Control of stagewise extractors using neural-based approximate predictive control as compared to nonlinear MPC

The control of liquid-liquid extraction processes is still one of the major areas of research due to the complexity of the process and the inherent nonlinearity and varying dynamics involved in its operation. Traditional linear control schemes may have limited performance when applied in situations involving unknown process time delays, loop interactions, and processes with unknown order such as the extraction process itself. The objective of this work is to present a comparative study for the application of a nonlinear model predictive control technique for the control of an agitated type extractor as compared to an approximate nonlinear one. The process model used in the two control algorithms is based on a neural network approach. The implementation of these two algorithms covers the performance of the control system for the servo as well as regulatory control of the column. Both controllers were able to drive the process for good set-point tracking and disturbance rejection. The approximate model predictive control (MPC) was slightly faster in response and achieved its control objective in much lower computation time.