Research on the prediction algorithm for experts’ set values of hot air temperature for secondary conditioning cylinders in threshing and redrying

The research aims at secondary conditioning cylinders that suffer high-frequency manual intervention and lagged adjustment in tobacco threshing and redrying. In the case of the secondary conditioning cylinders processing the ACDFAPK, DBBFAH, and DCAKAHK Tobacco Modules (2021 Season) in the Yunnan Chuxiong Tobacco Redrying Factory, it takes the equipment parameters and the processing environment’s temperature and moisture as the test data and considers the correlations as the evaluation indexes, so as to analyze the relationships between the equipment parameters given by the experts and the environment temperature and moisture. The prediction model for experts’ set values of hot air temperature is built by virtue of the deep neural network (DNN) machine learning (ML) algorithm. The results show that: (1) the primary parameter requiring adjustment based on the environment temperature and moisture is hot air temperature; (2) the prediction model for experts’ set values based on the DNN algorithm has a mean square error (MSE) of 5.745, a coefficient of determination (R-square) of 0.66, and a prediction accuracy of 98.49%. Thus, this prediction model boasts favorable fitting performance, small prediction errors, and high prediction accuracy.