Analysis on the difference of volatile flavor components in strong-flavor Baijiu base liquor with different grades

In order to analyze the difference of volatile flavor components in strong-flavor (Nongxiangxing) Baijiu base liquor with different grades, the volatile flavor components in 4 grades of strong-flavor Baijiu base liquor were analyzed by gas chromatography-mass spectrometry (GC-MS) technology. Princip...

Full description

Saved in:
Bibliographic Details
Main Author: YUE Xiaolin, XIANG Shuangquan, QIAN Yu, LIAN Shengqiang, YANG Guorui, LAN Xiaoqin, TAN Wenyuan
Format: Article
Language:English
Published: Editorial Department of China Brewing 2025-06-01
Series:Zhongguo niangzao
Subjects:
Online Access:https://manu61.magtech.com.cn/zgnz/fileup/0254-5071/PDF/0254-5071-2025-44-6-258.pdf
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In order to analyze the difference of volatile flavor components in strong-flavor (Nongxiangxing) Baijiu base liquor with different grades, the volatile flavor components in 4 grades of strong-flavor Baijiu base liquor were analyzed by gas chromatography-mass spectrometry (GC-MS) technology. Principal component analysis (PCA) and orthogonal partial least squares-discriminant analysis (OPLS-DA) were conducted based on the results, the key flavor components were screened based on odor activity value (OAV), the different flavor components were screened using variable importance in the projection (VIP) values, and the discriminant model was established combined with machine learning. The results showed that a total of 44 common volatile flavor components were detected in 4 grades of strong-flavor Baijiu base liquor by GC-MS, including 17 alcohols, 12 esters, 5 acids, 5 aldehydes and ketones, and 5 others. The 4 grades of strong-flavor Baijiu base liquor could be effectively distinguished by PCA and OPLS-DA. A total of 21 key flavor components (OAV>1) and 12 differential flavor components (VIP>1) were screened out. The 3 discriminative models were constructed through machine learning, among them, the random forest model had the optimal discriminative effect with an accuracy of 100%.
ISSN:0254-5071