本文用厚膜氧化锡半导体气敏传感器阵列对三种品牌卷烟烟气进行分析,详细阐述了实验过程及确定传感器阵列的组成方法,并从样本中用马氏距离选取合适的样本,用主成分分析法和神经网络聚类分析法对样本进行分析,主成分分析结果已较好的把各品牌的烟分开,神经网络对三种品牌烟的识别率分别为中华烟85%、云烟90%、红梅烟95%。 In this paper, a thick film of tin oxide semiconductor gas sensor array for three brands of cigarette smoke analysis, elaborated the experimental process and to determine the composition of the sensor array method, and Mahalanobis distance from the sample to select the appropriate sample, with the main component Analysis and neural network clustering analysis of the sample analysis, the principal component analysis has better separation of the various brands of cigarettes, neural network identification of three brands of cigarettes were 85% of China Tobacco, Yunyan 90% Red plum smoke 95%.