The improvement of the observability and the development of a method for an
active process-control of the beer fermentation process were main goals of
this Work. Current offline measurement methods in breweries were substitut
ed by online measurement and modelling methods. Especially the online deter
mination of diacetyl, a biochemical key compound during the fermentation Pr
ocess, using an artificial neural network-model ("Cognitive Estimator"), en
tailed a great improvement of the process-observability. Applying the "Cogn
itive Estimator for diacetyl" in practice, postulated accuracies of the onl
ine determination were reached during the whole fermentation process. A fuz
zy-logic expert-system specified the actual state of the fermentation based
on the delivered online information. A potential for a reduction of fermen
tation time up to 25% was exposed in practice by the process control module
which is based on an automated change between the detected fermentation st
ates.