Abstract

A new method for the direct quantification of tile fat content of leather based on near infrared diffuse reflectance spectrometry (NIRRS) and multivariate calibration was developed. Wet-blue cow hide samples (15 for calibration and 10 for prediction) containing 0- 13% fatty matter were prepared for this purpose. Spectra were recorded by using fibre optics and rotary cuvettes, and the spectral information acquired was processed by stepwise multiple linear regression (SMLR) and partial least-squares regression (PLSR). Results were correlated with the initial amount of fat in the fat-liquoring bath, which was depleted by 96-99%. The relative standard error of prediction for the calibration set (RSEPC) ranged from 6 to 8%. and for the prediction external set, yielded a relative standard error (RSEPP) of about 8%.