Material-based characterization of multilayer plastic film and textile wastes using NIR spectroscopy

The transition toward a circular economy requires reliable identification and separation of ma-terials in post-consumer waste streams. Near-infrared (NIR) spectroscopy is already widely applied in automated sorting, but its performance is generally optimized for well-defined mate-rial fractions. More complex waste streams, such as multilayer plastic films and textiles, re-main difficult to classify due to low thickness, layered or blended structures, and susceptibility to contamination. This thesis focuses on fundamental research to understand how NIR spec-troscopy can be adapted and applied to improve material identification and feedstock quality in these challenging waste fractions. For multilayer plastic packaging films, handheld NIR spectroscopy was shown to accurately differentiate the most common multilayer combinations using only a single-sided measure-ment. The measurement setup, particularly the reflectance properties of the background, was found to strongly influence spectral quality and classification stability. These findings support the use of handheld NIR devices for representative sampling and supplier-related quality con-trol, complementing automated sorting systems. For textiles, NIR spectroscopy was evaluated for quantifying cotton content in polyester/cotton blends under both laboratory conditions and handheld measurements. Accuracy was affected by blend ratio, textile structure, and surface treatments. Moisture contamination was identified as the most influential factor reducing classification reliability. The results highlight the need to account for contamination effects in model development and to build broader, more repre-sentative datasets, while future work should evaluate performance under larger-scale sorting conditions. Overall, the thesis establishes the conditions under which NIR spectroscopy can reliably clas-sify multilayer films and textiles and clarifies how handheld and automated systems can be used in complementary roles within sorting workflows. The findings provide fundamental knowledge that supports further research and industrial implementation.