Quantification of LDPE [Low Density Poly(Ethylene)], Lldpe [Linear Low Density Poly(Ethylene)], and Hdpe [High Density Poly(Ethylene)] In Polymer Film Mixtures “As Received” Using Multivariate Modeling with Data Augmentation (Data Fusion) and Infrared, Raman, and Near-Infrared Spectroscopy

Abstract

This work was completed to develop a routine method for rapid or at-line quantitative measurement of the blend composition of poly blends using NIR, IR, and Raman spectroscopy. The initial results indicate that spectroscopy can be used for predicting percent LDPE [low density poly(ethylene)], LLDPE [linear low density poly(ethylene)], and HDPE [high density poly(ethyiene)] with accuracy values ranging from 1-5% absolute. The best results have been obtained by using mean-centered Raman spectra. Successful quantitative analysis was also achieved by combining NIR and Raman spectra using data augmentation (i.e., data fusion), autoscaling, and a variety of regression methods (chemometrics). Initial accuracy (as SEE, standard error of estimate for the regression) was 0.53-1.4% absolute using Raman spectral data, and 1.5-4.5% absolute for NIR/Raman augmented data. Infrared was found to have the least useful information content as compared to both NIR and Raman due to in situ sample thickness.