Near Infrared Spectroscopy In Usability Studies, The Analysis Of Tobacco Ingredients

There exists inherent relationships between organic molecular structure and its near infrared spectrum (NIR). Many quality characters of tobacco can be analyzed by the means of NIR. For it's lossless and fast, NIR spectroscopy technology can be used to analyze main components e.g. nicotine and total sugar in tobacco, polyhydroxybenzene and olding impact of tobacco leaves as well as smoking tar in cigarettes by building a suitable mathematic model. This method can also be used to distinguish if the cigarettes are eligible or fake. So the NIR spectroscopy technology has great application potential in flowing fields of tobacco industry: detecting components of tobacco in laboratories, manufacturing of cigarettes, olding of tobacco leaves, quality analysis of tobacco leaves during shopping and stocking, quality control of tobacco leaves during re-curing and quality cautioning during trading. For the above reasons, the tobacco group corporations vie each other to develop near infrared spectroscopy technology for the maximal benefit. But the variance of the raw materials system and the interests make it impossible for them to build a general mathematic model of NIR. The data from different model are not comparable.In the first part of the paper, the developing of near infrared Spectroscopy technology and the technology's application to tobacco components analysis was summarized.In the second part, the possibility to establish a general protecting model for tobacco components was studied. After mainenervations being drawn out and being dried for 4 hours at 60℃,1156 samples of tobacco leaves from different tobacco producing-area were made into powder (≤0.13 mm). Reducing sugar, total sugar, nicotine, nitrogen, potassium (K) and chlorine of more than 1000 samples from different tobacco producing-area were determined with classical chemical or instrumental methods. By scanning these samples with Fourier transform Near Infrared (FT-NIR) spectrometer (MARIXT-I produced by Bucker) we obtained their spectrograms, pretreated the spectrograms with a chemometrics software (OPUS 5.5) to debase the noises' influence on spectrum information, and built a chemometrics model by means of Partial Least2Squares (PLS). The chemometric model was optimized some times and validated with some other samples. Result indicates that it is feasible to develop a general chemometric model to detect the main components in tobacco samples from all over the tobacco producing-areas, and the data obtained from the model are very close to that of classical chemical or instrumental methods; there is no remarkable difference between the data from the general model and that from regional one. A general model has a higher usability.In the third part, Study on the relationship between the relativity and the veracity was also done, the result shows that the veracity is mainly determined by the veracity of original data, not the relativity; a higher relativity (R~2) of the model does not mean higher detection veracity. If the relativity beyond the limit would induce a bigger Root Mean Square Error of Cross-Validation (RMSECV).