Modelling Fed-Batch Fermentation Processes: An Approach Based on Artificial Neural Networks

Abstract

Artificial Neural Networks (ANNs) have shown to be powerful tools for solving several problems which, due to their complexity, are extremely difficult to unravel with other methods. Their capabilities of massive parallel processing and learning from the environment make these structures ideal for prediction of nonlinear events. In this work, a set of computational tools are proposed, allowing researchers in Biotechnology to use ANNs for the modelling of fed-batch fermentation processes. The main task is to predict the values of kinetics parameters from the values of a set of state variables. The tools were validated with two case studies, showing the main functionalities of the application.

Optimization of Nosiheptide Fed-Batch Fermentation Process Based on Hybrid Model

“…A mechanism model of the nosiheptide fermentation process can be used to explain the basic nature of the process directly, but the process is so complex a biochemical reaction that the mechanism model we can get is not precise enough. , Black-box modeling methods have already been successfully used in many fields. − We can build the accurate mathematical model of a certain process using black-box modeling methods without knowing detailed knowledge of the process, if there were sufficient modeling data, , which are difficult to get in the nosiheptide fermentation process, however. When only limited amount of data is available to build a black-box model, the model often tends to overfit the training data and results in significant errors when applied to unseen data .…”