基于机器学习与含水率影响因子的叶丝烘干参数自适应控制算法  

Adaptive Control Algorithm of Leaf Drying Parameters Based on Machine Learning and Water Content Influencing Factor

作　　者：樊盛炯 邱晓峰[1] FAN Shengjiong;QIU Xiaofeng(Shanghai Tobacco Factory of Shanghai Tobacco Group Co.,Ltd.,Shanghai 200082,China)

机构地区：[1]上海烟草集团有限责任公司上海卷烟厂,上海200082

出　　处：《计算技术与自动化》2025年第1期125-130,共6页Computing Technology and Automation

摘　　要：为提升烟丝生产质量,使不同批次间叶丝含水率均符合标准,研究了基于机器学习与含水率影响因子的叶丝烘干参数自适应控制算法。选取包括加料流量及出口含水率、烘丝机筒壁温度等13个叶丝烘干含水率影响因子,经递归特征消除法完成影响因子筛选后,构建筛选后叶丝烘干含水率影响因子的样本数据集,作为以卷积神经网络LeNet-5构建叶丝出口含水率预测模型的输入样本,预测叶丝出口含水率;在PID反馈控制中输入预测含水率与标准含水率之间的差值,输出最佳叶丝烘干参数,实现叶丝出口含水率控制;并利用双层EWMA控制模型平稳控制叶丝批次内的出口含水率、协同控制批次间的出口含水率。实验表明:该算法可准确预测叶丝烘干含水率,增加叶丝烘干含水率控制稳定性,提高卷烟制丝生产过程中叶丝烘干含水率的均质化水平。In order to improve the quality of tobacco production and make the water content of leaf silk in different batches meet the standard,an adaptive control algorithm of leaf silk drying parameters based on machine learning and water content influencing factor was studied.Thirteen influencing factors of leaf drying moisture content,including feed flow and outlet moisture content of moistening leaves,temperature in cut-tobacco drier moisture reclaimed area,were selected.After the selection of influencing factors by recursive feature elimination method,sample data sets of influencing factors of leaf drying moisture content after screening were constructed,which were used as input samples for the prediction model of leaf drying moisture content of leaf filaments constructed by convolutional neural network LeNet-5,to predict leaf wire outlet water content.The difference between the predicted moisture content and the standard moisture content is input in PID feedback control,and the optimal drying parameters are output to realize the moisture content control at the outlet of the filament.The double-layer EWMA control model is used to control the export water content of the fillet batch smoothly and coordinate the export water content of the batch.Experiments show that the algorithm can accurately predict the drying moisture content of leaf silk,increase the control stability of drying moisture content of leaf silk,and improve the homogenization level of drying moisture content in the process of cigarette silk production.

关 键 词：机器学习 含水率 影响因子 叶丝烘干参数 自适应控制 卷积神经网络 

分 类 号：TS432[农业科学—烟草工业]