Model based optimization of fed-batch bioprocesses

Project members: J.A. Roubos, P. Krabben, R. Babuška, J.J. Heijnen, H. Verbruggen

Sponsored by: DSM Anti Infectives, DIOC-6

Many biotechnological production systems are based on batch and fed-batch processes. Optimization of the product formation currently requires a very expensive and time consuming experimental program to determine the optima by trial and error. The aim of this project is to find a more efficient development path for fed-batch bioprocesses by an optimal combination of experiments and process models. The two main research topics of this project are:

• Development of a user friendly modeling environment for fed-batch processes. The software tool must be able to use different types of knowledge coming from experts, experiments and first-principles, i.e., conservation laws. New modeling methods such as fuzzy logic, neural networks and hybrid models will be used.
• Iterative optimal experiment design. First some basic experiments can be done to estimate some preliminary parameters for the system. The idea is to make a rough model to design the next experiment. First, a stoichiometric model is made and thereafter a structured biochemical model that will be gradually improved according to the fermentation data. The main objective is to predict the right trends. The actual values are less important at the initial stages.