Study On Raman And Infrared Spectroscopy To Evaluate The Oxidation Of Nut Oils

Nuts are one of the most popular leisure foods with relatively high fat content.Extensive oxidation and deterioration of nut oil may be easily occurred due to the high unsaturated level.In this study,the complementary Raman spectrum and Fourier transform infrared spectrum（FT-IR）were adopted to monitor and track the oxidation process of eight types of nut oil.Combined with chemometrics methods,four quantitative prediction models of oxidation indexes were established,which provided a simple,effective and sensitive mean of rapid detection,quality evaluation and oxidation-index prediction of nut oil.The main research contents are as follows:Raman spectroscopy was used to characterize the lipid oxidation during accelerated storage of nut oils.Looking at the oxidation process,the Raman characteristic peak was determined to be 1656 cm^-1.The relative intensity ratio of characteristic peak and the internal standard peak,i.e.I_（1656）/I_（1746）represented the stretching vibration of cis double bond.The I_（1656）/I_（1746）value was decreased gradually.The relative intensity ratio of characteristic peak at970 cm^-11 and the internal standard peak,i.e.I_（1656）/I_（1746）represented the bending vibration of trans double bond.The I_（1656）/I_（1746）value was increased gradually.Meanwhile,results of four oxidation indexes,i.e.peroxide,p-anisidine,conjugated diene and carbonyl values were increased with the extend of storage duration.Pearson correlation analysis showed that the I_（970）/I_（1438）value was positively related to the characteristic value of chemical oxidation,the correlation coefficients were determined to be 0.7490 to 0.9900（p<0.01）.The I_（1656）/I_（1746）value was negatively related to the characteristic value of chemical oxidation;the correlation coefficients were determined to be-0.7140 to-0.9790（p<0.01）.Our results clearly indicated that the Raman spectroscopy has a great potential to determine the oxidation procedure of nut oils.The quantitative prediction model of four oxidation indexes of nut oil was established based on Raman spectrum.Firstly,the leverage value and the concentration of residual error were combined with spectral residual in order to eliminate the abnormal samples.The total of24 methods of single pretreatment and the combination of two pretreatment methods were adopted to pre-process the spectral data.Results showed that SNV and 1^st D were suitable for pre-processing peroxide value model;MSC and OSC methods were suitable for p-anisidine value and conjugated diene model;OSC and 1^st D were suitable for preprocessing carbonyl value model.The four oxidation indexes of unknown samples were predicted through the established partial least square regression（PLSR）model.The prediction determination coefficients were 0.8672,0.9133,0.9589 and 0.9325,respectively.Moreover,the predicted value of the model correction set and external cross-validation showed a good correlation with the measured value,which proved that the Raman spectroscopy combined with the PLSR model can rapidly quantify and detect the oxidation process.In addition,The PC1 and PC2 principal component analysis loading plots of the model with the best oxidation index showed that the main variables were located at four different bands,which indicated that the prediction of oxidation index was mainly affected by the vibration changes of methylene,C-H and C=C functional groups,respectively.Based on the whole spectrum range,the random forest algorithm was further used to optimize the peroxide value model.The total of 36 wavenumbers was selected to build up the model,and the effect and prediction ability were largely improved.Fourier transformed infrared spectroscopy was used to monitor the oxidation process of nut oil.It further combined with the chemometrics and achieved quickly detection and analysis of the oxidation index.The PLSR was used to establish the quantitative prediction model of different pretreatment methods and band range of oxidation index.The OSC pretreatment method was selected;the peroxide value model selected the full wavenumber range of 400～4000 cm^-1;the anisidine value model selected 3070～3800 cm^-1 wavenumber range（hydroperoxide-OO-H）;carbonyl value and conjugated diene model selected 930～990 cm^-1wavenumber range（trans=CH）.The prediction coefficients R²p of the unknown samples were higher than 0.9000.Results again proved that the infrared spectroscopy technology can be used to rapidly detect the oxidation index of oil.