Data-Driven Tube-based Model Predictive Control of an Industrial Thickener

Abstract

In this work, a data-driven tube-based model predictive control (MPC) is presented to track the setpoints of the underflow concentration. By defining the terminal admissible set to consider all the possible steady-states, the controller can ensure tracking for all reachable operating setpoints. Besides, a data-driven general polyhedral uncertainty set is constructed by employing the principal component analysis (PCA) technique, which can effectively capture correlations among uncertain variables. Based on the constructed uncertainty set, the feasible region could be enlarged while reducing the conservatism of control performance. In addition, recursive feasibility and stability of the controller can be guaranteed. The effectiveness of the proposed method is verified by tracking problems of the thickening process.