Constructing in-situ and real-time monitoring methods during soy sauce production by miniature fiber NIR spectrometers.

The digestion rate of steamed soybean (DRSS), protease activity of koji (PAK) and formaldehyde nitrogen content of moromi (FNCM) are key indicators to monitor soy sauce production. Currently, monitoring these indicators relies on workers' experience, which can sometimes lead to low material utilization rates and even fermentation failures. Near-infrared spectra were collected during soybean steaming, as well as koji and moromi fermentation, using miniature fiber spectrometers. These spectra were optimized using four pretreatment methods, and regression models were constructed using PLS, iPLS, and Si-PLS. The evaluation of models in prediction sets was based on the correlation coefficient (Rp) and root mean square error (RMSEP). Results indicated that Rp = 0.9327, RMSEP = 4.37% for DRSS, Rp = 0.9364, RMSEP = 228 U/g for PAK, and Rp = 0.9237, RMSEP =0.148 g/100 mL for FNCM were obtained. The above results coupling with validation experiments demonstrated that the developed in-situ and real-time spectroscopy system could ensure high-quality soy sauce production.