Huang Zugang (Physical electronics) Directed by Prof. Li JianpingElectronic nose(EN) technique has become an international research focus and has made rapid progress during the latest years. And practical EN products have been extensively applied in many fields.In this dissertation, the construction and application of EN systems in the identification of different brands of cigarettes and the detection of alcohol with gasoline interference were analyzed, and progress was achieved in the design and realization of gas sampling system, pattern recognition algorithm and experimental strategies.A gas sampling system was designed to accurately provide testing gases in demanded concentrations, and two other were specialized for the application in the identification of different brands of cigarettes and the detection of alcohol with gasoline interference. Array chambers of different structures were fabricated, and the influence of the chamber's dimension and structure on the performance of the gas sampling system was discussed in detail.EN pattern recognition system was studied. Feature extraction through 2-order polynomial fit of the descending part of the response curve made possible a timesaving measurement process. The performances of two pattern recognition algorithms, namely principal component analysis (PCA) and linear discriminant analysis (LDA) in practical problems were discussed. Artificial neural network(ANN) was utilized with back-propagation algorithm (BPA), and the combination of PCA/LDA with ANN improved the identification performance of the system.The influence of the array chamber, carrier gas flux and sample temperature on the performance of the system was discussed in the identification of different brands of cigarettes. While in the detection of alcohol with gasoline interference, the methods of signal preprocessing showed great effects. And at the end, suggestions were made on the optimization of the system construction and measurement strategies, which advanced towards portable EN systems. |