Real time in-line monitoring of large scale Bacillus fermentations with near-infrared spectroscopy.

NIR spectroscopy was used to monitor Bacillus fermentations in 50 m(3) reactors under harsh industrial conditions. The BioPAT(®) Spectro NIR sensor was attached directly to the bioreactor and provided fast, sensitive, non-destructive and robust measurements without interfering with the microorganism metabolism. Multivariate data analysis techniques related the spectra collected in real time during the fermentation with reference analyte concentrations. Analyte concentrations of future batches can be determined in real time with these models. The SugarSUM parameter was modeled with a SEP of 1.33 g/L in a range of 0-35 g/L. The models for AnalyteSUM (SEP = 0.81 g/L in 0.5-43 g/L range), OD(600) (SEP = 2.88 OD in 3.5-50 OD range), dry mass (SEP = 0.09 in 0.4-1.7% range) and Acetoin (SEP = 0.94 g/L in 0-11 g/L range) also show a great prediction performance in the complex media matrix. Sophisticated process control strategies such as a feeding control of the sugar source can be implemented in the future, potentially increasing spore yield due to a reduction of carbon overflow mechanisms. Media classification with PCA identified media formulation errors. Batch evolution models, built with spectra data only, monitored the evolution of new batches by comparing it with a "golden batch" trajectory.