多次衰减全反射IR指纹图谱结合PLS-DA法在烟用香精分类中的应用

Classification of Tobacco Flavors by PLS-DA Combined with Multi-Bounce Attenuated Total Reflectance IR Fingerprint Spectra

摘要: 为识别不同风格特征的烟用香精,采集3类烟用香精的多次衰减全反射(ATR)红外指纹图谱,结合偏最小二乘-判别分析,建立了3类不同风格特征烟用香精的类模型,并用30个验证样品对类模型进行了验证。结果表明:在P>0.05条件下,验证样品均获得了正确的分类,类模型判别效果良好,同时,采用Hotelling T²统计量监控不同批次烟用香精的质量稳定性也获得了满意的结果。

Abstract: In order to identify the tobacco flavors of different style characteristics,the multi-bounce attenuated total reflectance infrared spectra of three tobacco flavors were collected,the class models for tobacco flavors of different style characteristics were established by combining with partial least square-discriminant analysis,and the models were verified with 30 validation samples.The results showed that all of the validation samples were correctly classified at P>5% level,and the class models featured good discrimination effect.Moreover,the quality consistency of tobacco flavors of different batches monitored with Hotelling T² statistics obtained satisfactory results as well.