Rapid characterize the changes of chemical components closely related to EEFS and LF NMR in Chuju among different drying times

Abstract

The combination of excitation-emission fluorescence spectroscopy (EEFS) and low-field nuclear magnetic resonance (LF NMR), integrated with deep learning algorithms, enhances the intelligent and rapid discrimination capability of Chuju quality under complex thermal processing conditions. First, EEFS and LF NMR were used to record EEFS and LF NMR spectra of representative Chuju samples during different drying times. Then, characteristic variables were transformed, extracted, and combined from EEFS and LF NMR spectra recorded data. Finally, parallel factor (PARAFAC), partial least squares regression (PLS), and support vector regression (SVR) have been comparatively combined in developing the prediction models. In a comparison of the constructed models, the PLS model achieved the best result, with R = 0.9950, RMSECV = 0.0617 for the training set, and R = 0.9997, RMSEP = 0.0037 for the test set, respectively. The overall results revealed that EEFS and LF NMR combined with chemometric analysis could be successfully used as a rapid method in determining the Chuju quality, and perhaps some chemical compositions of Chuju during drying processes.