Research on infrared spectroscopy analysis method of cotton-polyester blended fabrics based on blind source...

In the qualitative analysis of complex components using mid-infrared spectroscopy (MIR), it is often difficult to directly resolve pure component spectra due to peak overlap, baseline drift, and nonlinear interactions between components. Therefore, spectral decomposition via blind source separation (BSS) to extract pure spectra of individual components, followed by comparative analysis with measured or quantum chemically calculated spectra of pure components, is an excellent and crucial choice for achieving accurate identification. In this study, MIR ATR spectra of cotton-polyester blends and their pure constituents were obtained, and three BSS methods including natural gradient-based BSS (NGBSS), independent low-rank matrix analysis (ILRMA), and non-negative matrix factorization (NMF) were applied for spectral decomposition. The molecular models of cotton and polyester were constructed, and their theoretical vibrational spectra were simulated. Both decomposed and simulated spectra were evaluated against experimental pure component spectra using peak-based cosine similarity (PBCS).

Results demonstrate that NMF outperformed other BSS methods in spectral extraction. Specifically, PBCS between the NMF-separated spectra and the measured reference spectra reached 0.953 for cotton and 0.976 for polyester. In contrast, the PBCS values between the simulated vibrational spectra and the measured fiber spectra were lower, at 0.869 (cotton) and 0.923 (polyester). Furthermore, the NMF-separated spectra showed a strong correlation with the simulated pure vibrational spectra, yielding PBCS values of 0.784 for cotton and 0.797 for polyester.

This study establishes quantum chemical simulation as a robust framework for validating BSS decomposition. It offers a sophisticated evaluation paradigm for complex mixture analysis, particularly when reference standards are unavailable. Data and algorithms are available at https://gitee.com/LILILLLI/blc.