为提升稻谷干燥过程中的品质,并准确预测干燥过程中稻谷的水分变化及其对质量安全的影响,本研究以粳稻为研究对象,通过爆腰增率和干燥时间作为评价指标,结合单因素试验与正交试验分析,对干燥工艺进行优化。文章探讨了不同干燥温度、风速及初始含水率条件下,稻谷水分含量及品质的变化规律；并提出了一种融合自适应变异和精英策略优化(Adaptive Mutation and Elite Strategy Optimization,AEO)的遗传长短期记忆神经网络模型(AEO-GA-LSTM),用于稻谷干燥过程中的水分预测。结果表明,干燥温度和风速对稻谷的爆腰增率和干燥时间均具有显著影响(P<0.01),各因素的影响顺序为:干燥温度>干燥风速>稻谷初始水分,随着温度和风速的升高,干燥速率加快(P<0.01),稻谷的爆腰增率也显著增加；通过构建并对比BP、LSTM、GA-LSTM和AEO-GA-LSTM模型在不同干燥条件下的时序数据预测效果,结果显示,改进的AEO-GA-LSTM模型综合决定系数R2为0.9970,均方根误差为0.0784,优于BP、LSTM和GA-LSTM模型的误差值0.2252、0.1964和0.1420,表现出更好的拟合度和更高的准确率。因此,本研究建立的AEO-GA-LSTM水分预测模型可以为热风干燥条件下稻谷的水分预测提供一种新的思路和方法参考,有助于提升稻谷干燥工艺的自动化水平和品质控制能力。

To enhance the quality of rice drying processes and accurately predict moisture changes during drying and their impact on quality and safety, this study focuses on japonica rice, using fissuring rate and drying time as evaluation metrics. The study optimizes the drying process through single-factor experiments and orthogonal tests, and explores the effects of different drying temperatures, wind speeds, and initial moisture content on the moisture content and quality of rice. An Adaptive Mutation and Elite Strategy Optimization (AEO) integrated with a genetic algorithm-long short-term memory neural network model (AEO-GA-LSTM) is proposed for predicting moisture content during the rice drying process. The results demonstrate that drying temperature and wind speed significantly affect the fissuring rate and drying time of rice (P<0.01), with the order of influence being: drying temperature > wind speed > initial moisture content. As temperature and wind speed increase, drying rates accelerate (P<0.01), leading to a significant rise in the fissuring rate. By constructing and comparing the time-series prediction performance of BP, LSTM, GA-LSTM, and AEO-GA-LSTM models under different drying conditions, it was found that the improved AEO-GA-LSTM model achieved a coefficient of determination (R2) of 0.9970 and a root mean square error (RMSE) of 0.0784, outperforming the BP, LSTM, and GA-LSTM models, which had RMSEs of 0.2252, 0.1964, and 0.1420, respectively. The AEO-GA-LSTM model exhibited better fitting and higher accuracy. Therefore, the AEO-GA-LSTM moisture prediction model established in this study offers a new approach and methodological reference for moisture prediction in hot-air drying conditions, contributing to the improvement of automation and quality control in rice drying processes.

河南省科技创新领军人才,粮食减损增质智能干燥关键技术研究(244200510021)；河南省高校科技创新团队,粮食智能无人系统及装备(24IRTSTHN030)