Online rapid detection of biomass in avilamycin fermentation process based on near-infrared spectroscopy technology

Biomass is one of the key parameters to be monitored during avilamycin fermentation. Currently, periodic sampling is commonly employed for biomass detection, which often entails significant manpower and cost expenditure. Therefore, it is crucial to develop a method capable of real-time, online monitoring of biomass changes during avilamycin fermentation. This study selected wet weight as the biomass measurement indicator during avilamycin fermentation, and employed online near-infrared spectroscopy (NIRS) technology to perform rapid, non-destructive monitoring of wet weight content throughout the fermentation process. Comparing the models' performance based on three algorithms, including partial least squares regression (PLSR), principal component regression (PCR) and support vector regression (SVR), it was confirmed that the PLSR model demonstrated optimal performance. After outlier removal, preprocessing method selection, and feature wavelengths screening, the results indicated that the dynamic biomass prediction model constructed using the SG+1^st Der-i-PLSR method exhibited the optimal predictive performance. Within the validation set, the prediction set determination coefficient (R²_p), root mean square error of prediction (RMSEP), residual predictive deviation (RPD), and standard error of prediction (SEP) were 0.9319, 21.19, 3.351 and 20.98, respectively. For wet weight content ranging from 109.5 to 436.5 g/L, the intuitive metric mean absolute error (MAE) was 16.30 g/L after external validation across five fermentation batches. The foregoing indicates that this model holds promising application prospects for online non-destructive biomass monitoring during avilamycin fermentation. It will lay the foundation for process optimization and control in the Industrial-scale production of avilamycin.