摘要
雪茄烟茄胚吸阻作为雪茄烟设计制造中的核心指标,对其产生影响的因素繁多且具备复杂的非线性特性.传统经验模型基于大量实践所得,往往难以准确量化并指导雪茄烟的设计和生产过程.为快速便捷地测定全叶卷雪茄茄胚的吸阻,建立了以雪茄烟卷制中9个关键参数为输入层的多层感知神经网络预测模型.经过验证,该预测模型的平方和误差为4.908,相对误差为0.572,在茄胚吸阻处于300~500 Pa的区间内,模型的预测值更趋近于实测值.结合两种雪茄烟卷制手法,建立神经网络模型并对模型进行了生产验证,结果表明,模型预测值与实际值之间的差异并不显著,成功实现了生产场景下茄胚吸阻的快速预测.
The suction resistance of cigar wrappers,serving as a core metric in the design and manufacturing of cigars,is influenced by numerous factors exhibiting complex nonlinear characteristics.Traditional empirical models,derived from extensive practical experience,often struggle to provide accurate quantitative guidance for the design and production processes of cigars.To rapidly and conveniently determine the suction resistance of full-leaf cigar wrappers,a multi-layer perception neural network prediction model was established,utilizing nine key parameters from the cigar rolling process as the input layer.Upon validation,the prediction model demonstrated a sum of squared errors of 4.908 and a relative error of 0.572.Notably,within the suction resistance range of(300-500)Pa,the model s predictions aligned closely with actual measurements.Additionally,neural network models were developed by incorporating two cigar rolling techniques used in production,and subsequent production validation confirmed that the differences between the model s predictions and actual values were insignificant.This successful implementation enables rapid prediction of wrapper suction resistance in production scenarios.
作者
时永楠
柏明月
郭天恩
仕小伟
庄铭慧
王振亚
公斌
SHI Yongnan;BAI Mingyue;GUO Tianen;SHI Xiaowei;ZHUANG Minghui;WANG Zhenya;GONG Bin(Jinan Cigarette Factory of China Tobacco Shandong Industrial Co.,Ltd.,Jinan,Shandong,China 251000)
出处
《昆明学院学报》
2024年第6期18-23,共6页
Journal of Kunming University
基金
山东中烟工业有限责任公司科技重点项目(202101003).
关键词
雪茄
茄胚吸阻
神经网络
预测
cigar
embryonic suction resistance
neural network
prediction
