基于逐步判别分析的烤烟中上部烟叶香型判断

摘要

The aim of this study was to establish mathematical models for judging the aroma types of middle and upper flue-cured tobacco leaves according to the contents and proportions of aroma compositions. [Method] The aroma types of tobacco leaves were judged based on stepwise discriminant analysis, using 63 C3F and 65 B2F tobacco leaf samples from 13 tobacco producing regions in 11 provinces of China (Huili in Sichuan, Baokang in Hubei, Wulong in Chongqing, Lu-oyang in Henan, Zhucheng in Shandong, Wuyi Mountain in Fujian, Malong in Yun-nan, Chuxiong in Yunnan, Bijie in Guizhou, Liuyang in Hunan, Suiyang in Guizhou, Kaiyuan in Liaoning, Nanxiong in Guangdong) as calibration samples, and 67 aroma components as indices. And the Fisher discriminant functions were verified using 21 C3F and 19 B2F tobacco leaf samples. [Result] Variation coefficients of the propor-tions were lower than that of contents of most aroma components in middle and upper leaves of the samples, indicating that the proportions were more stable than contents of aroma components. The proportions of benzyl alcohol, solanone, β-dam-ascone, neophytadiene, farnesylacetone A, palmitic acid, thunbergol, methyl linole-nate and cembratriene-diol were al over 1% in both middle and upper leaves, al-though the dominant aroma components of the same aroma type varied between middle and upper leaves. Moreover, 11, 18, 7 and 11 aroma components were re-spectively introduced into the Fisher discriminant functions established based on the contents and proportions of middle and upper flue-cured tobacco leaves, which ex-hibited accuracy rates of 91.7%, 100%, 91.7% and 91.7% in the judgments of other tobacco leaf samples. The results revealed that the components those determined aroma types in middle leaves were obviously more than in upper leaves. ln middle leaves, the accuracy rates of aroma type judgment could be improved by using the proportions rather than the contents of aroma components as indices. However, the functions based on the proportions and the contents of aroma components in upper leaves gave close accuracy rates. [Conclusion] The results of the study wil provide references for identifying aroma types of flue-cured tobacco leaves in future work.%[目的]通过揭示不同香型和不同部位烤烟的香气风格特征及其形成原因，建立起中上部烟叶香型风格定量判断的数学模型。[方法]以中国11个主要产烟省的13个地区的63个C3F和65个B2F烟叶作为校正样品，并以67种致香物质为指标，采用逐步判别分析的方法对不同香型烤烟样品进行判别分析，并建立判别函数，并分别利用21个 C3F烟叶和19份B2F烟叶对该模型进行验证。[结果]中上部烟叶的大部分致香物质含量占致香物质总量的比例变异系数小于相对应的致香物质绝对含量，且在中上部烟叶中比例超过1%的物质只有苯甲醇、茄酮、β-大马酮、新植二烯、金合欢基丙酮 A、十六酸、寸拜醇、亚麻酸甲酯、西柏三烯二醇，且在不同部位烟叶中，主导某种香型的致香物质种类不同；在分别以中部叶和上部叶致香物质含量和比例为判断指标构建判别函数时，各引入了11种、18种、7种、11种致香物质指标，并以构建的模型对新样本进行预测，其整体准确率分别为：91.7%、100%、91.7%、91.7%。可见，影响中部烟叶香气风格特征的致香物质指标明显多于上部叶，但在中部叶中采用致香物质比例作为判别指标能明显提高其判别准确率。[结论]该研究为特色优质烟叶的开发与生产布局提供理论依据。