The Relationship Between Main Physical Properties And Chemical Components And Neutral Aroma Matters In West Henan Province

The experiments were conducted in Luoyang and Sanmenxia flue-cured tobacco growing area in west Henan. The relationship between main physical properties and chemical components and neutral aroma in west Henan Province was studied. Statistical methods was used to analyze the relationship between the physical properties and chemical components in western Henan; the regression models were made between physical properties and chemical components, and the changing trend of neutral aroma matters under the effect of different physical properties was also studied. With the help of BP neural network technology, the neutral networking models were established by the relationship between physical properties and chemical components, and physical properties between aroma matters. The main conclusions are as follows:There was a wide variation among physical properties, chemical components and neutral aroma matters of western Henan tobacco. Leaf thickness ranged from 58.80μm to 118.52μm; single leaf weight ranged from 7.67g to 12.95g; leaf density ranged from 55.44g/m2 to 95.37g/m2. Variation coefficients of leaf density was small(12.85%), and the average of leaf density was 68.86g/m2; variation coefficients of leaf thickness was big(16.79%), and its average was 81.83μm. There was a wide variation among chemical components of Henan tobacco. The variation coefficient of total sugar was 13.62% and the average was 28.08%; the average value of potassium to chloride ratio was 7.16, with the highest variation coefficient (50.15%). In neutral aroma matter terms, neophytadiene had the highest concentration (401.48μg/g), and carotenoid degradation products concentration was the second (70.44μg/g); aromatic amino acid degradation products had the lowest concentration(13.20μg/g), and the variation coefficient of aromatic amino acid degradation products was the highest(64.76%), followed by the concentration of neophytadiene (58.19%), and the concentration of carotenoid degradation products was lowest(28.77%).The liner or curve regressive equations were built between every chemical index and leaf thickness, single leaf weight and leaf density of flue-cured tobacco. The relationship could be described by quadratic equation y? = ?1.316+0.067x+0.001x2 between total nitrogen content and leaf thickness; relationship of reducing sugar could be described by linear equation y? = 12.815+0.122x with leaf density; the relationship could be described by linear equation y? = ?1.721+0.237x between leaf density and the ratio of sugar/nicotine; relationships of leaf density could be described by complex equation y? = 0.683×1.010xwith potassium content; relationships of nicotine could be described by complex equation y? = 2.955×0.991x with leaf density; the relationship of the ratio of nitrogen/nicotine could be described by linear equation y? = 0.335+0.010x with leaf density. All above equations were significant at the 5% level.The relationship between physical properties and the neutral aroma of flue-cured in western Henan was analyzed. It was found that the relationship of the neutral aroma with leaf thickness could be described by S curve equation y? = e(5.573+53.659/x), and its relationship with single leaf weight could be described by S curve equation y? = e(5.803+5.408/x), and its relationship with leaf density could be described by composite function y? = 1557.168×0.987x. All the above equations were significant at 5% level, and the relationships between the neutral aroma and leaf thickness, leaf density were significant at 1% level. Cluster analysis indicated that the neutral aroma content of the leaves, which had the medium thickness, was highest, and that of thinner leaves was low. The neutral aroma content of the leaves, which had the light single leaf weight, was highest, and that of the leaves, which had the heavy single leaf weight was lowest. The neutral aroma content of the leaves, which had the small leaf density, was lowest, and that of the leaves, which had the high leaf density, was highest.Based on BP neural networks, data samples used for network training and testing were structured. Simple classification, normalization and principal component analysis were applied to analysis the experimental data according to the actual situation. 70% of the sample data were used for training the neural network model, and the remaining 30% were for validating the trained network prediction capabilities. After the selection of hidden layer neurons, the choice of activation function, training function selection, BP neural network model and multiple linear regression model were compared. The results showed that BP neural network model can predict the chemical components and neutral aroma components in tobacco very well, and prediction accuracy is usually higher than the linear regression model.In this paper, the artificial forecast of tobacco quality was studied from new perspective, which deepened the understanding of the tobacco quality forecast. With the development of the analysis technology in quantity of tobacco chemical components, the improvement of diagnostic technology, the further study of neural network and optimized technology, it will be helpful for solving the problem of the tobacco quality forecast.