[Quality assessment of tobacco flavor by classification of principal component analysis-mahalanobis distance combined with FTIR-ATR fingerprint].

In normal condition, 150 Fourier transform infrared-attenuated total reflection spectra of three kinds of tobacco flavor were measured. Mean centering and Karl Norris second derivative filter were employed for treating the spectra data. Whereafter, class models of tobacco flavor were established with classification of principal component analysis-Mahalanobis distance. Class models have been applied to multivariate statistical process control(MSPC) for the quality assessment of tobacco flavor in site. The upper control limit(UCL) was based on this kind of average Mahalanobis distance within +3sigma. The results of monitoring are correct.