关键词: 烟叶, 打叶复烤, Lasso, 预测模型, 常规化学成分

Abstract: To construct a prediction model for the routine chemical components of tobacco leaves during threshing and redrying, a dual feature selection mechanism based on the Lasso algorithm was employed. The chemical composition of tobacco leaves before redrying and temperature parameters from various processing stages were used as input variables to elucidate the linkage mechanism of "raw material characteristics before redrying-process temperature parameters-quality after redrying" during the threshing and redrying process. The results showed that the prediction model for chloride ion content performed the best, followed by the model for potassium oxide, and then the model for nicotine. The models for total sugar, reducing sugar, and total nitrogen content performed relatively poorly. In the chloride ion prediction model, the regression coefficients for chloride ion content before redrying, primary moistening hot air temperature, and drying zone 2 temperature were 0.34, 0.01, and -0.01, respectively. For the potassium oxide content prediction model, the regression coefficients for potassium oxide content before redrying and tempering zone 1 temperature were 0.42 and -0.14, respectively. In the nicotine content prediction model, the regression coefficients for primary moistening hot air temperature and nicotine content before redrying were 0.32 and 0.26, respectively. For the total nitrogen content prediction model, the regression coefficients for total sugar content before redrying, drying zone 2 temperature, drying zone 3 temperature, and tempering zone 1 temperature were -0.12, 0.08, -0.05, and 0.03, respectively. In the reducing sugar content prediction model, the regression coefficients for total sugar content before redrying, chloride ion content before redrying, and drying zone 3 temperature were 1.95, -0.90, and 0.60, respectively. For the total sugar content prediction model, the regression coefficients for total sugar content before redrying, chloride ion content before redrying, and drying zone 3 temperature were 2.20, -1.00, and 0.25, respectively. In conclusion, the prediction models for chloride ion, potassium oxide, and nicotine content after threshing and redrying demonstrated superior performance, whereas the models for total nitrogen, total sugar, and reducing sugar content were less effective. The temperature in the drying zones significantly influenced the routine chemical components of tobacco leaves after redrying.

Key words: flue-cured tobacco leaves, threshing and redrying, Lasso, prediction model, routine chemical components