Research And Application On Rapid Identification Of Vegetable Oil Types Using LF-NMR Technology

As an important component of human diet,vegetable oil not only improve the taste of food,but also provide nutrients essential such as fatty acids and fat-soluble vitamins for the human body.However,the yield and nutritional values of different species of vegetable oils are different from each other,and the price is also quite different.To get more profits,the phenomenon of shoddy and fake vegetable oils occurs frequently in the market,which harms the health of consumers and affects the market order.Therefore,it is particularly important to develop a method for the rapid identification of vegetable oil types.In this study,the low-field nuclear magnetic resonance（LF-NMR）technology combined with chemometrics was used to identify the species of vegetable oil.The theoretical basis of the identification of vegetable oil species based on LF-NMR technology was systematically studied in detail.The physicochemical properties,glyceride contents and fatty acid compositions of seven different types of vegetable oils were analyzed.The differences in LF-NMR signals of seven vegetable oils were compared.The relationship between the contents of fatty acids and glycerides and the relaxation properties of oil was explored.In addition,the effects of sample temperature and color on LF-NMR signal were also studied.The results showed that there was no significant differences in the physicochemical properties of rapeseed oil,soybean oil,peanut oil,sunflower oil,corn oil and sesame oil.However,the iodine value,melting point and viscosity of palm oil were significantly different from those of the other six vegetable oils.There were no significant differences in the contents of monoglyceride,diglyceride and triglyceride among the seven vegetable oils,while significant differences existed in the contents of palmitic acid,oleic acid,linoleic acid and polyunsaturated fatty acid.To some extent,the CPMG echo attenuation curves and transverse relaxation graphs of LF-NMR of different vegetable oils were different.The order of CPMG echo attenuation rate of seven vegetable oils was Palm oil>Peanut oil>Sesame oil>Rapeseed oil>Corn oil>Soybean oil>Sunflower oil,and the relaxation properties T₂₃,S₂₃,S_total,R₂₂,R₂₃ were significantly different（P<0.05）.The change of fatty acid and glyceride contents in the oil had a significant effect on the relaxation properties T_2W,T₂₂,T₂₃,S₂₃ and S_total（P<0.05）.The attenuation rate of CPMG echo,transverse relaxation spectra and relaxation properties of LF-NMR changed with the temperature of the sample,while the difference of color in sample had no significant effect on the LF-NMR signal.The CPMG echo attenuation curve of LF-NMR combined with chemometric method was used to identify the species of vegetable oils.The effects of modeling data and preprocessing methods on the identification performance of vegetable oil species by LF-NMR CPMG echo attenuation curve were analyzed.The CPMG echo attenuation curve of LF-NMR was combined with principal component analysis（PCA）,hierarchical cluster analysis（HCA）,linear discriminant analysis（LDA）and support vector machine（SVM）to establish the identification model of vegetable oil species,and also tried to the identification of mixed vegetable oil.The results showed that different combinations of modeling data and preprocessing had different effects on the modeling performance based on the CPMG echo attenuation curve.The best combination of modeling data and preprocessing method was the 1-1000 ms data segment without preprocessing.Compared with other modeling methods,the identification model of vegetable oil species established by SVM can get the best performance with the prediction accuracy of 81%in 21 blind samples.The prediction accuracies of this method for binary and ternary mixed vegetable oil blind samples were 93.8%and 100%,respectively.The LF-NMR transverse relaxation curve combined with chemometric method was used to identify the species of vegetable oil.The effects of modeling data and preprocessing methods on the identification of vegetable oil species by LF-NMR transverse relaxation curve were analyzed.The transverse relaxation curve of LF-NMR was combined with PCA,HCA,LDA and SVM to establish the identification model of vegetable oil species,and applied to the identification of mixed vegetable oil.The results showed that the selection of transverse relaxation curve modeling data and preprocessing methods can directly affect the performance of the identification model.The best combination was 1-3000 ms data segment without data preprocessing.Compared with other modeling methods,SVM model had the best identification performances with the prediction accuracy of 66.7%in 21 blind samples.The prediction accuracies of this model for binary and ternary mixed vegetable oil blind samples were 92.2%and 100%,respectively.The relaxation properties of LF-NMR combined with chemometric method were used to identify the species of vegetable oil.The effects of the number of relaxation properties variables on the identification of vegetable oil species by LF-NMR technology were investigated.The LF-NMR relaxation properties were combined with PCA,HCA,LDA and SVM to establish the identification model and applied to the identification of mixed vegetable oil.The results showed that the number of relaxation properties variables had a certain influence on the performance of LDA and SVM models.The optimal number of variables based on the LF-NMR relaxation properties was 11.Compared with other modeling methods,the model established by SVM had the best identification performance with the accuracy of 57.1%in 21 blind samples.The prediction accuracies of this model for binary and ternary mixed vegetable oil blind samples were 95.3%and 97.5%,respectively.In conclusion,the best modeling method for vegetable oil specie identification based on LF-NMR was SVM,and the performance of the model established by CPMG echo attenuation curve combined with SVM method was the best.