Hierarchical hybrid fuzzy strategy for column flotation control

Abstract

Column flotation is widely used in the concentration of low grade ores. Often column flotation concentrate is the final product of a very complex circuit, and therefore control of the metallurgical performance has direct impact in the plant performance. Several control schemes has been implemented for the stabilization of column flotation process, including decentralized control. model predictive control and fuzzy approaches, which attempt to control froth depth, water bias and air holdup. At the same time many efforts have been oriented to improve process instrumentation, with the aim of providing better measurements for control purposes. Instrumentation improvements have made feasible the design of strategies focused on recovery and concentrate grade control. In this work we present the design and implementation of a new advanced controller for column flotation process. The controller was implemented in a 10 columns cleaning stage following a hierarchical scheme with two control levels: an improving level with the aim of metallurgical performance control of the whole process, and a stabilizing level in charge of the distribution of control actions in each column. The controller design was made based on a hybrid scheme with three different operation scenarios, defined by a recovery-concentrate grade domain partition. Results show that the controller is able to keep the process in the normal operation scenario 80% of the analyzed time; on the other hand, when the process was operated only with local control it achieved the normal operation scenario 43% of the analyzed time. Results also show that the controller is capable of increasing concentrate grade and recovery mean values, despite variations on feed grade; while reducing recovery and concentrate grade standard deviations. (C) 2009 Elsevier Ltd. All rights reserved.