Near-infrared Spectroscopy Analysis Of Major Physical Properties Of Flue-cured Tobacco

In order to study the feasibility of rapid determination of major physical properties of flue-curedtobacco using Near-infrared spectroscopy, comprehensive analyses on the relationship between physicalproperties and chemical compositions were carried out and near infrared mathematical models of majorphysical properties (leaf weight, filling value, thickness, equilibrium moisture content and tensilebreaking strength) in flue-cured tobacco were established using the chemometrics method. The bestmethods of modeling and testing were identified and model transfer experiments were conducted withtwo of the parameters that fit the models best with better practicability. The results are as follows：1. Simple correlation path and path analysis were used in studying the correlation between physicalproperties and chemical compositions. The results showed that the physical properties and regularchemical compositions are closely correlated. The most significant correlation was identified betweenleaf thickness and leaf weight. Significant correlation was observed between leaf weight, filling value,mass burning rate and chemical compositions/derived values. No significant correlation was observedbetween tensile strength and chemical compositions.2. Near infrared mathematical models of the major physical properties of flue-cured tobacco wereestablished via chemometrics method. The results showed that nondestructive determination of themajor physical properties of flue-cured tobacco using NIR is feasible. Better models were establishedfor those physical properties having higher correlation with chemical compositions of tobacco leaf, suchas leaf thickness, leaf weight and filing value. The results of modeling with tensile strength that is notwell correlated with chemical compositions, on the other hand, are not very good.3. Using PLS, near infrared calibration models for the major physical properties of flue-curedtobacco were established after screening different spectra pretreatments and wavelength range, externalcross-validation was then conducted. The results showed that:(1) Good near infrared models wereobtained when the sampling methods were the same as conventional measurements. Suitable modelswere obtained, when leaves were used for leaf thickness, leaf weight and tension, and when cut tobaccowas used for filling value and equilibrium moisture content. The procedure of modeling and testing forthese parameters, however is complicated, time-consuming and is not as efficient as conventionalmeasurements.(2) Good modeling results were obtained with two of the physical property parameters,leaf weight and filling value. For the leaf weight model, tobacco powder was used and the firstderivative pretreatment was chosen with the wavelength range of7848.24-3948.24cm-1Themodel’s RMSECV is0.660with a average relative deviation6.77%between the calculated values andthe actual values. For the filling value model, tobacco powder was used with the first derivativepretreatment and wavenlength range8000-4000cm-1. The model has a RMSECV of0.225and thecross-validation model’s average relative deviation is3.71%.4. The NIR models of leaf weight and filling value were successfully transferred from “maininstruments” to “auxiliary instruments” via model transfer test. Cross-validation and t-test have showed that there was no significant difference between the calculated values and the actual values. The modeltransfer is therefore successful.