Discrimination of LongJing green-tea grade by electronic nose

An investigation was made to evaluate the capacity of an electronic nose (E-nose, PEN2) to classify the tea quality grade. Four tea groups (A120, A280, A380 and A600) with a different quality grade were employed. In the experiment, the volume of vial and the headspace generated time were considered corresponding to the 5 g tea samples, and the optimum experimental procedure was determined by using the variance analysis (ANOVA), multivariance analysis and principle components analysis (PCA). The results showed that the volume of vial affected the result of discrimination, and the headspace generated time had no significant effect on the E-nose response. The four tea groups were measured and response values at four different collection times were conducted by PCA, linear discrimination analysis (LDA) and artificial neural network (ANN). Only A120, A380 and A600 could be discriminated by PCA. However, the four tea groups were discriminated completely by LDA. The response value of the E-nose at 60 s was optimum to be used for discrimination. The method of ANN (network topology 20-12-4) was performed and 90% of the total tea samples were classified correctly by using the back-propagation neural network.