rougher-scavenger recovery. This control system represents an application where an understanding of the constraints imposed by the ore type, and the operating philosophy of the concentrator, is more significant than the actual control techniques which are employed. For these reasons the control approach hit application to other flotation circuits treating porphry copper ores.The lead cleaner circuit at HIH was characterised by dynamic behaviour which was found to be not stationary with time. This resulted in a complex environment, and hence the need to develop an adaptive control system, at both optimising and stabilising levels.A self-tuning minimum variance controller was applied to stabilise the lead final concentrate grade. It was discovered that a low order dynamic model could be used to predict the response of this complex process. This resulted in the development of a simple and robust self-tuning control system, based on minimum variance control theory. The self-tuner was able to improve the overall stability of the lead cleaner circuit, when compared with manual control techniques.The setpoint of the self-tuning controller was adjusted by an optimising control system, in order to maximise the economic efficiency of the flotation circuit. Process optimisation was based on determining the effect of lead final concentrate grade on economic performance. The method used was based on fitting simple linear equations to stored process data. Linear regression was used to fit the equations. This procedure was carried out by the control computer, before each optimising step.Deficiencies were observed in both rougher and lead cleaner control systems. Techniques incorporating several aspects of adaptive control theory, including adaptive prediction and recursive parameter estimation, are discussed as approaches for rectifying the said deficiencies. The use of a new steady state adaptive optimisation algorithm, and its potential application to rougher flotation in general, is considered. It appears that this technique has wide appeal for flotation in the future.