NEURAL NETWORK MODELING FOR REAL-TIME VARIABLE ESTIMATION AND PREDICTION IN THE CONTROL OF GLUCOAMYLASE FERMENTATION

Efficient real-time bioprocess control requires accurate models for the estimation and prediction of key state variables. Conventional modelling techniques are, however, often unsuitable owing to the uncertainties and complexity of the biochemical reactions involved, noisy or even unavailable data, etc. Artificial neural network models offer here a novel alternative. In the present work neural network programming was applied both in real-time estimation and multi-step ahead prediction of enzyme activity and biomass dry matter in fungal Aspergillus niger glucoamylase fermentation. Three-layered feedforward artificial neural network models of varying topology were constructed in Microsoft Quick C version 2.5, operated in a personal computer with an 80486/33 MHz processor, and applied in real-time estimation and prediction. A backpropagation algorithm with a momentum term was used in the training of the neural network on the basis of varying input/output pair data sets. The neural network models developed performed quite satisfactorily when the results were compared with real values obtained by off-line analyses.