Using surface-enhanced Raman spectroscopy combined with chemometrics for black tea quality assessment during its fermentation process

Developing a reliable and convenient method for monitoring the quality of black tea during fermentation could lead to a significant improvement in fermentation process. This work presented a rapid method based on surface-enhanced Raman spectroscopy (SERS) technology and chemometrics to determine the optimal fermentation stage and monitor the changes in 10 types of quality indicators of black tea throughout fermentation. First, the 10 different fermentation time points were clustered into 5 fermentation stages. Based on the SERS data, the fermentation stages were distinguished with an accuracy of 83.33% by one-dimensional ResNet18 (1D-ResNet18). Furthermore, important Raman peaks at 317.71, 619.59, 731.48, 956.08 and 1326.70 cm -1 were found for monitoring quality changes of black tea by density functional analysis and correlation analysis. The prediction r 2 for catechin (C) and epigallocatechin gallate (EGCG) reached 0.81 and 0.82, respectively, by integrated SERS with a one-dimensional convolutional neural network (1D-CNN). In conclusion, this study revealed the Raman fingerprint characteristics of key compounds associated with the fermentation quality of black tea, presenting an opportunity to quantify the quality changes of tea during fermentation using SERS data. With the monitoring method developed in this research, the optimal fermentation stage can be determined accurately, thus decreasing fermentation costs and improving tea quality. • A novel and highly interpretable SERS method for evaluation of black tea quality • A novel method for determining the optimal fermentation degree of black tea • SERS model simultaneously measure the content of nine quality indicators of black tea • Raman fingerprint peaks associated with the main components of black tea were found • Deep learning successfully mined feature of SERS for quantitative detection