Hotelling T^2结合多组分NIR校正模型在卷烟生产过程质量监测中的应用  

Quality Monitoring in Cigarette Manufacturing Process by Combining Hotelling T^2 Statistics and Multicomponent NIR Calibration Model

作　　者：李伟 冯洪涛[2] 周桂圆[2] 王家俊[2] 李滟芳[2] 钱启福[3] 杨盼盼 杨勇 

机构地区：[1]云南同创检测技术股份有限公司 [2]红云红河烟草(集团)有限责任公司技术中心 [3]红云红河烟草(集团)有限责任公司昆明卷烟厂

出　　处：《烟草科技》2014年第7期5-9,共5页Tobacco Science & Technology

基　　金：红云红河烟草(集团)有限责任公司科技项目"FT-NIR结合Hotelling T2在卷烟生产质量监测中的应用"(HYHH2012GY06)"卷烟品质特征的多维空间化学指纹图谱表征研究"(HYHH2012HX07)

摘　　要：为有效利用近红外技术结合化学计量学方法监测卷烟生产配方模块质量,通过建立卷烟配方模块主成分类模型,提取Hotelling T2统计量对配方调整前后的样品进行验证,其中配方调整前150个样品的Hotelling T2统计量分布于95%(Hotelling T2=22.70)置信范围内,配方调整后50个样品的Hotelling T2统计量均超过99%(Hotelling T2=28.16)置信范围;同时采用所建立的校正模型预测配方调整前后样品的烟碱、总氮和总糖含量,预测结果显示,150个配方调整前样品的烟碱、总氮和总糖含量处于M±3S质量控制范围内,50个配方调整后样品的烟碱、总氮和总糖含量均超出M±3S质量控制范围。该方法既可监测配方模块整体质量的稳定性和均匀性,也可预测配方模块化学成分含量的变化。To effectively monitor the quality of blend modules during cigarette manufacturing via near infrared spectroscopy combined with chemometrics, Hotelling ff statistics were derived from the established categorized models for principal components of blend modules and used to monitor the samples before and after blend modifying. The Hotelling T~ statistics of 150 samples before b!end modification were of 95% confidence interval （Hotelling T2=22.70）, the 50 samples of modified blends were of 99% confidence interval （Hotelling T2=28.16）. The contents of nicotine, total nitrogen and total sugar in samples before and after blend modification were predicted with the established calibration model, the results indicated that the contents of nicotine, total nitrogen and total sugar in the 150 samples of unmodified blends were within the quality control limits of M±3S, while those in the 50 samples of modified blends all exceeded the limits of M±3S. The method is capable of monitoring the quality consistency and uniformity of blend modules and predicting the variation of chemical component contents in blend modules.

关 键 词：近红外光谱 HOTELLING T2 多组分近红外校正模型 卷烟生产 过程质量 

分 类 号：TS411[农业科学—烟草工业]