Development of switched predictive control system for continuous purification process of bioproducts

The continuous adsorption recycle extraction (CARE) process is an innovator system of enzyme purification with large potential of industrial applicati on, firstly proposed by [Pungor, E., Afeyan, N. B., Gordon, N. F., Cooney, C. L. (1987). Continuous affinity-recycle extraction - a novel protein sepa ration technology. Biotechnology, 5 (6), 604.] The CARE is originally const ituted by two stages, adsorption and desorption, each one being approximate d by a perfectly mixed reactor. The adsorption stage is by affinity and the desorption stage takes place due to changes in some properties of the syst em, as pH for instance. The process works in a continuous operation mode wi th an adsorbent recycle. A washing intermediate stage was included in the o riginal CARE process and this is the process studied in this work. In order to achieve the goal of this work, a representative mathematical model was developed using the kinetic laws and the mass balance for each component in the system stages. The adsorption and desorption kinetic parameters for th e lisosyme enzyme were obtained in the literature. There are some difficult ies to control the system, such as strong interaction among the input and o utput variables, non linearity and inverse response. The concentration of e nzyme in the liquid phase of two stages of the process is controlled by a d ynamic matrix control (DMC) multivariable predictive controller. To carry o ut this task, the controller DMC manipulates the flowrates of the fermented feed of the desorption stage and of recycle stream of the process. The exi stence of the restriction of the fermented flowrate, in the adsorption stag e, is solved through the strategy of switching of the fermented and recycle flowrate. The control actions are taken in values that guarantee the linea rity of the system around the steady state operation point. The simulation results in closed loop of this predictive control system showed an excellen t performance. (C) 2000 Elsevier Science Ltd. All rights reserved.