Value-iteration-based Adaptive Optimal Reagents Control for Antimony Flotation Process

This paper investigates a data-driven adaptive optimal control approach for antimony flotation process in presence of input time-delay and disturbance. The integration frame of adaptive dynamic programming (ADP) and value iteration (VI) is applied to the optimal controller design without requirement of system dynamics. Fundamentally different from the existing reagents control methods, the input time-delay and disturbance are simultaneously considered in the VI-based ADP control scheme. Specifically, the disturbance is compensated directly by adding an inner model as the feedforward component to the control action and the optimal feedback gain is computed by iteratively solving Riccati equation. By exploiting industrial collected data, the numerical simulation proves that the proposed data-driven methodology can enable the concentrate and tailing grade to keep tracking the target trajectories with a minimum reagents consumption.