4. Conclusion
The signals from the gas sensor array described here showed a high correlation toward ethanol concentration during cultivations of S. cerevisiae growing on glucose . Similar to any other indirect measurement method, a chemometric model is required for predicting the ethanol concentration from the signals of the gas sensor array. This approach normally requires off-line sampling for calibration purposes which is expensive and time consuming. Alternatively, a simulation model can be used, if the correct parameters for the model are known. In the proposed method, the only requirement for calculating the parameters of the simulation model is the response of the gas sensor array from a single cultivation run. Then the parameters for the simulation model can be calculated by minimizing the prediction error by optimizing the kinetic parameter values of the simulation model as well as the parameter values of the chemometric model.
The proposed model-based calibration method provided comparable results to the reference ethanol concentration values obtained by HPLC. Furthermore, compared to spectroscopy supported models for ethanol prediction which have applied separate prediction models for each diauxic growth phases of the cultivation, in this investigation only a single model was applied for the two diauxic growth phases.
The mechanical stress on the sensor as well as the number of interfering substances is smaller in the gas phase compared to the liquid phase and, furthermore, the so-called sensor fouling by cell adhesion is not a problem here, because a sterile barrier in the form of a mass filter can be easily introduced into the gas stream, this gas sensor arrays have a big potential. Compared to HPLC measurements the time delay is much shorter. Besides these advantages of using this method for ethanol detection, the proposed method is also inexpensive to implement and has just a few maintenance requirements. Therefore the gas sensor array seems to be a useful non-invasive tool for continues monitoring and might also be used for closed loop controlling of processes involving ethanol measurements.