Self-Tuning Regulators For Control Of Comminution Circuits

Introduction Self-tuning regulators are a class of controllers built around the concepts of sampled data theory, and are especially suitable for implementation in a digital microprocessor or minicomputer control system. The algorithms used are distinctly different from those based upon analog control techniques and include automatic adjustment features which permit them to adapt to unmeasured changes (in, for example, ore grindability) or process nonlinearities (as exhibited by cyclone classifiers). They are also designed to handle transport lags in a simple manner, for example, anticipating the effects of a variable recycle rate on particle size distribution. The advent of distributed control systems will encourage the exploration of these techniques in the minerals processing industries, and particularly for improving the control of comminution circuits, which are especially energy intensive. This paper outlines some current work in progress on this topic. Background Those engaged in the practice of process control have long had two philosophical problems for which solutions have been actively sought - what is optimal control and how can we attain it in some heuristic fashion automatically. The basic analog controller can be represented by the differential equation: [V = Ko + K1 f y • dt + K2 dt(1)] where: v and y are the instantaneous values of output and error respectively and K0, K1, and K2 are tuning constants. The empirical rules for establishing the values of K0, K1, and K2, which provide stability, have evolved from the initial suggestions of Ziegler and Nichols(9) in 1942 and have served their purpose well. However, later analyses demonstrated that there are in fact many combinations of the values of K0, K1, and K2 which, besides being stable, behave better or worse in relation to other process control criteria; some examples are duration of error, amplitude of a damped response, magnitude or rate of movement of control valve or adjustment of speed regulator, etc. The definition of optimality had broadened. In 1964, Kalman(4) published his classic paper which tried to resolve these problems. He defined a