Near-infrared reflectance spectroscopy was used in its absorbance and first-derivative modes in conjunction with two multivariate calibration procedures, viz., stepwise multiple linear regression (SMLR) and partial least-squares linear regression (PLSR), for the determination of cotton in cotton-polyester yarns. The results provided by the two procedures were similar, with prediction errors of less than 1%. The PLSR procedure proved to be more robust and gave slightly better results over the wavelength range 1400-2300 nm. Analysis of the samples requires virtually no pre-treatment and takes very little time (about 15 s), which makes the proposed method highly suitable for quality control analyses.
|Publication status||Published - 1 Dec 1994|