Abstract:Rapid Raman quantitative analysis models were constructed for four fatty acids (oleic acid, linoleic acid, stearic acid, and palmitic acid) from edible vegetable oil. The regions corresponding to the Raman active groups in the molecular structure of the fatty acids were selected as multiple characteristic regions in the measured spectra, and 46 edible vegetable oil samples were collected as experimental materials to construct the model for these four fatty acids via the partial least squares method. The experimental results indicated that compared with the model built according to the full spectrum, use of the characteristic spectral region reduced the number of variables required for effective modeling and significantly improved the prediction performance of the models. In the experiment, variation in the degree of fluorescence interference was caused by the different colors of edible oils. Therefore, the selected characteristic spectral regions were combined with the spectral bands that showed significant interference (ranging from 295 to 325 cm-1) for joint modeling, thus significantly enhancing the performance of the Raman quantitative models. Furthermore, a mathematical-statistical method was employed to optimize the characteristic spectral regions for the construction of near infrared (NIR) models, which were compared with the established Raman models. The results indicated that the two kinds of models had similar performance and both can be applied for the rapid and accurate detection of the fatty acids in edible oils. However, the Raman-based model has more advantages due to the smaller sample size required and the explicit mechanism.
