Quality Control of fermentation processes

The production of biofuels from renewable feedstock has grown immensely in the past several years. Bioethanol is one of the most interesting alternatives for fossil fuels, since it can be produced from raw materials rich in sugars and starch. Ethanol fermentation is one of the oldest and most important fermentation processes used in the biotechnology industry. Although the process is well-known, there is a great potential for its improvement and a proportional reduction in production costs. Due to the seasonal variation of feedstock quality, ethanol producers to need to monitor the fermentation process to ensure the same quality product is achieved.

Near-infrared spectroscopy (NIRS) offers rapid and reliable prediction of ethanol content, sugars, Brix, lactic acid, pH, and total solids at any stage of the fermentation process.

Figure 1. Fermentation mash sample placed on the DS2500 Solid Analyzer.

Production of ethanol from corn goes through three typical steps: milling / liquefaction of corn into starch mash, fermentation of starch mash with yeast, and finally purification of the resulting ethanol by distillation. A total of 206 samples (117 for Brix index) of fermentation mash were analyzed on the DS2500 Solid Analyzer. Due to the large amount of solids present in the samples, all measurements were performed in reflection mode using the DS2500 Large sample cup (Figure 1). The samples were measured in rotation to collect spectral data from several areas. Spectral averaging of signals from several spots helped to reduce sample inhomogeneity. The Metrohm software package Vision Air Complete was used for all data acquisition and prediction model development.

**Table 1.** Hardware and software equipment overview
Equipment	Metrohm number
DS2500 Analyzer	2.922.0010
DS2500 Large Sample Cup	6.7402.050
Vision Air 2.0 Complete	6.6072.208

All 206 measured Vis-NIR spectra (Figure 2) were used to create a prediction model for quantification of the key fermentation parameters. The quality of the prediction model was evaluated using correlation diagrams, which display a very high correlation between the Vis-NIR prediction and the reference values. The respective figures of merit (FOM) display the expected precision of a prediction during routine analysis.

Figure 2. Vis-NIR spectra of fermentation mash samples analyzed on a DS2500 Solid Analyzer.

Result ethanol content

Figure 3. Correlation diagram for the prediction of ethanol content using a DS2500 Solid Analyzer. The ethanol content lab value was evaluated using HPLC.

**Table 2.** Figures of merit for the prediction of ethanol content using a DS2500 Solid Analyzer.
Figures of merit	Value
R²	0.998
Standard error of calibration	0.21%
Standard error of cross-validation	0.22%

Result solid content

Figure 4. Correlation diagram for the prediction solid content using a DS2500 Solid Analyzer. The lab value was evaluated by LOD balance.

**Table 3.** Figures of merit for the prediction solid content using a DS2500 Solid Analyzer.
Figures of merit	Value
R²	0.982
Standard error of calibration	0.87%
Standard error of cross-validation	1.06%

Result brix index

Figure 5. Correlation diagram for the prediction of Brix index values. The lab value was measured using a refractometer.

**Table 4.** Figures of merit for the prediction of Brix index values.
Figures of merit	Value
R²	0.987
Standard error of calibration	0.66
Standard error of cross-validation	0.87

Result total sugar content

Figure 6. Correlation diagram for the prediction of the total sugar content. The total sugar content lab value was measured using HPLC.

**Table 5.** Figures of merit for the prediction of the total sugar content.
Figures of merit	Value
R²	0.981
Standard error of calibration	1.09%
Standard error of cross-validation	1.30%

Result glucose content

Figure 7. Correlation diagram for the prediction of glucose content. The glucose content lab value was measured using HPLC.

**Table 6.** Figures of merit for the prediction of the glucose content.
Figures of merit	Value
R²	0.920
Standard error of calibration	0.70%
Standard error of cross-validation	0.86%

Result lactic acid content

Figure 8. Correlation diagram for the prediction of lactic acid content. The lactic acid lab value was evaluated using HPLC.

**Table 7.** Figures of merit for the prediction of lactic acid content.
Figures of merit	Value
R²	0.722
Standard error of calibration	0.09%
Standard error of cross-validation	0.10%

Result pH value

Figure 9. Correlation diagram for the prediction of pH value. The pH lab value was measured using a pH meter.

**Table 8.** Figures of merit for the prediction of pH value.
Figures of merit	Value
R²	0.734
Standard error of calibration	0.17
Standard error of cross-validation	0.19

Result maltotriose content

Figure 10. Correlation diagram for the prediction of maltotriose content. The maltotriose lab value was measured using HPLC.

**Table 9.** Figures of merit for the prediction of maltotriose content.
Figures of merit	Value
R²	0.928
Standard error of calibration	0.36%
Standard error of cross-validation	0.42%

Result dextrin content

Figure 11. Correlation diagram for the prediction of dextrin content. The dextrin lab value was measured using HPLC.

**Table 10.** Figures of merit for the prediction of dextrin content.
Figures of merit	Value
R²	0.964
Standard error of calibration	0.60%
Standard error of cross-validation	0.68%

This application note demonstrates the feasibility to determine multiple key parameters of the fermentation process with NIR spectroscopy. Corn fermentation is a well-established process which typically runs for 55–60 hours. Samples are extracted from fermenters every few hours and sent to the laboratory for analytical measurement. Several analytical methods need to be used to monitor key quality parameters for the fermentation process. Vis-NIR spectroscopy enables a fast alternative with high accuracy, and therefore represents a suitable single method to monitor the fermentation process.

**Table 11.** Time to result overview for the different parameters
Parameter	Method	Time to result
Ethanol, sugars	HPLC	∼30–45 min
Brix index	Refractometer	∼3–5 min
pH	pH meter	∼3–5 min
Solids	LOD Balance	∼10–15 min

Author

Metrohm Middle East FZC

B2-21 SAIF Zone, Sharjah, UAE
120747 Sharjah