Temperature and water activity control in a lipase catalyzed esterification process using nonlinear model predictive control

Abstract

AbstractControlling the batch esterification process is a difficult task because the kinetics of the process incorporate intrinsic nonlinearity, process uncertainty, and model mismatch. The model predictive control (MPC) was created and used in the lipase‐catalyzed esterification process in this study. The Autoregressive with Exogenous Input (ARX) and Nonlinear Autoregressive with Exogenous Input (NARX) models were embedded in the MPC. The controller's goal is to manipulate the jacket flow rate and air flow rate, respectively, to control reactor temperature and water activity. To identify the best controller performance, the ARX‐MPC and NARX‐MPC parameters of horizon time (P), number of control moves (M), and weighting factor (wk and rk) were tuned in tracking set point. In terms of set point tracking, disturbance rejection, and robustness test, the best‐tuned ARX‐MPC and NARX‐MPC controllers were assessed and compared. Due to smaller integral square error (ISE), quicker settling time, streamlined response, and manipulated variables remaining within their permitted constraints, the NARX‐MPC controller outperformed the ARX‐MPC controller.