34139

Determination of diethylene glycol content in fibers of polyethylene terephthalate using near infrared hypespectral images (HI-NIR)

Favorite this paper

Polyester fibers, in particular, polyethylene terephthalate (PET), contribute greatly to the synthetic fiber market worldwide. In the PET fiber manufacturing process, the diethylene glycol (DEG) content is an important parameter that must be considered with caution because it influences the physical and chemical properties of the PET fibers, even in small quantities. The presence of DEG in the fibers can cause a decrease in polymer crystallinity, which leads to a decrease in its melting point, heat resistance, oxidative thermal stability and UV stability as well as influencing the dyeing of the polymer. Of all the negative effects that the presence of DEG in the polymer causes, the worst is related to the dyeing. The points with higher concentration of DEG will have a stronger staining as compared to the regions with lower concentration. This is due to the fact that the presence of DEG decreases the crystallinity of the polymer, thus allowing greater absorption of the dye. In this work, the images of the PET fibers were acquired using the SisuCHEMA SWIR camera (SPECIM Spectral Imaging Ltd.), in the range of 928-2,338 nm, at a spectral resolution of 10 nm and pixel size of 150x150 µm. Partial Least Squares (PLS) regression models were constructed with the average spectra of the near infrared hyperspectral images (HI-NIR), using 96 samples for calibration and 39 samples for external validation. Different preprocessing techniques were evaluated: Standard Normal Variation (SNV), Savitzky-Golay smoothing and derivative (and combinations of them). The best model was built using 4 latent variables with the spectra of the HI-NIR preprocessed with SNV and Savitzky-Golay derivation (7 points window), resulting in a root mean square error of Prediction (RMSEP) of 0.08 % (w/w) and R2 of 0.91. The RMSEP value obtained by this model is equivalent to the repeatability of the conventional method (chromatography), which is 0.10 % (w/w). In order to map the DEG content in the images of the fibers, samples from the external validation set were used. For each sample, an area of 251 x 61 pixels was selected and distribution maps of the concentration of DEG were obtained using the PLS model. The model was able to predict the points with higher concentrations of DEG, which consequently would have a more pronounced color when the fiber is dyed. These distribution maps are useful for process optimization in order to avoid high concentrated areas in the fibers.