Neural network for modelling and control of fed batch fermentation process

In a typical industrial fermentation process, important variables such as product concentration are determined by slow infrequent off-line laboratory analysis, making this set of limited use for control purposes. In this paper the artificial neural network approach has been adopted for the online estimation of the state variables in the fed batch fermentation process with the neural network taking on the task of both modelling and state estimation. The ability of the neural network to estimate the state variables is compared with the conventional identification approach based on an autoregressive identification followed by the Kalman filter technique. In the second part of the paper, the ability of the neural network to control the state variables around a desired trajectory by controlling the amount of carbon fed is illustrated