Data fusion strategy for rapid prediction of critical quality attributes in JianWeiXiaoShi extract during pulsed vacuum drying process based on FT-NIR and Vis/NIR-HSI

Abstract

This study explored the feasibility of using two optical sensing methods - Fourier Transform Near-Infrared Spectroscopy (FT-NIR) and Visible/Near-Infrared Hyperspectral Imaging (Vis/NIR-HSI) - to quantitatively predict the critical quality attributes (CQAs) of JianWeiXiaoShi extract during pulsed vacuum drying (PVD) process. Additionally, a data fusion strategy was implemented to integrate the two spectral datasets, aiming to enhance the prediction accuracy and robustness of the quantitative models. Comparative analysis revealed that the FT-NIR model demonstrated higher accuracy in predicting moisture content, narirutin, and hesperidin levels, while the Vis/NIR-HSI model performed better in predicting color changes during the drying process of the extract. In addition to moisture content, the prediction model established by integrating the two spectral datasets through the data fusion strategy demonstrated more accurate predictive performance compared to single-spectrum models. Therefore, integrating FT-NIR and Vis/NIR-HSI spectral datasets through the data fusion strategy for online monitoring of quality changes during PVD of extract represents a rapid, non-destructive, and accurate approach to predict CQAs of materials. This study also provides essential technical support and valuable insights for advancing non-destructive analytical technologies in drying processes.