Design of Dynamic Experiments for the Optimization of Batch Fermentation Processes: The Case of Penicillin.
Abstract: This work
aims to investigate the use of a systematic methodology to optimize the operating
conditions of batch fermentation processes, presented by Georgakis (Georgakis, 2009).
This methodology is a novel model-free technique, as opposed to model-based optimization
techniques. The methodology consists of designing certain experiments, obtaining a
response surface model, and optimizing ... read morethe response surface model. This methodology has
been called Design of Dynamic Experiments (Georgakis, 2009) and is an extension of the
well-studied and widely used classical Design of Experiments technique (Montgomery,
2005)(Box & Draper, 2007). The main difference is that the DoDE methodology allows
for the design of experiments in which at least one of the decision variables is a
time-varying one. This allows us to explore several substrate feeding strategies, and to
determine the optimal one. Two different designs of interest to fed-batch fermentations
are studied. One in which the substrate is fed in a systematic fashion throughout the
fermentation (centralized), and one in which the fermentation is split into two
segments, corresponding to the growth phase and the production phase (decentralized).
The results of the two designs are compared. The production of penicillin is used as a
case study for this methodology, using a well-established and widely studied model by
Bajpai and Reuss (Bajpai and Reuss, 1980). Centralized designs are found to be more
efficient than decentralized designs. Using four dynamic subfactors gives the optimal
penicillin production when using Centralized designs. Using three dynamic subfactors
gives the optimal penicillin production when using Decentralized design. However, the
number of experiments required for each optimal design is the same. Centralized Design
has the advantage of only needing to add one extra factor to test the significance of
adding one more dynamic subfactor, whereas the Decentralized Design needs two extra
factors to test the significance of adding one more dynamic subfactor to each
Thesis (M.S.)--Tufts University, 2011.
Submitted to the Dept. of Chemical and Biological Engineering.
Advisor: Christos Georgakis.
Committee: Blaine Pfeifer, and Christoph Borgers.
Keyword: Chemical Engineering.read less
- Component ID:
- To Cite:
- TARC Citation Guide EndNote