Post-Consumer Textile Waste Classification through Near-Infrared Spectroscopy, Using an Advanced Deep Learning...

Area of Science:

• Textile Engineering
• Materials Science
• Artificial Intelligence

Background:

• The linear 'take-make-dispose' model of the textile industry causes significant environmental issues.
• Fast fashion accelerates consumption and waste, necessitating a shift towards a circular economy.
• Effective end-of-life management, including collection and classification, is crucial for textile recycling.

Purpose of the Study:

• To develop an automated method for classifying post-consumer textile waste.
• To enable the separation of high-quality textile fiber streams for reuse.
• To support the transition to a sustainable, circular textile economy.

Main Methods:

• Utilized near-infrared (NIR) spectroscopy to analyze textile samples.
• Applied convolutional neural networks (CNNs) for spectral data processing and classification.
• Trained and tested the model on 370 textile samples, including pure fibers and binary mixtures.

Main Results:

• Achieved 100% accurate classification for pure textile fibers.
• Demonstrated 90-100% accuracy for classifying binary mixtures of common textile fibers.
• Validated the proposed methodology's power and trainability for industrial-scale automation.

Conclusions:

• The NIR spectroscopy and CNN-based method offers a powerful solution for automated textile waste classification.
• This approach facilitates the separation of valuable fiber streams, promoting high-value recycling.
• The technology is compatible with industrial automation, paving the way for a more sustainable textile industry.