Application of a novel approach for DMC predictive controller design by response surface analysis in a fed-batch bioreactor

Abstract

This work presents a novel approach for tuning a predictive DMC controller implemented in a fed-batch penicillin bioreactor in order to stabilize the dissolved oxygen concentration by agitation speed manipulation. The operating process variables were calculated by a deterministic and non-structured model solved by a fourth order Runge-Kutta-Gill numerical technique with variable steps. The parameters of the model were obtained from experiments and the literature. The estimated parameters of the DMC controller were model, prediction and control horizons, suppression factor and reference trajectory. The tuning approach employed complete factorial design in order to estimate the influence of these parameters on the integral of the absolute error between the controlled variable and the set point. Response surface analysis provided the optimal parameters. This study showed negligible influence of model, prediction and control horizons while the suppression factor and reference trajectory were very important for the controller. Another important feature of the DMC controller was the that the parameters had negligible influence on each other making design of the controller easier. The performance of the DMC controller was evaluated using several delay times and sample periods of the controlled variable. The behavior of this predictive controller was better than a PID controller tuned by the Modified Simplex method.