Hybrid Modeling and Optimisation of Industrial Fed-Batch Fermentation Process

摘要

This paper describes the formulation of a hybrid(mechanistic/meural network) model of a industrial fermentation process. A number of hybrid techniques were considered and the best approach was found to be more precise than using the individual models alone. The hybrid model was verified using process data and found to acurately predict biomas and product concentration. Subsequently the model is used for optimisation purposes. The Chemotaxis algorithm was used to determine the parameters of polynomials describing the feed profiles necessary to optimise fermentation productivity.