Development Of An Adaptive Optimal Control Strategy For An Industrial Ball Mill Grinding Circuit ? Introduction

Ball mill grinding circuits can be subject to upsets that arise from variable characteristics of the feed ore. These upsets are transients in critical circuit variables such as ball mill mass holdup, circulating load, sump level, cyclone efficiency and product particle size. They may occur frequently so that the operation is continuously fluctuating or they may be of such magnitude that some of the equipment is overloaded, resulting in shutdown, or an undesirable product particle size. The feed ore characteristics that influence ball mill circuit behavior in this manner include the grindability of the ore, the eventual transport characteristics of the ore in the mills and the feed ore size distribution. Control strategies for grinding circuits using conventional proportional-integral (PI) control have limitations to their effectiveness in controlling against feed ore disturbances. The first of these are the interactive nature of two or more control loops on a single circuit. Under some circumstances they begin to fight each other, increasing control error and settling time. PI control loop gain constants can be tuned to give best response for some set of conditions but once these change, as they would with a change in the feed ore, they are no longer the "best" ones (Shinsky, 1967). Due to these limitations, alternatives to conventional PI control strategies have been investigated for grinding as well as other mineral processing operations. These alternatives, all computer-based, include self tuning regulators, decoupling control and optimal control (Herbst and Bascur, 1983).