Prediction of Water-Soluble Sugars in Tobacco Casings with Models Based on Near Infrared Spectrometry

In order to evaluate the stability of the preparation quality of tobacco casings, models for the rapid prediction of water-soluble total sugar and reducing sugar in tobacco casings were established based on near infrared spectrometry. Continuous flow method was used as reference, the total sugar and reducing sugar in 120 tobacco casing samples from 5 different brands were determined. The samples were analyzed with near infrared spectrometer, and the original spectra were preprocessed with second derivative and Norris filtering smoothing. The quantitative analysis models for total sugar and reducing sugar were established by using partial least square method in the spectral ranges of 5 400-7 600 cm^-1 and 4 000-4 600 cm^-1. It was shown that when the number of principal factor was 6, the correlation coefficients of calibration (R_c), prediction (R_p) and cross validation (R_cv) for total sugar in the quantitative analysis model were 0.998 5, 0.998 1 and 0.998 8, respectively, and the root mean square errors of calibration (RMSEC), prediction (RMSEP) and cross validation (RMSECV) were 0.257, 0.261, and 0.284, repectively; R_c was 0.999 0, with R_p of 0.998 7, R_cv of 0.999 2, RMSEC of 0.235, RMSEP of 0.250, RMSECV of 0.266 for reducing sugar in the quantitative analysis model. Though paired T-test statistical analysis, there was no significant difference between continuous flow method and this method in the determination of water-soluble sugars in tobacco casings. The contents of water-soluble sugars in 112 test tobacco casing samples were predicted with the established models. The stability of the preparation quality of test tobacco casings were estimated by the 3σ process control principle of statistical process control technology, and the accuracy of results was verified according to the physical and chemical indexes of the samples. The results showed that No. 5, 14, 28 samples from brand A and No. 21 sample from brand C were unnormal, and the contents of water-soluble sugars in brand B, D and E were all within the quality control ranges, indicating that the quality of sample preparation was stable.