Establishment And Application Of Processing Tolerance Model Of Tobacco Leaves Based On Physical And Chemical Properties

Tobacco processing resistance is one of the important indicators in the process of tobacco processing,which directly affects the degree of shredding in the process of processing,which in turn affects the loss of tobacco raw materials and the quality of processing.It is an important mechanical property of tobacco.The processing resistance of tobacco leaves is not only affected by the mechanical properties of tobacco leaves,but also closely related to its own characteristics and chemical components.In order to comprehensively establish a model for evaluating the processing resistance of tobacco leaves,this topic studied the physicochemical properties of tobacco leaves in different regions and grades,used BP neural network to establish a comprehensive prediction model of tobacco leaf processing resistance,and verified the model in production.The results show:(1)There are great differences in the physical properties of tobacco leaves between different parts in different regions.There are varying degrees of correlation between the processing resistance of tobacco leaves and the main physical properties.The mathematical model of the comprehensive factor score of the physical properties of tobacco leaves is F=0.46F_Fac-1+0.23 F_Fac-2+0.20 F_Fac-3+0.11 F_Fac-4.According to the results of grey correlation analysis,11 other physical indicators except adhesiveness were selected as the key indicators of comprehensive physical properties to characterize the processing resistance of tobacco leaves.(2)There were significant differences in the conventional chemical components of different grades of tobacco leaves in different regions.The processing resistance of tobacco leaves was significantly correlated with conventional chemical components.Through grey correlation analysis,the effect of conventional chemical components on processing resistance was reducing sugar*nitrogen>reducing sugar>potassium>total alkaloids*reducing sugar>nitrogen>chlorine>total alkaloids*chlorine>total alkaloids.Reducing sugar,potassium,sugar-to-nitrogen ratio,sugar-to-alkali ratio and nitrogen content were selected as key indicators of comprehensive chemical properties to characterize the processing resistance of tobacco leaves through grey correlation analysis.(3)The BP neural network model can better express the nonlinear relationship between the processing resistance and physicochemical properties of tobacco leaves.The canonical correlation analysis found that the two groups of canonical variables reached a very significant level,explaining 20.1%of the variation in physical properties and 27%of the variation in chemical components.The explanatory power was low,and the model could not truly reflect the relationship between the two groups of variables.The mathematical model of comprehensive factor score of tobacco leaf physical and chemical properties index was established by factor analysis as F=0.504 F1+0.204 F2+0.152 F3+0.078 F4+0.062 F5.Through grey relational analysis,11 indicators were selected as the physical and chemical indicators to characterize the processing resistance of tobacco leaves.Through the BP neural network,the prediction accuracy of the overall model reached 99.697%,and the root mean square error(RMSE)was 0.58641 and the mean square error(MSE)was 0.34387.,the mean absolute error(MAE)is 0.47199,the mean absolute percentage error(MAPE)is 0.017574,and the error is1.7%,indicating that the established prediction model is effective.(4)The BP neural network model was selected for the production verification of the loosening and moisturizing process in the silk production line.The coefficient of variation was less than 5%,indicating that the model was relatively stable and had good reproducibility.According to the results of the prediction model,the tobacco leaves were clustered and analyzed to provide a theoretical basis for grouping processing.