Power and limitations of model based bioprocess optimization

Since the last few decades the application of system theory concepts and modern model based control techniques to the optimization of biotechnological processes has received a lot of attention. This represents a challenging field of research for the following reasons. First, the dynamics of bioprocesses are determined by the non-linear behavior of living micro-organisms. Therefore, knowledge of reaction kinetics is most often only partial, which represents a serious limitation for mathematical modeling. In addition, bioprocess characteristics are time-varying, which requires on-line adaptation of the model parameters or even the model structure. Second, until now the number of on-line measurement systems is very small, while the measured quantities are most often only indirectly related to the important biological state variables. This inhibits the on-line validation of complex model structures. Therefore, in the past bioprocess optimization used no model at all. However, an adequate mathematical bioprocess description is the basic ingredient for the development of model based control strategies. As a case study, we consider in this paper the optimization of fed-batch fermentation processes involving one limiting substrate for biomass growth and product synthesis, with respect to the volumetric feed rate of this substrate. Three examples illustrate that feed rate profiles which are optimal in the sense of process performance, can contribute at the same time to solving model structure selection and parameter estimation problems as well.