Modeling and prediction of tobacco cut stem drying process using an ARIMA-LSTM hybrid model

To address the complex mechanistic modeling challenges in tobacco cut stem drying, this study proposes an ARIMA-LSTM hybrid model for accurate prediction of outlet moisture content over 180 timesteps. Leveraging the dynamic characteristics of moisture content, the method employs ARIMA to capture linear features and LSTM to learn residual nonlinear relationships. Experimental results demonstrate the hybrid model's superior accuracy over individual ARIMA or LSTM models, effectively establishing a mapping between dryer process parameters and outlet moisture content. This provides a novel technical approach for optimizing drying process control.