Multivariable Control of a Continuous Crystallization Process Using Experimental Input-Output Models

Abstract

Results from closed-loop control of a continuous evaporative crystallization process on the basis of identified input-output models are presented. Experimental results show that open-loop identification of a crystallization process may lead to poor models as the process has a tendency to oscillate. Improved experimental conditions are achieved under feedback control, using a simple single-loop PI controller. Identification of low-order models, on the basis of closed-loop data, is studied using linear multivariable input-output models. Controllability analysis reveals that only two output quantities can be controlled independently: the mass production rate and a quantity related to the density of small crystals (fines) in the system. A two input two output model predictive controller (MPC) is designed and tested on the pilot crystallizer. Experimental results show that an improvement of the performance of the closed-loop system is obtained.