SIMCA分类法与偏最小二乘法结合近红外光谱检测卷烟的内在品质

The quality analysis of cigarette by partial least squares and SMICA combined with fourier transform near infrared spectra

针对卷烟内在品质常规化学检测过程烦琐、速度慢的特点,本文探讨了基于化学计量学方法与近红外光谱分析技术结合判别卷烟品质等级和同时检测多个品质指标的快速方法。通过采用SIMCA分类法与偏最小二乘法结合卷烟的傅里叶变换近红外漫反射光谱,建立了3个不同级别卷烟的分类模型,用于甄别卷烟品质真伪,结果准确；同时,又建立了检测卷烟中总氮、总糖、烟碱主要化学成分含量和其主流烟气中焦油、烟碱、一氧化碳释放量等品质指标的校正模型,其相关系数分别为：0．9844、0．9918、0．9968、0．9505、0．8945和0．9625；交互效验均方残差分别为：0．0663、0．6682、0．0607、0．2262、0．0350和0．3685。另取未知样品对模型进一步验证,经配对t-检验表明,在显著性水平大于0．05的条件下,其测定结果与标准方法的测定结果对比,两者均无显著性差异。该法应用于卷烟品质真伪鉴别和品质指标检测,操作简单、快速、准确。

The routine chemical analysis of cigarette quality was loaded down with trivial details, and slow. A fast approach based on methods of chemometrics combined with Fourier transform near infrared spectrometry, which, was discussed in this paper. The approach was intended for discrimination of cigarette quality, and simultaneous determination of many quality indexes. By using partial least squares and SMICA techniques combined with Fourier transform near Infrared spectra, the classification methods were established for discrimination of three grades cigarettes, which, was used to identify bogus cigarette with correct results. Meanwhile, the calibration models were established for determination of total nitrogen, total sugar, nicotine in cigarettes, and tar, nicotine, carbon monoxide in mainstream smoke. Correlation coefficients of the models were 0. 9844,0. 9918, 0. 9968, 0. 9505, 0. 8945and 0. 9625, and the root mean square error of cross validation （RMSECV） were 0.0663, 0.6682, 0.0607, 0. 2262, 0.0350 and 0.3685, respectively. The calibration models had been applied to determination of another unknown samples. By paired-samples t-test, the results were compared with those of standard methods and neither of them was significant difference over 0.05 level. The approach can be used to discrimination and determination of cigarette quality, which, was simple and fast with accurate predicted results.