基于聚类和逐步判别分析的烤烟感官质量特征分类评价

Classified Evaluation of Sensory Quality Characteristics of Flue-cured Tobacco Based on Clustering and Stepwise Discriminant analysis

为优化烟叶原料感官质量的评价方法,促进配方设计中不同质量特点烟叶的合理配置,采集陕西中烟曲靖沾益基地县不同等级的初烤烟叶,运用系统聚类和Fisher线性判别法,建立基于化学成分的烤烟感官质量特征分类评价方法。结果表明:聚类分析可将陕西中烟曲靖沾益基地县烟叶的感官质量特征合理分类,三类样品的主要特征分别为香气和烟气浓度低、口感有优势;杂气大、口感差、生理满足感有优势;香气品质有优势。经交叉分析,糖氮碱的协调对香气品质的影响较大,蛋白质、总氮对生理满足感和口感特征的影响较大,两糖比对香气、烟气的充足程度和口感特征的影响较大。建立了烟碱、总氮、还原糖、淀粉、氮碱比、两糖比、糖蛋比和糖氮比等8项指标的烟叶感官质量特征分类评价模型,模型的自身和交互验证平均准确率达80.8%,验证效果较好。

The classified evaluation method of flue-cured tobacco sensory quality characteristics was established based on chemical components by the system clustering analysis and Fish linear discriminant method after the sensory quality of flue-cured tobacco leaves with different class collected from Zhanyi County was analyzed to optimize the evaluation method of sensory quality of flue-cured tobacco raw material and promote the rational allocation of flue-cured tobacco leaves with different quality characteristics in formulation design. Results： The flue-cured tobacco samples from Zhanyi County can be rationally divided into three groups by clustering analysis according to their sensory quality characteristics. The main characteristics of Group Ⅰ ,Group Ⅱ and Group Ⅲ is low aroma and smoking concentration with good taste, poor offensive odor and taste with good physiological satisfaction and good aroma quality respectively. The cross analysis shows that the harmony among sugar, nitrogen and nicotine has the great effect on aroma quality of flue-cured tobacco, protein and total nitrogen have obvious effects on physiological satisfaction and taste characteristics of flue-cured tobacco, and the sugar-reducing sugar ratio has the significant effects on aroma, smoking sufficient degree and taste characteristics of flue-cured tobacco. The classified evaluation mode of flue-cured tobacco sensory quality characteristics is established based on eight indexes of nicotine, total nitrogen, reducing sugar, starch, nitrogen-nicotine, sugar- reducing sugar ratio, sugar-protein ratio and sugar-nitrogen ratio. The self and cross validation indicates that the average accuracy of the classified evaluation mode is 87.4% and 74.1% respectively.