Application of Robust Model Predictive Control Using Principal Component Analysis to an Industrial Thickener

A data-driven robust model predictive control (DRMPC) method is used in this brief to control an industrial thickener. To estimate the future states of the thickener, a discrete-time linear time-invariant (LTI) model is employed using the data information from the pressure sensors installed inside the thickener. The hard constraints on the amount of dry ores in the thickener and the underflow concentration are considered in the thickening process, and a robust model predictive control (RMPC) framework based on the affine disturbance feedback (ADF) technique is developed. To better describe the prediction errors of the LTI model of the thickener, principal component analysis (PCA) is employed to properly construct the uncertainty set from the historical data. Case studies based on a simulation platform and an industrial thickener are presented to evaluate the efficiency of the method.