基于电子鼻技术的信阳毛尖茶咖啡碱检测方法  

Detection of Caffeine in Xinyangmaojian Tea by an Electronic Nose

作　　者：张红梅[1] 王俊[2] 余泳昌[1] 高献坤[1] 花恒明[1] 何玉静[1] 

机构地区：[1]河南农业大学机电工程学院,郑州450002 [2]浙江大学生物系统工程与食品科学学院,杭州310029

出　　处：《传感技术学报》2011年第8期1223-1227,共5页Chinese Journal of Sensors and Actuators

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：本文采用电子鼻系统对三个等级的信阳毛尖茶进行了检测。采用Loading分析和相关分析对传感器阵列进行优化,选出四个传感器为最终的新传感器阵列,用于信阳毛尖茶的品质识别。PCA分析结果显示,可以将不同等级的茶叶完全区分开,而且效果比较好。利用PCR、MLR和QPSR方法分别建立信阳毛尖茶基于气敏传感器阵列的咖啡碱预测模型,并用预测集对模型进行验证。3种模型咖啡碱含量预测值与实测值之间的相关系数、预测标准误差SEP和平均误差误差百分比分别为0.660、.32和3.42%;0.800、.19和2.8%以及0.94、0.19和2.3%。QPSR模型效果最好。研究结果表明,电子鼻技术可以用于信阳毛尖茶中咖啡碱含量的检测。Three different grades of Xinyangmaojian tea were detected by an electronic nose system.The optimization of a gas sensor array consisting of four gas sensors was selected by loading analysis and correlation analysis.Principle component analysis of the data provided by the electronic nose completely distinguished all the tea samples.The principle component regression（PCR）,multiple linear regression（MLR）,and quadratic polynomial step regression（QPSR）models of the caffeine content of Xinyangmaojian tea were created based on the optimization of gas sensor array and were validated by a prediction set.The correlation coefficient,standard error prediction,and average percent error between the predicted content and the measured content of caffeine of these three models were 0.66,0.32,and 3.42%;0.80,0.19,and 2.8%;and 0.94,0.19,and 2.3%,respectively.The QPSR model was better than the PCR and MLR models.The results show that the electronic nose system can detect the caffeine content of Xinyangmaojian tea.

关 键 词：电子鼻 咖啡碱 信阳毛尖 主成分回归 多元线性回归 二次多项式回归 

分 类 号：TP212.6[自动化与计算机技术—检测技术与自动化装置]