Discrimination Of LongJing-green Tea Quality By Electronic Nose

The tea output of our country is in the second place in the world. Tea not only brings the people enjoying, but also is an important exportation commodity bringing in foreign exchange income. Especially in Zhejiang province the green tea-Xihu Longjing tea, is known far and wide! However, presently tea quality is mainly evaluated by a human taste panel in domestic and foreign. The organic evaluating may vary with different factors, such as: experience, emotion, exhaustion and infection, then the reproducibility is bad. Other methods for tea quality evaluating are also in the developing stage and have some insufficiencies. Which request to establish more scientific tea quality evaluating rules.The tea odor is one of the important factors in tea quality evaluating, moreover the tea 'odor' and 'flavor' have quite close relationship, generally, it is inevitably that tea could have the good flavor when it has the good odor. The odor of the tea consists of many kinds of fragrance ingredient together. The electronic nose (e-nose) technology simulats mammal's sense of smell mechanism, and scanning the information of odor from the overall, which lets the people realize the probability to appraisal smell overall information. Present the reports of application of the e-nose on the tea quality evaluating are few. In this paper the e-nose was applied in the tea quality classification and storage time prediction.The preparation experiment was carried out based on two groups of the tea under the flow speed of pump 100ml/min and the tea sample 5g. The variance analysis showed that different headspace volume had significance affect on the response signal of the e-nose, especially it was more significance between the big volume beaker and little volume beaker. Then the result of the relative standard deviation analysis showed that greater headspace volume was advantageous to the stability of the response signal. Different headspace generation time had significance affect on the e-nose response, too, but the difference mainly because of too long time settling. When the headspace generation time became short, the difference was not very significant. The vapor could cause the signal change, the result of relative standard deviation analysis showed that the vapor had significant effect on the e-nose response signal. The superior sampling and clearing time was 60 s and 70 s respectively.The research was performed based on five groups of tea, tea solution and tea-leaf respectively. Many feature values were extracted and the original feature vector was consisted. Principle component analysis (PCA) was used to decrease the dimension of the data sets and genetic analyisi (GA) was used to choose the optimum feature values. Linear discriminant analysis (LDA) and BP neural network were used in the pattern recognition to classify different tea samples.The results of LDA showed that the response signal of the tea was prone to distinguish different storage time. The response signal of the tea solution was prone to distinguish different tea quality grade. Regarding the tea-leaf, the result was bad both for the quality grade and the storage time.The storage time of the tea samples were tested by BP neural network. The testing results showed that in the all tested samples the testing error of exceeding 10 days was only 8.8% based on the response signals of tea. The maximum testing error for 120 days was 31 days (T 100) and the maximum testing error for 180 days was 21 days (T 100). The mean testing error based on the response signals of tea solution and tea-leaf was bigger than that of the tea.The correlation between volatile component of tea and response signal of e-noseThe change of the volatile components was very complex, some component decreaseed according to the storage time prolonging and the quality grade deteriorating. Contrarily, some components would increase. The odor property of the tea is determined by many components mixed with an appropriate rate. The relation of the volatile components and the response sigals of the e-nose was analysized by the linear regression model.The category of the volatile components was regarded as the dependent variables, the response singals of 10 sensors were regarded as the independent variables. Step linear regression model was established. The relation of the volatile component and response signal was analysized. The result showed that according to the properties of the sensors, some independent variables were positive correlated with the dependent variables, some independent variables were negative correlated with the dependent variables, and some independent variables were not introduced to the regression model because that they were not correlated with the dependent variables.The volatile components under different storage time were analysized. The response singals of the sensors were regarded as dependent variables, the volatile componets that were significant correlated with the storage time were regarded as independent variables. The linear regression model was established. The correlation between the volatile components and response signal was analysized under different storage time.