不同形态烟叶近红外光谱转移模型研究  

Calibration Transfer Model of Near-Infrared Spectra for different physical form of tobacco

作　　者：吴丽君 张磊[2] 段如敏 刘晶 朱杰 唐丽 白晓莉 Wu Lijun;Zhang Lei;Duan Rumin;Liu Jing;Zhu Jie;Tang Li;Bai Xiaoli(China tobacco yunnan industrial Co.,Ltd,Kunming 650202,China;Ocean University Of China,Qingdao 266700,China)

机构地区：[1]云南中烟工业有限责任公司,云南昆明650202 [2]中国海洋大学,山东青岛266700

出　　处：《现代科学仪器》2019年第4期75-79,共5页Modern Scientific Instruments

基　　金：云南中烟科技项目“烟叶热水可溶物与烟叶品质相关性研究及应用”(2018JC01)。

摘　　要：烟叶原料的物理形态会对扫描的近红外光谱产生影响,从而使近红外模型预测结果产生差异。以同一烟叶原料烟丝和粉末两类光谱为分析对象,结合光谱变量优化提取和光谱转移的研究成果,基于改进的直接标准化(DS)算法构建烟丝和粉末近红外光谱转移模型,使得烟气指标近红外定量模型的预测性能显著提高。结果表明:光谱转移模型能够有效降低烟叶物理形态对近红外光谱造成的影响,烟叶焦油、CO、烟气烟碱三项指标的平均相对误差分别由7.16%、12.02%和18.69%降低为6.46%、6.62%和9.6%。单位焦油、CO、烟气烟碱用于建模和验证的分布区间为:18.6~27.01mg/g,16.89~24.41mg/g,1.5~3.41mg/g。研究结果为不同形态烟叶的烟气预测一致性提供了保障。The physical form has an influence on the near-infrared(NIR)spectra,so it affects the prediction results of the near-infrared model accordingly.In this study,we develop tobacco cut and tobacco powder models by the improved direct standardization(DS)method via characteristics selection strategy.As a result,the predictive performance of the tobacco smoke indexes is significantly improved.Spectral transfer model can effectively reduce the effect of physical morphology of tar,carbon monoxide and nicotine on tobacco.Specifically,the relative errors of tar,carbon monoxide and nicotine are reduced from 7.16%,12.02% and 18.69%to 6.46%,6.62% and 9.6%,respectively.Moreover,the prediction ranges for tar,carbon monoxide and flue gas nicotine are from 18.6 to 27.01 mg/g,16.89 to 24.41 mg/g,and 1.5 to 3.41 mg/g,respectively.The research provides a technical support for ensuring the consistency of tobacco smoke prediction.

关 键 词：近红外光谱 烟气指标 模型转移 直接标准化算法 

分 类 号：O657.63[理学—分析化学]