Improving the identification of a penicillin fermentation model

This work concentrates on the selection and improvement of differential equation based models of the penicillin G fermentation. Published penicillin fermentation models have been reviewed and compared with regard to their abilities to predict fermentation behaviour, genetic algorithms have been applied to the design of optimal experiments for model parameter estimation, and a new approach to assessing the theoretical identifiability of model structures has been proposed. When applied to the best penicillin fermentation model yet found, this new approach suggests that the model's parameters are uniquely identifiable. The best performing model was shown to be a morphologically structured model for which measurement data related to the various morphologically distinct regions were obtained using image analysis. This model was modified to increase its speed of execution, and extended to describe fermentations where lactose was present in the inoculum. Design criteria from the field of optimal experiment design were combined with genetic algorithms as a technique for searching through the range of possible input combinations, subject to constraints on the fermenter operation, to develop experimental feed profiles. The theoretical identifiability of the fermentation model has been assessed for the first time, using a novel approach to identifiability testing which uses a symbolic mathematics package, along with subsequent post-processing, to determine almost at a glance whether or not a fermentation model should be uniquely identifiable.

Name dc:type.qualificationname

d_ph

Level dc:type.qualificationlevel

d_ph

Grantor dc:publisher.institution

University of Birmingham

Year dc:date.issued

1999

Author dc:creator

• Syddall, Mark Timothy

• TP Chemical technology