JOURNAL OF THE AMERICAN OIL CHEMISTS SOCIETY, vol.78, no.5, pp.495-502, 2001 (SCI-Expanded)
Use of near-infrared (NIR) transmittance spectroscopy for rapid determination of the oxidation level in soybean oils (SBO) was investigated, and calibrations were developed for quantitative determination of peroxide value (PV), conjugated diene value (CD), and anisidine value (AV) of SBO. Partial least squares (PLS) regression and forward stepwise multiple linear regression were used to develop calibration models from spectral data in log 1/T, first derivative and second derivative of log 1/T formats for both 1- and 2-mm path lengths. The models were validated by comparing NIR results from independent sample sets to the values; obtained by official methods. The spectral region from 1100 to 2200 nm was best for measuring oxidation when using a 2-mm path length. PLS regression using first-derivative spectra gave the best results for PV. For the validation sets, linear relationships were obtained for PV (r = 0.99), and CD (r = 0.95), compared with accepted reference procedures. However, measurement of AV by NIR was less successful than measurement of the other two indices of oxidation, especially for an external validation sample set. Results obtained in this study indicate that NIR spectroscopy is a useful technique for measuring oxidation in soybean oil.