Effects of surface contamination on automated textile sorting using NIR-spectroscopy.

Area of Science:

• Materials Science
• Analytical Chemistry
• Environmental Science

Background:

• EU member states must implement separate textile collection by January 2025.
• Contamination in post-consumer textiles poses a significant challenge for automated sorting technologies.
• Near-infrared (NIR) spectroscopy shows promise for automated textile identification but is affected by material contamination.

Purpose of the Study:

• To investigate the impact of contamination, particularly moisture, on NIR spectroscopic identification of textiles.
• To analyze how contamination affects classification accuracy for cotton-rich, polyester-rich, and polycotton textiles.
• To identify spectral differences caused by contaminants and assess their influence on sorting models.

Main Methods:

• Analysis of residual waste textile samples with varying contamination levels.
• NIR spectroscopy to measure spectral properties of contaminated textiles.
• Development and evaluation of classification models for cotton-rich, polyester-rich, and polycotton textiles.
• Investigation of spectral differences and contaminant identification.

Main Results:

• Moisture significantly influences textile spectral behavior and NIR classification model performance.
• Excluding moisture-sensitive wavelengths (1368-1459 nm) improved polyester-rich sample accuracy but complicated polycotton classification.
• Average classification accuracy decreased from 83.1% to 78.9% after excluding moisture-sensitive wavelengths.
• Other contaminants had a lesser impact, but heavy surface contamination may necessitate lowering classification thresholds.

Conclusions:

• Excluding moisture-sensitive wavelengths can enhance classification for specific textile types (e.g., polyester-rich) but risks misclassification for others (e.g., polycotton, cotton-rich).
• Tailoring NIR classification models to target textile fractions is essential for optimizing sorting accuracy.
• Adjusting classification thresholds may be necessary for heavily contaminated textiles to mitigate misclassification.
• This research supports improved NIR classification of contaminated textiles, aiding the circular economy by enabling resource recovery.