Abstract

A highly expeditious method for the direct analytical control of leather fat as an alternative to the officially endorsed methods was developed based on solvent extraction. The proposed method uses multivariate calibration with partial least-squares regression (PLSR) of diffuse reflectance spectra (NIRRS) for leather specimens. The reference values of the calibration matrix are used to derive those for powdered leather extracted with methylene chloride using the official method. Two different approaches based on matrices containing spectral information for powdered leather and spectra directly recorded on unpowdered leather samples were developed.

The optimum calibration model for powdered samples is based on the second spectral derivative and 3 PLS components, and provides overall relative calibration and prediction errors in the region of 5%. On the other hand, the optimum prediction model uses the first spectral derivative and 5 PLS components; the results are similarly as good as those of the calibration model, but the prediction of some external samples results in larger errors arising from the inability to construct calibration sets encompassing the whole variability of the prediction set.

Fibre optic-based NIR spectroscopy in combination with multivariate calibration appears to be an effective alternative to existing methods for the analysis and control of fat in the leather tanning industry as it is a rapid method and uses no reagents.