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。

关键词: 聚氨酯, 拉曼光谱, 偏最小二乘, 异氰酸酯(—NCO)

Abstract: Because the free isocyanate in polyurethane couldn't be determined fastly and nondestructively, the optical fiber spectroscopy and raman spectrometer based on the principle of chemical metrology and partial least squares, were used to study the quantitative analysis method of -NCO 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 free-NCO in polyurethane. When the number of principal components is 12, correlation coefficient of calibration set r=0. 9978, interactive authentication root mean square error(RMSECV)=0. 1107, and prediction correlation coefficient r=0. 9834, interactive authentication root mean square error(RMSECV)=0. 2647, 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 minute, and the determination efficiency of -NCO was improved greatly.

Key words: polyurethane, raman spectrum, partial least squares, isocyanate(-NCO)