Data-Driven Predictive Control Applied to a Thickening Process

In this work, we investigate the use of data-driven predictive control for a thickening process, in which the availability of an accurate mathematical model is often not available. The proposed framework combines elements of predictive control with a Nonlinear Autoregressive Model with eXogenous inputs (NARX) model built from process data using the Lazily Adaptive Constant Kinky Inference (LACKI) learning technique. Simulation results indicate that the proposed approach achieves satisfactory performance in controlling the process, with improvements of up to 21% observed when process data refinement is applied.