원문정보
초록
영어
The biotechnology industry demands new and high-performance online measurements to identify important process variables. The ultrasonic measurement in combination with mathematical models provides online control of bioprocesses. Characteristically, yeast cultivations show different process phases such as glucose, ethanol and acetate consumption which can be identified via ultrasonic velocity measurements. In addition, a prediction was carried out with measured and simulated data and thus, the process variables were calculated. Because the ultrasonic signals depends on several unspecific variables, this research uses theoretic and chemometric models to correlate the online measurement of the ultrasonic velocity with the offline measurements of the process variables. The prediction of different process variables were performed by so-called software sensors, which are based on multilinear regression, principal component analysis, extended Kalman filter and neural networks. The simulated values and the measurements of ultrasonic velocity were used for the evaluation as well as the prediction of ethanol concentration.