茶叶中茶多酚含量电子鼻技术检测模型研究  

Testing models for tea polyphenol content based on electronic nose techonlogy

作　　者：张红梅[1] 田辉[1] 何玉静[1] 常粉玲[1] 余泳昌[1] 

机构地区：[1]河南农业大学机电工程学院,河南郑州450002

出　　处：《河南农业大学学报》2012年第3期302-306,共5页Journal of Henan Agricultural University

基　　金：中国博士后基金项目(2009046054);中国博士后基金特别资助项目(201003396);河南省教育厅自然科学基金项目(2009B210017)

摘　　要：为探索茶叶茶多酚含量的快速检测方法,利用电子鼻技术对3个品质等级信阳毛尖茶的挥发性气味进行了研究.采用多元线性回归、二次多项式逐步回归分析和BP神经网络分别建立传感器信号和信阳毛尖茶的茶多酚含量之间的预测模型,并用测试集样本对模型进行验证.试验结果表明,3种模型茶多酚含量预测值与实测值之间的相关系数分别为0.86,0.90和0.92;预测标准误差分别为0.61,0.5和0.14;平均误差分别为2.5%,1.5%和1.0%.3种建模方法对茶多酚含量的预测结果都很好,最优模型为BP神经网络.研究结果表明电子鼻技术结合有效地模式识别方法可以用于茶叶理化成分的快速检测.In order to explore the rapid detection method of tea polyphenols content the electronic nose technology was used to evaluate three different grades of Xinyangmaojian tea. The relationship between sensor signals and content of tea polyphenols for Xinyangmaojian tea was developed using the multipine linear regression（MLR） , quadratic polynomial step regressin（QPSR） and BP network respectively. And they were validated by prediction set. The correlation coefficient between predicted content of tea polyphenols measured by one of these models was 0.86, 0.90 and 0.92 respectively; standard error prediction 0.61,0.5 and 0.14 respectively; the average error was 2.5% , 1.5% and 1.0%. And BP network was the best method compared with MLR and QPSR. The results showed that electronic nose technology combined with effective pattern recognition method can be used for the rapid detection of chemical components of tea.

关 键 词：电子鼻 信阳毛尖茶 茶多酚 多元线性回归 二次多项式回归 BP神经网络 

分 类 号：TP212.6[自动化与计算机技术—检测技术与自动化装置]