基于在线近红外光谱分析技术对七种常规烟丝化学成分的实时检测

【作者】 朱红波；

【导师】 彭黔荣； 杨敏；

【作者基本信息】 贵州大学 ， 食品科学， 2009， 硕士

【摘要】 目前,近红外（NIR）分析技术在烟草研究领域已得到了较广泛的应用,但是,这些研究采用的近红外光谱仪主要是以离线为主。本研究应用在线傅立叶变换近红外光谱仪收集制丝线上的烟丝光谱,采用偏最小二乘法（PLS）建立了在线烟丝含水率、总植物碱、总糖、总氮、还原糖、氯、钾的预测模型。主要的研究内容有:1.通过实验对光谱仪器的分辨率和扫描次数两个仪器参数进行了选择和优化,最后得到仪器的分辨率参数为:8cm^-1,扫描次数参数为:64次。采用主成分-马氏距离法从115个样品中剔除了7个光谱异常样品,采用化学值F检验法分别剔除了含水率、总植物碱、总糖、总氮、还原糖、氯、钾化学值异常样品数分别为:2、7、2、2、5、5、4。2.通过OPUS软件优化,选择了最佳的谱区范围、最佳的光谱预处理方法以及最佳的主成分维数,建立了7个化学指标的数学模型,TS、RS、NIC、TN、H₂O、CL、K模型的内部均方差（RMSECV）分别为:0.5450、0.4520、0.0499、0.0505、0.0841、0.0270、0.0786。采用外部验证法验证模型的可靠性,用建好的模型预测8个样品,得到近红外方法与常规方法的比较结果。TS、RS、TN、NIC、K、CL、H₂O含量预测的最大绝对误差的绝对值分别为:0.42、0.67、0.08、0.10、0.14、0.07、0.15;标准偏差分别为:0.2127、0.4276、0.0266、0.0603、0.0924、0.0385、0.1936;平均相对误差分别为:0.64%、1.59%、2.49%、1.9%、4.19%、6.58%、0.97%;预测均方差（RMSEP）分别为:0.1994、0.4027、0.0595、0.0581、0.0923、0.0395、0.1823。离线近红外模型已经满足分析检测的要求,通过在线模型和离线模型的比较,在线模型的各个参数都达到了预测精度要求,能够用于生产线的监测。3.用WinCC V6.0控制开发软件,设计了系统的登陆界面,生产数据显示界面以及烟丝化学含量变化趋势界面,且用WinCC组态软件的Web Navigator选件将这些界面面发布到网上,各个客户机通过IE浏览器就可以接收定量分析数据,了解烟丝生产线的生产情况,可以及时、直观地了解到当前卷烟生产成品烟丝的化学值变化状况。实验结果表明,在线近红外光谱分析法结果与常规分析法结果比较,具有很好的相关性,预测标准偏差小。经实际生产实验,近红外光谱法能应用于烟厂制丝线的生产分析,且经济效益显著,在烟丝常规质量指标的监测中有着广泛的应用前景和推广价值。更多还原

【Abstract】 At present,the near-infrared(NIR) analysis techniques in the field of tobacco research has been more widely used,but these studies used near infrared spectroscopy is mainly offline.In this study,the prediction models of Total sugar content,reducing sugar content,total nitrogen content,nicotine content,potassium content,chloride content,moisture content in cut tobacco during primary processing were established with its spectra acquired by on-line FT-NIR and the partial least two Multiplication(PLS).The main contents as follows:1.The spectral resolution and scanning frequency of the instruments is optimized through the experimental,the final parameters to be instruments for the resolution:8cm-1,the number of parameters for the scan:64 times.Using principal components-Mahalanobis distance from sample 115 to remove the abnormal spectra of seven samples,the value of F test chemical excluded moisture content,respectively,with a total alkali plant,the total sugar,total nitrogen,reducing sugar,chlorine, potassium chemical Samples number of abnormal values are:2,7,2,2,5,5,4.2.The seven mathematical models of TS,RS,NIC,TN,H2O,CL,K are optimized by selecting the best area of the spectrum,the best spectral pre-processing methods and the best principal component dimension through the OPUS software,its RMSECV are as follows: 0.5450,0.4520,0.0499,0.0505,0.0841,0.0270,0.0786.Verifying by external validation the reliability of the model,the prediction model built eight samples,getting the near-infrared method result comparing with routine method.TS,RS,TN,NIC,K,CL,H2O content in the prediction of the absolute value of the maximum absolute error are as follows:0.42,0.67,0.08,0.10,0.14,0.07,0.15;the standard deviation are asfollows:0.2127,0.4276,0.0266,0.0603,0.0924,0.0385,0.1936;the average relative error are as follows:0.64%,1.59%,2.49%,1.9%,4.19%,6.58%,0.97%;prediction mean square error(RMSEP) are as follows:0.1994,0.4027,0.0595,0.0581,0.0923,0.0395,0.1823.Comparing the online models with the offline models,the parameters of the online models achieve the prediction accuracy,so it can be used to monitor the production line.3.Using WinCC V6.0 with control software to design of the landing system interface,data interfaces,as well as the production of the chemical content of tobacco trend of the interface,using configuration software with WinCC options Web Navigator interface to noodle posted to the Web, various client through the IE browser will be able to receive data quantitative analysis to understand the tobacco production line can be timely and intuitive to learn that the current cigarette tobacco products of chemical production value changes.Experiment results show that the pertinence is good between on-line NIR spectroscopy and the chemical analysis and the standard deviation of predict is small.so we think the NIR spectroscopy can be used for analysis in in the production of silk thread tobacco factory system analysis,which has great economic benefit,extensive prospect and value of popularization in the quality of conventional tobacco monitoring.更多还原