Bioprocess engineering, the technology that is focused on the exploitation of the metabolic potential of biological agents, has attracted growing interest throughout the past 50 years from both scientific and industrial communities. The products that have been brought to market range from pharmaceuticals and enzymes to food products and vitamins. The quality of human life has been improved through these efforts in many ways.

Despite a strong research effort and the fact that microbial transformations often reach yields close to the theoretical maximum. many bioprocesses still operate at relatively low yields. One of the obstacles in effective operation is the extraction of useful information from the bioprocess. Sensors that acquire real-time information about the cells' state and their interaction with the environment in the bioreactor are seldom available. Hence, the implementation of sophisticated process control is prevented.

In this thesis a new approach of non-invasive on-line bioprocess monitoring is evaluated. Chemical multisensor arrays (i.e. electronic noses) are used to extract information from the composition of volatiles emitted from the cell culture. The focus is on two specific areas: (i) monitoring of key variables in the bioreactor environment and (ii) monitoring of cell states and physiological events. The overall concern is, besides the increase of yield and reproducibility, the safety operation of bioprocesses.

To cover a comprehensive area of modern bioprocessing, several organisms are investigated under different modes of operation in laboratory- and production scale processes. In repeated batch cultivations of recombinant Escherichia coli it is shown that an electronic nose can monitor biomass and specific growth rate with high accuracy. Glucose and ethanol concentration are monitored in batch cultivations of Saccharomyces cerevisiae. Bioproduct monitoring is presented in production-scale mammalian cell cultivation. The concentration of a therapeutic protein is monitored on-line in this long-term bioprocess thereby also outlining the stability of the sensor technique.

In production-scale mammalian cell culture it is possible to follow cell transition states and monitor the reproducibility of the process. The physiological state of the cell population is revealed in laboratory-scale cultivations. It is shown that microbial contamination can be identified earlier than with conventional methods. Finally, the metabolic burden imposed on bacterial cells through strong overexpression of recombinant protein is monitored in fed-batch cultivation.