Synchronous Detection Of Physicochemical Indicators Of Vegetable Oils In Frying Process Based On LF-NMR

As one of the Chinese traditional foods,the quality and safety of deep-fried food is closely associated with the quality of fat and oil.The frying process of fats and oils undergoes complex chemical reactions with the internal components of food,which not only degrade the nutritional quality of fats and oils,but also produce substances that are harmful to the human body.Therefore,it is essential to develop a rapid and scientific method for measuring the physicochemical index of frying fats and oils.In this study,the theoretical basis for the detection of the physical and chemical indexes of vegetable oil frying by low-field nuclear magnetic resonance（LF-NMR）was systematically investigated,and the simultaneous models of the physical and chemical indexes of vegetable oil frying process were established by combining the LF-NMR technique with multiple linear regression（MLR）and partial least squares（PLS）.This work can provide theoretical and technical support for the supervision of frying oil enterprises and quality inspection departments.The theoretical basis of LF-NMR for the determination of physicochemical parameters in vegetable oil frying process was systematically investigated.The changes in triglyceride contents,fatty acid compositions,and physicochemical properties of different oils and fats during frying were investigated in detail,and the influence of oil type and frying time on the LF-NMR relaxation properties was analyzed to investigate the correlation between physicochemical indices and LF-NMR relaxation properties.The results showed that the acid,polar component,anisotropy,absorbance and viscosity of rapeseed,soybean and rice oils all increased during frying,while the iodine value gradually decreased.The rate variation of glycerides and triglycerides content was significantly greater than that of glycerides,and the degree variation was in the order of Rapeseed oil>Rice oil>Soybean oil.With the increase of frying time,the decay rate of LF-NMR echo curves of fats and oils gradually increased,the relaxation times T₂₁,T₂₂and T₂₃gradually decreased,and the peak areas of S₂₁,S₂₂and S₂₃also changed with the rate variations S₂₁>S₂₃>S_22.Within 48 h of frying,S₂₁had the biggest variation,and the order for variation was Soybean oil>Rapese oil>Rice oil.The acid,polar component,anisidine,viscosity,absorbance,iodine,and carbonyl values of the three oils and fats showed significant or extremely significant correlations（R>0.8）with the relaxation indices T_2w,T₂₁,T₂₂,S₂₁,and S₂₃.The optimal conditions for the detection of vegetable oils by LF-NMR were systematically analyzed,and the synchronization models for the frying physicochemical indexes of rapeseed oil,soybean oil and rice oil were established by using LF-NMR relaxation characteristics and echo curve data combined with multiple linear regression.The results showed that the optimal conditions for the detection of the physicochemical parameters of the three frying vegetable oils by LF-NMR were as follows:0.5 ms,Echo Count 3000,the sample volume 3 m L,and the sample temperature 45℃.In the MLR synchronization model based on LF-NMR relaxation properties,R²was bigger than 0.91 for all indicators except anisidine,and RAD and RMSECV were smaller than 0.28 and 7.5,respectively.Compared with the echo-curve-based MLR synchronization model,this model can improve the prediction accuracy by more than 40%.As for the three frying oils,the MLR synchrotron model for the relaxation characteristic index of rapeseed oil had the better performance,with RAD and RMSECV values of 0.047 and 1.62,while the synchrotron model for soybean oil has relatively poor performance.The best MLR synchronization model was selected for the prediction of the unknown samples,with R²values all larger than0.96 and RAD values all smaller than 0.01.However,the interaction prediction performances for the physicochemical indices of different types of vegetable oils were poor.Synchronization models for the physicochemical parameters of deep-fried vegetable oils were developed based on LF-NMR relaxation properties and echo curve data combined with a partial least squares method.The prediction performances of synchronization models for different types of oils and fats were compared.The results showed that the numbers of principal factors in the PLS synchronization model using relaxation characteristic indexes for rapeseed oil,soybean oil and rice oil were 6,5 and 6,respectively,while the numbers of principal factors in the PLS synchronization models using echo curve data for rapeseed oil,soybean oil and rice oil were 9,6 and 6,respectively.Among the PLS synchronization models using the relaxation characteristic index,all models except the anisotropic model have R²of at least 0.91,RPD value greater than 3.3,and RMESCV value less than 10.0.Compared with the echo-curve-based PLS synchrotron model,the prediction accuracy improved by more than 19.4%.In contrast to the MLR synchrotron model,the RMSECV values of the physical and chemical indices for rapeseed oil,soybean oil and rice oil decreased by 7.1%,1.2%and 0.2%respectively.Among the three frying oils,the prediction performances of the PLS synchronization model for rapeseed oil were better than those for soybean oil and rice oil.The best PLS synchronous model can obtain the performances of the RPD greater than 2.7 and RMSEP values less than 7.4,respectively.Compared with the best MLR synchronous model,the prediction accuracy improved by more than 0.46%,which indicated that the simultaneous detection of the physicochemical indicators of deep-fried vegetable oils by LF-NMR techniques was feasible.