Research On Quality Detection Method Of Vegetable Oil Based On Near Infrared Spectroscopy

The vegetable oil market is disturbed by doping and inferior oils,which seriously damages food safety and market order.Therefore,the development of rapid,reliable and compact vegetable oil quality testing methods is of great significance and application value for improving food safety and quality supervision.The traditional vegetable oil quality testing relies mainly on chemical means,the operation is cumbersome,time-consuming and laborious.Based on the combination of near-infrared spectroscopy and two-dimensional correlation analysis,this paper studies the identification of vegetable oils.Based on the combination of near-infrared spectroscopy and chemometrics,qualitative analysis of the doping of olive oil and quantitative analysis of the doping amount of frying oil in vegetable oil were carried out.The main contents are as follows:（1）Utilizing NIR absorption spectroscopy and two-dimensional correlation analysis,six common vegetable oils such as rapeseed oil,soybean oil,olive oil,peanut oil,sunflower oil and corn oil were classified and identified.Firstly,the perturbation spectra of various vegetable oils were obtained with the concentration as the disturbance factor,and then the two-dimensional correlation analysis was carried out to obtain the two-dimensional correlation synchronization spectrum.The autocorrelation spectrum of the synchronization spectrum can reflect the extent to which the spectral intensity changes with the disturbance factor.The principal component analysis is used to reduce the dimensionality of the autocorrelation spectrum,and the first eight principal components are retained as input parameters of the classification model.By calculating the Euclidean distance between the main components of the autocorrelation spectrum of various types of vegetable oils,different types of vegetable oils were identified on the basis of the principle of proximity.（2）The combination of near-infrared absorption spectroscopy and pattern recognition algorithms in chemometrics was used to qualitatively identify three samples of pure soybean oil,pure olive oil,and doped olive oil（mixed with soybean oil）.Principal component analysis discriminant（PCA-DA）,partial least squares discriminant（PLS-DA）and back propagation artificial neural network（BP-ANN）were used to establish qualitative discriminant models.The three models were optimized by optimizing modeling bands and spectral pretreatment methods.Finally,the model based on BP-ANN had the highest discriminant rate,reaching 96.71%.（3）The concentration of fried oil in pure soybean oil was quantitatively analyzed by near-infrared absorption spectroscopy combined with partial least squares regression（PLSR）.The optimal modeling band(8500～8000 cm^-1)is determined by the band selection,and the preprocessing method is optimized on this band:standard normal variate correction（SNV）combined with the first order derivative.Then,using the competitive adaptive reweighted sampling（CARS）algorithm for wavelength selection,the dimensional reduction of the spectral data is performed,and the uncorrelated nonlinear variables are eliminated.The PLS quantitative analysis model optimized by the pretreatment method and the PLS model established after the wavelength selection of the CARS algorithm are compared.The effect of the latter is better.The correlation coefficient of the training set can reach 0.993,and the test correlation coefficient can reach 0.971.