Detection of degraded high-density polyethylene via near-infrared hyperspectral imaging

The quality of recycled plastics is crucial to make them competitive in more demanding applications and to extend their range of applications. However, there are many influencing factors that can reduce the quality and limit the use of recyclates. One of these factors is degradation, which can occur at different stages of a plastic's life cycle. Degraded material can affect the quality of recyclates. Therefore, it would be beneficial to sort out heavily aged plastics from the recycling stream before further processing. This work investigates the possibility of separating severely degraded polyethylene (PE) from unaged or less degraded PE using near-infrared (NIR) hyperspectral imaging. For this purpose, PE samples were artificially aged using two methods: (i) exposure to UV light, (ii) exposure to aqueous chlorine dioxide solution. The ageing state of the samples was assessed by means of Fourier Transform Infrared (FTIR) spectroscopy and tensile tests and their NIR spectra were recorded on a laboratory NIR sorter. The ability to separate highly degraded from non-/less-degraded samples, which was defined by their mechanical performance, was then analysed using multivariate data analysis and machine learning algorithms applied to the NIR data. These analyses showed promising results for separating highly degraded PE samples, with the classification between degraded and less degraded PE achieving accuracy and F1 scores exceeding 90%.