Software sensor for biomass concentration monitoring during industrial fermentation

For the on-line monitoring of the parameters and states of aerobic fermentation batch processes three parallel software sensors for biomass concentration monitoring were developed, which use mutually independent measured input signals: the oxygen uptake rate, the carbon dioxide excretion rate and the reaction heat flow. The software sensors are based on a linear model, where the parameters of the model are estimated using the recursive least squares method. The model was upgraded with a mechanism for a time-varying forgetting factor to enable good tracking of the estimated parameters during rapid and large changes in the non-linear and in certain phases of the production, very dynamic process. The estimates given by all three software sensors are in good agreement with the measurements acquired off-line. The developed software sensors make it possible to achieve accurate process identification, mutual control of individual sensors and fault detection in the process. In this way, better supervision and more accurate monitoring of the industrial fermentation process is possible, where reliability of the operation is of key importance.