Development of Corn Drying Process Optimization System Based on LSTM and NSGA ‐ II Algorithms

ABSTRACT The process optimization of corn drying is difficult due to the strong coupling of multidimensional parameters and time‐varying lag characteristics of hot air material interaction. In view of this, this study proposes a drying process optimization system that combines dynamic feature weighted long and short‐term memory networks with elite oriented multi‐objective evolutionary algorithms. It achieves accurate prediction of moisture content and real‐time generation of Pareto near optimal set of process parameters through adaptive feature weighting and dynamic guidance mechanism of archives. The experiment showed that the prediction model of the proposed system significantly reduced the prediction deviation of key indicators such as moisture content and crack rate, and remained within 0.03% and 0.06% throughout the process. In terms of optimizing performance, the Pareto solution set obtained by the system had higher quality, with a key evaluation index of 0.98, which was more than 23% higher than the existing optimal methods. Under actual testing and application, the system has achieved a drying energy consumption as low as 0.72 MJ/kg, with an average energy saving of over 40% compared to existing optimization schemes. At the same time, the crack suppression effect was stable, the energy consumption of key process nodes was reduced by 48%, and the computational resource requirements were dynamically adapted to changing working conditions. The constructed system can effectively improve the stability of drying quality and the synergy of energy efficiency, providing reliable technical solutions for postharvest loss reduction of grain and intelligent upgrading of drying processes.