基于热分析图谱的卷烟配方原料分类分档模型构建与应用

Construction and application of the classification model for cigarette blend raw material based on thermal analysis spectrum

何伟 ¹杨韧强 ¹刘泽春 ¹蔡国华 ¹邓小华 ¹叶仲力 ¹熊哲 ²余志强 ¹梁晖 ¹李菁菁 ¹李巧灵¹

扫码查看

作者信息

1. 福建中烟工业有限责任公司,技术中心,福建省厦门市集美区滨水路 298 号 361021
2. 华中科技大学,煤燃烧国家重点实验室,湖北省武汉市洪山区珞喻路 1037 号 430074
折叠

摘要

[目的]提高卷烟配方原料分类分档的工作效率,帮助配方人员客观掌握配方原料的年度间波动.[方法]提出一种热分析图谱结合支持向量机算法(SVM)对卷烟配方原料进行分类分档的新技术.采用python3 中sklearn中的SVM模块,通过核函数和one-against-all方法并选择合适的惩罚参数,对 2016-2018 年度的 129 个样品进行训练学习,并对 33 个样品进行测试.当模型预测准确率达到应用需求后,对 2019 年度46 个卷烟配方原料进行分类预测.[结果]129 个训练集样品的分类准确率为 93.02%,33 个样品测试集样品的分类准确率为 84.85%,46 个卷烟配方原料的分类准确率为 84.78%.

Abstract

[Background]This study aims to improve the work efficiency of classification and grading of cigarette blend raw materials,and help formula personnel to objectively grasp the annual fluctuation of formula raw materials.[Methods]A new technique is proposed for classifying and grading cigarette blend raw materials by combining thermal analysis spectra with the Support Vector Machine(SVM)algorithm.The SVM module in sklearn of python3 is used,and through the kernel function and one-against-all method and selecting appropriate penalty parameters,129 samples from the years 2016-2018 are trained,and 33 samples are tested.After determining the prediction accuracy of the model,the model was applied to the classification evaluation of 46 samples in 2019.[Results]The classification accuracy rate of the 129 training set samples was 93.02%,the classification accuracy rate of the 33 test set samples was 84.85%,and the classification accuracy rate of the 46 cigarette blend raw materials was 84.78%.

关键词

卷烟配方原料/分类/热分析图谱/支持向量机

Key words

cigarette blend raw material/classification/thermal analysis spectrum/support vector machine

引用本文复制引用

基金项目

福建中烟工业有限公司科技项目(FJZYJH2021ZD007)

出版年

2024

中国烟草学报

中国烟草学会

中国烟草学报

CSTPCD北大核心

影响因子：1.182

ISSN：1004-5708

参考文献量27

段落导航