Modeling and optimization of the fed-batch penicillin fermentation

Venkatesh Kalyanaraman Chittur, Purdue University

Abstract

A systematic and practical approach is developed for the optimization of the fed-batch production of penicillin in cultures of Penicillium chrysogenum. The proposed optimization strategy involves development of a simple but reliable mathematical model, computation of the optimal feeding policy using singular control theory, and experimental verification of the success of the optimal feeding policy. A simple model for the fed-batch fermentation of Penicillium chrysogenum is developed. The model is a simplified version of a structured model developed by Cagney (1984), but is able to predict experimental results as well as the latter. In the model developed, the cell mass is classified into two different states: viable (and therefore active) and non-viable (inactive). All metabolic functions are attributed to only the viable cells; the non-viable cells do not enter into the optimization calculations. As a consequence, in becomes possible to obtain an optimal feed rate profile for the fed-batch culture using singular control theory. A cycle-to-cycle optimization technique is used to obtain the optimal feed rate profile. In order to test the convergence of the technique, it is first applied with the differentiation state model of Cagney (1984) as the reference system. Reasonable estimates of the model parameters are obtained by this simulated optimization process. The experimental implementation of this procedure is then conducted using this as a starting point. The technique is also shown to work if one starts at any arbitrary point far away from the expected optimum. To demonstrate this, an arbitrary fed-batch run is used as a starting point, and, using the optimal feed rate profile generated by parameter values applicable to this run, the cycle-to-cycle strategy optimization is conducted. The results indicate a very desirable tendency of the technique to improve the performance index. Simulation studies are also conducted to investigate what happens when one uses a different model to conduct the optimization. The results of this are inconclusive. Also investigated is the situation when one starts with a functionally incorrect model. This is demonstrated by starting with a model that does not include penicillin inhibition by higher concentrations of glucose.

Degree

Ph.D.

Advisors

Lim, Purdue University.

Subject Area

Chemical engineering|Pharmaceuticals|Pharmacology

COinS