Abstract: Gas chromatography-mass spectrometry (GC-MS) and electronic tongue were used to quantitatively determine the volatile compounds and taste indices of 21 Jiang-flavor baijiu samples of different grades. These samples were differentiated by chemometrics, and key differential compounds among grades were identified. Finally, a discriminant model was established by machine learning. The results showed that there were differences in the contents of volatile compounds in Jiang-flavor baijiu of three grades, indicating the feasibility of further discriminant analysis. The total content of flavor compounds in second-grade baijiu (4 908 mg/L) was significantly lower than that in premium-grade (6 583 mg/L) and first-grade baijiu (8 254 mg/L), while the proportion of several esters responsible for floral and fruity aromas in total esters showed a decreasing trend as the grade decreased. Partial least squares-discriminant analysis (PLS-DA) identified 16 key differential compounds represented by ethyl palmitate and acetic acid. The results of electronic tongue showed that the taste indexes of premium-grade baijiu were more consistent, with lower bitterness and astringency aftertaste. The taste indexes of second-grade baijiu showed significant intersample differences. Principal component analysis (PCA) showed clear discrimination of Jiang-flavor baijiu of different grades according to their taste indexes. The above results provide a basis for the establishment of Jiang-flavor baijiu quality system. Four discriminant models were established based on 25 differential compounds and taste indexes identified. The accuracy of all models was higher than 90%, and the support vector machine (SVM) model performed best, with an accuracy of 100%.

Key words: Jiang-flavor baijiu; gas chromatography-mass spectrometry; electronic tongue; chemometrics; discriminant models