Classification And Identification Of Liquors Based On Gas Sensor Array

Liquors' production and consumption occupy an important position in the national economy. Along with the development of economy and society, as the improvement of people's living standard, the problem of liquors' quality is paid more and more attention. In recent years, poisoning incidents caused by forged liquors frequently occurre. liquors' market is filled with fake commodities and unlicensed business, etc., which is harmful to health. Therefore, it relates to consumers' health that liquors are quickly and accurately measured. Liquors evaluation based on electronic nose is more objective, rapid and accurate than sense evaluation and chemical analysis. In this paper, a sensor array composed by seven commercial TGS sensors of Figaro company, and the electronic nose system made in laboratory are adopted to dynamically measure liquors, classification and recognition of different kinds of liquors are analyzed.The influence of evaporation temperature on the results of liquors classification were studied. Liquor samples were respectively measured after being water-bath heated at different temperatures. Seven different kinds of liquors were respectively measured at 20℃,40℃,60℃and 80℃after water-bath heating. The classification of liquors according to the aromatic types and percentages of alcohol was analyzed by using principal component analysis (PCA) method. With the evaporation temperature rising, the results of liquors classification turned better and better. The best results of classification were obtained at 60℃.The analysis of liquors classification according to different evaporation temperatures by using PCA shows that, with the evaporation temperature rising, the concentration of trace components in saturated steam is increasing, and liquor samples presente better clustering features. Liquor samples were best clustered at 60℃; Because the evaporation temperature 80℃was higher, more water vapor in saturated steam affected evaluation of liquors based on electronic nose, so liquor clustering was worse than 60℃.Analysis of characteristics in liquors is very useful to liquors classification and liquors quality control. Eghit liquors were dynamicly measured by using the same gas sensor array. The brands, aromatic types, raw materials, percentages of alcohol, production areas and manufactories of liquors were choosen to evaluate liquors. The analysis of PCA showed that, aromatic types and raw materials were better to characterize liquors than percentages of alcohol,.The best results of classification were obtained when classified by raw materials. Liquors of the same manufactory mostly came from the same production areas. Therefore, the results of liquors classification were consistent when classified by production areas and manufactories. A few liquors overlapped together when classified by percentages of alcohol.Pattern recognition algorithm plays an important role in electronic nose system. According to percentages of alcohol, aromatic types and raw materials,10 liquors were recognized by using BP-ANN, probabilistic neural network (PNN), LVQ and K-nearest neighbor (k-NN). The accuracy of BP-ANN and PNN were 100%,100% and 95%according to aromatic types, raw materials and percentages of alcohol, respectively. The results of identification were ideal. K-NN also presented strong recognition ability in liquors identification of different characteristics. The accuracy of LVQ was 85%,90%and 70%according to aromatic types, raw materials and percentages of alcohol, respectively.The average accuracy was relatively low.