Monitoring of free fatty acid content in mixed frying oils by means of LF-NMR and NIR combined with BP-ANN

Abstract

The free fatty acid (FFA) is an essential indicator to determine the discard point of frying oils, however, the current detection method of FFA in oil is laborious. This research established two non-destructive approaches based on low field nuclear magnetic resonance (LF-NMR), near Infrared (NIR) spectra, and back-propagation artificial neural network (BP-ANN) algorithm for monitoring the FFA content of fried oil samples. 105 used frying oils, representing various frying degree, were detected using LF-NMR, NIR and reference method. HCA and PCA were used for natural clustering of LF-NMR parameters (S21, S22, S23, T21, T22, and T23) and NIR spectroscopy. Finally, the value of the correlation coefficient (R 2 ) manifested that the accuracy of LF-NMR model and NIR model reached 0.850, 0.963, respectively. The R 2 value of NIR model was 0.113 higher than that of LF-NMR model, indicating NIR spectroscopy of used frying oil could be a more accurate method for monitoring the FFA content in the oil using the BP-ANN model. • The changes of FFA of frying oil with the increase of frying batch were investigated. • HCA and PCA were used for natural clustering of LF-NMR and NIR parameters. • Both LF-NMR and NIR were applied for monitoring FFA values using the BP-ANN model. • NIR-FFA model was more accurate than LF-NMR-FFA model.