中国和美国赤霞珠干红葡萄酒产地鉴别模型构建的应用研究  

Application of chemometric determination model building of Cabernet Sauvignon wines from China and USA for geographical origin traceability

作　　者：梁娜娜[1] 刘萤[1] 王琳丽[1] 王珮玥[1] 吕美玲[2] 王金花[1] 张朝晖[1] 韩深[1] 

机构地区：[1]北京出入境检验检疫局检验检疫技术中心,北京100026 [2]安捷伦科技中国有限公司,北京100102

出　　处：《中国酿造》2014年第12期23-28,共6页China Brewing

基　　金：国家质检总局科技计划项目(2014IK272);国家质检总局科技计划项目(2012IK144);北京检验检疫局科技计划项目(2013BK002)

摘　　要：采用液相色谱串联四极杆飞行时间高分辨率质谱进行了中国三个葡萄酒主产区、美国两个酒庄的赤霞珠干红葡萄酒风味物质的检测分析,应用化学计量学工具对数据进行筛选、主成分分析和产地鉴别模型的构建,并通过模型判别准确率和置信度的比较,对构建模型所采用的偏最小二乘法判别分析(PLS-DA)、反向传播人工神经网络(BP-ANN)和朴素贝叶斯算法(NBM)三种算法进行了评价。结果表明,朴素贝叶斯算法(NBM)构建的模型适用于中国和美国赤霞珠干红葡萄酒的产地溯源,反向传播人工神经网络构建的模型适用于中国三个主产区的产地溯源。利用构建的模型进行了样品预测,准确率达86.7%,将为葡萄酒的产地溯源工作提供一定的思路。The flavor compositions of Cabernet Sauvignon wines （V.vinifera L.cv.） from three main regions in China and two wineries in USA were analyzed by ultra-performance liquid chromatography-quadrupole-time of flight mass spectrometer （UPLC-QTOF-MS）.Chemical metrology software was used for data filtering,principal component analysis and origin identification model establishment.The models for region authentication were established by partial least squares discriminant analysis,back-propagation artificial neural network （BP-ANN） and naive bayes model （NBM） algorithms and they were evaluated by comparing the discriminant accuracy and confidence.The results showed that NBM model was suitable for geographical origin traceability of Cabernet Sauvignon dry red wines from three regions in China and two wineries in USA and BP-ANN model was applied to region authentication for Cabernet Sauvignon wines from three main regions in China.The validation of the developed model was tested by samples with an accuracy of 86.7％,which will open a new way for geographical origin traceability of grape wines.

关 键 词：葡萄酒 产地溯源 模型构建 化学计量学 颜色 

分 类 号：O657.6[理学—分析化学]