AN EVALUATION OF A NEURAL-NETWORK FOR THE START-UP PHASE OF A CONTINUOUS RECOMBINANT FERMENTATION SUBJECT TO DISTURBANCES

A radial basis neural network was applied to a process for glyceraldeh yde-3-phosphate dehydrogenase produced by an Escherichia coli strain c ontaining the plasmid pBR Eco gap. A neural network trained with a pur e culture predicted the performance of a fermentation containing wild type cells and/or product in the inoculum better than in the reverse c ase; this is explained. In general, the network learnt the trends in t he concentrations of plasmid-containing cells and the recombinant prod uct more accurately than those of wild type cells and the substrate. T his similarity with deterministic networks and the good predictability with some training vectors suggests that neural networks can be used to simulate the start-up phase of recombinant fermentations corrupted by disturbances.