基于偏最小二乘和拉曼光谱检测聚氨酯中游离—NCO含量  

Determination of Free—NCO Content in Polyurethane Based on Partial Least Squares and Raman Spectra

作　　者：原长海 梁志岗 姚志湘[1] 粟晖[1] 李敏玲[1] 

机构地区：[1]广西科技大学生物与化学工程学院,广西柳州545006 [2]广西柳州鱼峰制漆股份有限公司,广西柳州545005

出　　处：《涂料工业》2013年第12期61-64,68,共5页Paint & Coatings Industry

基　　金：广西科技计划项目(桂科能129825-20)

摘　　要：基于聚氨酯中游离异氰酸酯不能快速无损检测的现状,利用光纤光谱技术和拉曼光谱仪,根据化学计量学和偏最小二乘原理,对聚氨酯中—NCO的定量分析方法进行研究。在—NCO的拉曼特征峰波段,对聚氨酯样本的拉曼光谱数据和—NCO实测值进行回归分析,建立了聚氨酯中游离—NCO的定量分析模型。选取的主成分数为12时,校验集的相关系数r=0.997 8,交互验证均方根误差(RMSECV)=0.110 7;预测集相关系数r=0.983 6,交互验证均方根误差(RMSECV)=0.264 7,预测集的相对误差<7%。结果表明:所建立的模型具有较好的预测能力,稳健性较强;检测时间从传统方法的几十分钟降低到1 min内,提高了—NCO的检测效率。Because the flee isocyanate in polyurethane couldn＇t be determined lastly and nondestructively, the optical fiber spectroscopy and raman spectrometer, based on the princi- ple of chemical metrology and partial least squares, were used to study the quantitative analy- sis method of INCO in polyurethane. The raman spectra data and measured --NCO values of the polyurethane samples were used for regression analysis to establish a quantitative analysis model of flee -NCO in polyurethane. When the number of principal components is 12, corre- lation coefficient of calibration set r = 0. 997 8, interactive authentication root mean square er- ror （RMSECV） =0. 110 7, and prediction con＇elation coefficient r =0. 983 4, interactive au- thentication rool mean square error （RMSECV） = 0. 264 7, the relative error of prediction set 〈7%. Results showed that the established model provided a good prediction ability and strong robustness. Test time was reduced from dozens of minutes of the traditional method down to 1 minule, and the determination efficiency of--NCO was improved greatly.

关 键 词：聚氨酯 拉曼光谱 偏最小二乘 异氰酸酯(-NCO) 

分 类 号：TQ630.72[化学工程—精细化工]