ASABE Technical Library - Abstract

Click on “Download PDF” for the PDF version or on the title for the HTML version.
If you are not an ASABE member or if your employer has not arranged for access to the full-text, view access and subscription options

Real-Time Monitoring of Microbial Protein Production Using Near-Infrared Spectroscopy and Machine Learning

Published by the American Society of Agricultural and Biological Engineers, St. Joseph, Michigan www.asabe.org

Citation:  2026 ASABE Annual International Meeting  2600104.(doi:10.13031/aim.202600104)
Authors:   Yu-Hang Miao, Jin-Hua Qiao, Mohammed Kamruzzaman, Douglas F. Barbin, Wen-Hao Su
Keywords:   Kluyveromyces marxianus, High-density fermentation, Near-infrared spectroscopy, Modeling, Feature wavelengths, Online monitoring system.

Abstract. Food-safe Kluyveromyces marxianusshows great potential for alternative protein production via high-density fermentation. To address the time-consuming, lagging, and reagent-dependent issues of traditional methods for detecting protein content and biomass, an in-situ immersion online monitoring system was developed using near-infrared (NIR) spectroscopy. This system enables rapid, non-destructive, and real-time detection. A hardware platform consisting of a fiber-optic NIR spectrometer (900-1700 nm), a halogen lamp, a custom immersion probe, a 5-L fermenter, and a computer was constructed. From multiple fermentation batches, 210 sets of spectral data and corresponding physicochemical values were collected. After comparing eight spectral preprocessing methods and three regression models (PLSR, SVR, RFR), the optimal methods for protein content and biomass were determined to be NOR and SG preprocessing with RFR modeling, achieving prediction set Rp of 0.919 and 0.974, and RMSEP of 0.021 and 0.591 g/100mL, respectively. Feature wavelength selection further improved the models: the best model for protein content was PCA-DT-RFR (Rp=0.940, RMSEP=0.017, 11 wavelengths), and for biomass was UVE-RFR (Rp=0.984, RMSEP=0.485 g/100mL, 87 wavelengths). Online monitoring software based on PyQt5 was developed and integrated. External validation showed Rp of 0.929 and 0.979, and RMSEP of 0.018 and 0.481 g/100mL for protein and biomass, respectively, meeting industrial inspection requirements. This study provides a reliable technical solution for intelligent monitoring and control of microbial protein fermentation.

(Download PDF)    (Export to EndNotes)