In almost every biotechnological process, it plays an important role whether the cells used are alive or dead (viability) and how efficient they are (vitality).  Fluctuations in these parameters can have a negative impact on both the process and the resulting product. Real-time determination of these key parameters is therefore necessary to improve biotechnological processes. This can shorten process times, improve the productivity and yield of a process as well as the quality characteristics of the products and save resources.

Despite the efforts of the process industry to measure critical process parameters inline and thus in real time, this has not yet been sufficiently successful for bioprocesses. The problem is that these parameters are often only determined using offline methods and the resulting time lag and low data frequency make it difficult to react effectively at an early stage. The project therefore aims to research soft sensing (the combination of sensor data and models) for an indirect estimation of these physiological variables.

The CellSENSING project focuses on the following bioprocesses: i) yeast management for beer production with the key parameters yeast viability and vitality, ii) beer maturation with metabolites such as diacetyl, iii) the cultivation of mammalian cells as a step in biopharmaceutical production with live cell density and viability.

Firstly, hardware sensors are implemented in the processes, which provide valid information on the concentration, viability and vitality of yeast cells, on important metabolites of the beer brewing process such as diacetyl and on the viable cell density and viability of CHO cells. Sensor data fusion, data processing, mathematical modelling and machine learning methods will be used to convert this initially partly encrypted information into precise quantitative information about the key parameters. A comparison of different methods should enable the development of robust soft-sensing systems for the precise and accurate real-time determination of these key parameters.

The research question therefore covers the investigation and evaluation of which methods of data processing and mathematical modelling can be used to fuse different hardware sensors in order to enable robust soft sensing.

Project leader:
Prof. Dr. Miriam Pein-Hackelbusch

Partners:
Hamilton Bonaduz AG, Schneider Electric GmbH, WiFö der Deutschen Brauwirtschaft e.V., ZIEMANN HOLVRIEKA GmbH