(National Institute of Standards and Technology) This paper employs machine learning to predict key physical properties of polyolefins from NIR spectra, including density, crystallinity, and short chain branching. By integrating rapid NIR measurement with property prediction, we propose a property-based sorting methodology for polyolefins, addressing the current challenges recyclers face in differentiating polyolefin subclasses. This approach also enhances our understanding in spectra-property relationships. Our work represents a significant advancement toward enhancing sorting efficiency and promoting circularity in plastic recycling.
Link to Dataset
Machine learning-enhanced near-infrared property prediction for improved polyolefin differentiation
Date last updated: July 24, 2025