Hybrid model predictive control for flotation plants

Abstract

Development of control strategies for large-scale processes has led to formulate novel methodologies to consider the global performance of a plant, facilitating the design, validation and evaluation of more complex control strategies. This paper describes a hybrid dynamic simulator for a flotation circuit with one production line composed by one cell and two banks in series, calibrated with industrial data. This simulator allows to represent a real plant behavior, facilitating the process of extracting input–output data. A PieceWise AutoRegresive eXogenous (PWARX) system is obtained from simulated plant by applying identification techniques for different operating conditions. A Hybrid Model Predictive Control (HMPC) methodology based on PWARX models has been developed and evaluated showing a good performance. Simulation results show that the proposed methodology is suitable for modelling the behavior of a flotation process and its control stage by minimizing the tracking error in the final tail grade.