Data Fusion for Enhanced Fermentation Process Tracking

Near-infrared spectroscopy along with process control variables, such as integral of airflow rate and the integral of alkali addition rate can be used as the basis for the monitoring of key analyte concentrations on a fermentation process. Within this paper, sequential data fusion modeling is applied first, embracing both physical and chemical information. Aiming to overcome the limitations of sequential modeling and to compare model accuracy, a novel data fusion methodology based on Partial Least Squares, weighted multivariate calibration, is introduced. The methodologies are applied to data from an industrial fermentation process and it is shown that the data fusion method results in a 50% improvement in the Root Mean Square Error of Cross Validation (RMSECV) compared to more traditional calibration approaches. An optimisation procedure was then considered in association with spectral window selection (SWS) to attain more accurate data fusion models.