ABSTRACT
A rapid method based on three-dimensional synchronous fluorescence spectroscopy was developed for emulsion oxidation evaluation. This method was selected because of its high sensitivity to dissolved organic matter typically occurring in the lipid oxidation. Spectral signal and chemical reference measurements were recorded for each emulsion sample as input and output data for the model construction. Characteristic values were extracted from the spectral data by the application of parallel factor (PARAFAC) analysis. Partial least squares regression (PLSR) was then used to construct a regression model for the rapid determination of emulsion oxidation. The correlation coefficient of the calibration and prediction sets were used as the performance parameters for the PLSR models as follows: R = 0.929, 0.973 for emulsion samples stored at 25â; R = 0.897, 0.903 for emulsion samples stored at 70â. The overall results demonstrated that the fluorescence spectroscopy, coupled with PARAFAC and PLSR algorithms, could be successfully used as a rapid method for the emulsion oxidation evaluation.
Subject(s)
Spectroscopy, Near-Infrared , Calibration , Emulsions , Feasibility Studies , Least-Squares Analysis , Spectrometry, FluorescenceABSTRACT
The purpose of this study was to construct a fusion model using probe-based and non-probe-based fluorescence spectroscopy and low-field nuclear magnetic resonance spectroscopy (Low-field NMR) for rapid quality evaluation of frying oil. Iron tetraphenylporphyrin (FeTPP) was selected as the probe to detect polar compounds in frying oil samples. Non-probe-based fluorescence spectroscopy and low-field NMR were employed to determine the fluorescence changes of antioxidants, triglycerides and fatty acids in frying oil samples. Compared to the models constructed using non-fusion data, the fusion-data models achieved a better regression prediction performance and correlation coefficients with values of 0.9837 and 0.9823 for the training and test sets, respectively. This study suggested that the multiple data fusion method was capable to construct better regression models to rapidly evaluate the quality of frying oil and other food with high oil contents.
