Watari Masahiro, Ozaki Yukihiro
Environmental and Analytical Products Business Division, Yokogawa Electric Corporation, Musashino 180-8750, Japan.
Appl Spectrosc. 2005 Jul;59(7):912-9. doi: 10.1366/0003702054411571.
The present study investigates calibration models for the vinyl acetate (VA) concentration in ethylene-vinyl acetate (EVA) copolymers and its on-line monitoring by near-infrared (NIR) spectroscopy and chemometrics. The key point in the present study is to make use of band shifts associated with concentration changes in the vinyl acetate (VA) for the improvement of the models. NIR spectra of EVA in melt and solid states were measured by a Fourier transform near-infrared (FT-NIR) on-line monitoring system and a FT-NIR laboratory system. Some of the bands in the NIR spectra for both states show significant shifts with the variations in the VA concentration. The peak shifts induced by the VA concentration changes are larger in the solid-state EVA than those in the melt-state EVA. We have developed calibration models for the VA concentration in the solid-state EVA and investigated how to improve the calibration models. The factor analysis of partial least squares (PLS) regression has suggested that the wavenumber shifts caused by the VA concentration changes affect the calibration models for the VA concentration in EVA. From the analysis, it has been proposed that the wavenumbers in the spectrum of one sample in nine EVA samples (VA concentration range: 0-41.1%) are shifted for the improvement of the calibration models, and the effects of the proposed method have been confirmed by using the PLS calibration models for the VA concentration in the solid EVA samples. As the next step, the effects of the wavenumber shift method have been explored for the calibration models for the VA concentration in the melt-state EVA. After that, the discrimination method using the score plots of PLS and the application sequence for the on-line monitoring to use the proposed wavenumber shift method were studied. The simulation results using the discrimination and wavenumber shift methods have shown that those methods are very effective to improve the predicted values of the calibration models for the on-line monitoring of the VA concentration in the melt-state EVA.
本研究探讨了乙烯-醋酸乙烯酯（EVA）共聚物中醋酸乙烯酯（VA）浓度的校准模型，以及通过近红外（NIR）光谱和化学计量学对其进行在线监测。本研究的关键点在于利用与醋酸乙烯酯（VA）浓度变化相关的谱带位移来改进模型。通过傅里叶变换近红外（FT-NIR）在线监测系统和FT-NIR实验室系统测量了EVA在熔融态和固态下的近红外光谱。两种状态下近红外光谱中的一些谱带随VA浓度的变化呈现出显著位移。VA浓度变化引起的峰位移在固态EVA中比在熔融态EVA中更大。我们开发了固态EVA中VA浓度的校准模型，并研究了如何改进校准模型。偏最小二乘（PLS）回归的因子分析表明，VA浓度变化引起的波数位移会影响EVA中VA浓度的校准模型。通过分析，提出对九个EVA样品（VA浓度范围：0-41.1%）中一个样品的光谱波数进行位移以改进校准模型，并通过使用固态EVA样品中VA浓度的PLS校准模型证实了该方法的效果。作为下一步，研究了波数位移方法对熔融态EVA中VA浓度校准模型的影响。之后，研究了使用PLS得分图的判别方法以及应用所提出的波数位移方法进行在线监测的应用顺序。使用判别方法和波数位移方法的模拟结果表明，这些方法对于改进熔融态EVA中VA浓度在线监测校准模型的预测值非常有效。
