MODELING AND COMPUTER CONTROL OF A FLOTATION CELL

During the last three decades flotation has developed into the most important separation method in the mineral industry. The diminishing supply of raw materials and the increasing rate of consumption of metal products is demanding higher efficiency of metal extraction. Efficient automatic control provides a promising answer for better operation of existing and future plants.; In spite of the availability of good instrumentation in modern flotation plants, current control strategies are far less effective than they should be. This failure to develop truly effective control strategies can be traced directly to our lack of understanding of the complex interactions between variables which occur in flotation. For systems of this complexity, our "engineering intuition" is not good enough for control development.; After a complete literature review a realistic phenomenological model of the flotation cell dynamics was developed. Particle/bubble and water transport description together with the hydrodynamic characteristics of a flotation cell have been linked for the first time providing a general model which includes potential controlled and manipulated variables for control system development. The model was parameterized for coal and copper sulfide ore systems using pilot plant data obtained during this research. A highly instrumented pilot plant was built for this purpose. The data obtained was adjusted using a hierarchical method of optimization prior to using it in a nonlinear estimation routine for parameter identification. The dynamic model was coded as a general computer program DYNAFLOAT and verified using dynamic data.; The implication of this model for control strategy development was explored using both coal and copper sulfide ore flotation systems. Identification of serious interactions between loops was observed and found that aeration rate manipulation should be joint operated with impeller speed to provide with optimal physical conditions, the manipulation of aeration by itself was found to be limited and in some cases detrimental.; Finally, a simplified version of the model was developed reducing computer mass storage and execution time for on-line applications. The development of a Kalman filter using this simplified model for on-line predictions of the concentrate grade and flowrate between samples given by the on-line chemical analyzer is presented. The availability of this technique is vital for further testing of control strategies in the current system.