During melt processing, the moisture inside polylactide (PLA) easily induces hydrolysis, which deteriorates the mechanical and thermal properties of the product. The state of dryness of resin pellets must be monitored to prevent PLA hydrolysis. In this study, near-infrared (NIR) spectroscopy was applied to measure water content in PLA. In addition, the shape of the NIR spectrum is also affected by crystallization, which could lead to a reduction in the accuracy of evaluating the water content. The objective of this research is to construct a robust model for estimating the water content with varying dispersive extents of crystallization. Two methods for estimating water content measured during a drying process were conducted: the integration of absorbance and partial least squares (PLS) regression were conducted to estimate the water contents in PLA considering the effect of crystallization. The slope of the calibration line of the water content obtained from integrating absorbance varied between PLA with different crystallinities. This is due to the overlap between the NIR band of water and that of PLA crystal in the range of 5100-5400 cm-1. We found that the shape of the NIR spectrum was changed by crystallization, and the crystallinity, compared to the thickness of lamellae, was the dominant factor determining such a change of NIR spectra. The PLS model of water content constructed from only amorphous PLA showed large error of estimation in crystallized PLA. In contrast, the PLS model constructed from both amorphous and crystallized PLA estimated the water contents with lower errors. This was because latent variables obtained from both amorphous and crystallized PLA cancelled the effect of crystallization on NIR spectra.
Keywords: NIR; Near-infrared; PLA; PLS; crystallinity; crystallization; hydrolysis; moisture; partial least squares regression; polylactide; spectroscopy.