基于双投影旋转图的烟叶样品香型聚类分析

Cluster analysis on aroma type of flue-cured tobacco samples based on rotated score-loading biplot graph

【目的】基于感官评价指标对烟叶样品进行聚类分析及类别特征分析,为烟叶原料的综合利用提供参考。【方法】对322个不同烟叶产区的样品进行分析,采用12个香韵特征指标的评吸结果,建立基于主成分分析的得分—载荷双投影旋转图,分析烟叶产区样品的聚类情况及各类别样品的主要特征。【结果】在双投影旋转图中烟叶样品可清晰地分为清香型、中间香型和浓香型三大类;12个香韵指标中,清甜香、蜜甜香和焦甜香的模长均大于0.8000,表达能力较强,是决定烟叶香型的重要指标;且香型聚类结果与产区聚类结果具有较高的一致性。【结论】双投影旋转图能够对烟叶感官评价指标及其相互关系进行定量描述,可用于烟叶样品的特征分析。

[Objective]In order to provide references for comprehensive utilization of raw tobacco leaves, the fluecured tobacco samples were clustered and the characteristics of all types were analysed based on sensory evaluation indexes. [ Method ]A total of 322 flue-cured tobacco samples were analysed, which were collected from different producing areas of China. According to sensory evaluation results of 12 aroma characteristic indexes, the rotated score-loading biplot graph based on principal component analysis were established to analyse clustering results of tobacco leaf samples and main characteristics of all types. [Result]The results showed that the obacco samples was clearly divided into three types viz., light-flavour type, moderate-flavour type and strong-flavour type in the rotated score-loading biplot graph. And among 12 aroma indexes, the modular lengths of fresh-sweetness, honey flavour and burnt-sweetness were more than 0.8000, with powerful expression ability, which was an important index for determining aroma type of tobacco. Besides, the clustering result of flavor was consistent with clustering results of producing area. [Conclusion]The rotated score-loading biplot graph can be used to describe quantitatively sensory evaluation indexes and their mutual relationships, thus which can be used to analyze characteristics of tobacco leaf.