The increasing demand for a sustainable plastics industry necessitates a more efficient utilization of post-consumer recycled plastics. In particular, the plastics processing industry is under pressure to incorporate higher proportions of recycled materials due to regulatory requirements such as the End-of-Life Vehicle Regulation (ELV). However, significant variations in the processing properties of recyclates hinder their reliable use. This study aims to enhance the predictability of recyclate quality and processability by combining artificial intelligence (AI) and material characterization to optimize the sorting process of plastic waste. Material detection was carried out using near-infrared spectroscopy and optical sensors. Further, the plastic waste was classified into three categories, based on the applied processing technology, being an injection molded, extruded or thermoformed product. This allows a correlation with the molecular weight and the molecular weight distribution of the plastic and thus the expected viscosity. Based on that a classification of plastics into different quality grades is introduced. A computer vision model is developed to capture and process the optical characteristics relevant for differentiation, enabling both material classification and a data-driven quality assessment of recyclates. Small and medium-sized enterprises (SMEs) in the plastics processing sector particularly benefit from improved recyclate reliability and a more stable manufacturing process.
Werner et al. (Thu,) studied this question.