Dynamic Optimization of Bioreactors using Probabilistic Tendency Models and Bayesian Active Learning

MARTÍNEZ E., CRISTALDI M., GRAU R., LOPES J.

Chalkidiki

Simposio; 21st European Symposium on Computer Aided Process Engineering; 2011

First-principles models of fermentation processes typically have built-in errors in theform of structural mismatch and parametric uncertainty. A model-based optimizationapproach for run-to-run improvement under uncertainty of fed-batch bioreactors byintegrating probabilistic tendency models with Bayesian inference is proposed.Probabilistic models grounded on first principles are used in the design of dynamicexperiments to bias data gathering towards the subspace of most promising operatingconditions. Results obtained in the fed-batch fermentation of penicillin G are presented.