AI-Assisted Characterization of Cotton/Polyester Textiles using IR Spectroscopy

Traditional textile separation and recycling methods are labor intensive, prone to inaccuracies and inefficient. A fast and environmentally friendly method for the simultaneous identification of multiple textile components is required for recycling, which is complex due to the presence of dyes and contaminants. The adoption of spectroscopic techniques for textile identification offers a pathway to more sustainable and efficient recycling processes [1,2]. The Josef Ressel Center ReSTex - aligned with the European Green Deal and Road to 2030 objectives - is dedicated to textile recycling. One objective is to characterize cotton-polyester blended textiles containing varying proportions of these fibers using Near-Infrared (NIR) and Mid-Infrared (MIR) spectroscopy for more efficient textile recycling process of different fibres and fabrics.
The methodology involves the establishment of robust spectroscopic analysis. Various sampling techniques and instrument parameters are systematically investigated to optimize the quality of the spectral data. Measurement of various cotton-polyester blends are carried out using FTIR-ATR for MIR and diffuse reflection for NIR measurements and the application. Various statistical and machine learning approaches, including Principal Component Analysis (PCA) and Partial Least Squares (PLS), are used to characterize these blends are tested.
This study successfully established MIR and NIR methods for accurately determining the composition of cotton-polyester blends in various proportions. It highlights the importance of selecting appropriate data analysis techniques for accurate evaluation and discusses the use of statistical and machine learning methodologies for data interpretation. The project's outlook involves expanding the dataset for model validation, refining the predictive model for cotton-polyester content, and comparing the results obtained from NIR and MIR spectroscopy with Raman spectroscopy results.