A Novel Approach to Mechanism Recognition in Escherichia Coli Fed-Batch Fermentations

Abstract In this work, a novel systematic approach to achieve an efficient mechanistic modeling and simulation of fed-batch fermentations is presented. In order to show the efficiency of the developed simulation framework, data of Escherichia coli fed-batch fermentations are used. Fermentation processes are characterized by its dynamic behavior described by parameters such as growth rate, substrate concentration and cellular metabolic activity. Although there are models able to describe individual fed-batch fermentations, they become unreliable when fitted to new fermentations. To overcome this drawback, in this work different models are used at different optimal time points enabling not only a better description of the process, but also a better understanding of non measurable characteristics. By these means, three models compete in different intervals of the process. The candidate models are: an Overflow metabolism model (OF), a Citric Acid Cycle model (CAC) and a Survival or Maintenance model (M). Using an adequate model sequence, acetate formation, substrate consumption and cell growth are predicted with high accuracy. Moreover, the data needed to fit the models are reduced and a standardization of the model to be applied in different process states is enabled. Besides, with the development of a robust and effective model, the possibility of an online implementation for monitoring and control of the fermentation is exhibited. The results show that an efficient process monitoring based on the Dissolved Oxygen Tension and the Mechanistic Recognition is only limited by the convergence velocity of the algorithm.