Supervisory control of flotation columns using multi-objective optimization

This work describes a method for supervisory control of flotation columns based on predictive trajectory generation of setpoints to be applied to the local controllers of the process. The system maximizes the profit while keeping quality constraints and engineering indices. It represents an advance over predictive supervisors using a single evaluation function or those that reduce a multi-objective problem to a single-objective one by means a weighted sum. The procedure is founded on multi-objective optimization using non-dominated genetic algorithms, and a new classification criterion of the Pareto optimal set. The system has been tested by simulation against usual perturbations affecting the flotation process and exhibits good performance, provided that the model properly represents the system operation.