Study On Quality Analysis Method Of Vegetable Oils Based On Gas Chromatography-Ion Mobility Spectrometry

China produces and consumes large amount of edible vegetable oils,among which soybean oil,rapeseed oil,peanut oil and other vegetable oils account for about 90%of total domestic consumption.The edible vegetable oils sold on the market are not only of many kinds and sources,but also of different quality and nutritional value due to the differences of breed,climate,soil composition,preparation and processing technology.Some illegal vendors or small vegetable oil processing factories were blinded by relevant interests to adulterate or substitute high-price vegetable oil with low-quality or cheap vegetable oil.As a result,vegetable oil safety issues still occur in the world.Such incidents seriously violate the health and benefit of consumers,and impede the development of vegetable oil market economy as well.In this study,the quality detection of vegetable oil was studied based on GC-IMS?Gas Chromatography-Ion Mobility Spectrometry?technology from the perspective of flavor chemistry.The main research contents are as follows:?1?Taking four standard volatile organic compounds that naturally present in vegetable oils as the research objects,the optimum experimental parameters were acquired with 60?of incubation temperature,10 min of incubation time,200?L of sample volume and program control mode of carrier gas with the help of single factor test.?2?In order to establish a fast identification method for different kinds of edible vegetable oils based on the flavor components,187 oil samples of sesame oil,rapeseed oil and camellia oil were detected by GC-IMS instrument.Two methods?manual feature peak selection and Otsu's threshold segmentation?were used respectively to extract feature variables from GC-IMS two-dimensional map data.With the help of principal component analysis and kNN?k-Nearest Neighbors?pattern recognition algorithm,the comparison of these two methods were analyzed and the recognition models of different kinds of vegetable oil were constructed based on the odor fingerprints.The results show that compared with the subjective method of selecting feature peaks by human eyes,Otsu's threshold segmentation algorithm combined with map difference visualization method could present the different regions between different kinds of vegetable oils,realize the automatic selection of feature peaks corresponding to flavor substances,and distinguish different kinds of vegetable oils successfully.The correct rate of predicted set was 98.24%.?3?Multi-dimensional principal component method was used to directly analyze GC-IMS two-dimensional matrix data of adulterated samples,which produced by blending different proportion of other adulterated oil with canola oil.The typical discriminant analysis method was applied to supervise and identify different proportion of adulterated oil samples.The results show that except for the misjudgment of peanut oil samples with 20%adulteration,the others could be well distinguished.Multi-Linear Regression,Principal Component Regression and PLSR?Partial Least Squares Regression?were applied to establish quantitative prediction model of adulterated canola oil samples respectively.Results indicate that the correlation coefficients of calibration set and prediction set are all smaller that 0.85 and the corresponding root mean square error is 4.78%⁹.73%;the introduction of preprocessing methods could eliminate the irrelevant information in two-dimensional spectrum of GC-IMS,reduce the error between the real value and the predicted value and improve the prediction accuracy;PLSR algorithm combined with Savitzky-Golay filtering and normalization pretreatment could obtain the best prediction model,the correlation coefficient of the prediction set is 0.833,and the root mean square error of prediction set is 5.16%.?4?In order to construct the flavor fingerprint which could indicate the refining degree of rapeseed oil,and to distinguish the refining grades of rapeseed oil,124rapeseed oil samples with different refining grades were analyzed by GC-IMS instrument.Fourier transform infrared spectrometer combined with the second derivative algorithm were used to analyze different grades of rapeseed oil samples.The results show that there are differences among components composition of different refined grade rapeseed oil samples,which are as follows:the first-grade rapeseed oil samples have characteristic absorption peaks at 2 854 cm^-1,2 924 cm^-1 and 2933 cm^-1,corresponding to alkanes containing methyl,methylene and other functional groups respectively.Apart from the first-grade rapeseed oil samples,other grade oil samples all have absorption peak at 912 cm^-1,the corresponding substances may be a kind of compound containing carbonyl functional groups,which is one of the important sources of flavor forming substances in the later stage of rapeseed oil.To verify the reliability of GC-IMS,gas chromatography-mass spectrometry technology combined with headspace solid phase micro-extraction was applied to analyze the characteristic volatile flavour components of rapeseed oils with different refining grades.It showed that the number of flavour component numbers from rapeseed oils decreased with the improvement of refining grade.The main flavor components of rapeseed oils were aldehydes,alcohols,alkanes,esters,furans and pyrazines.The first-grade rapeseed oil mainly contained small molecule substances,the odor components of second grade samples were aldehydes and alcohols,and the third-grade oil samples were aldehydes,nitriles and a small number of pyrazines and furans.The fourth-grade oil samples contained the most odor components.At the same time,these samples were also detected by GC-IMS instrument.The color difference visualization method combined with Otsu's threshold segmentation algorithm were used to extract characteristic feature variables,and principal component analysis and three kinds of pattern recognition methods?k-nearest neighbor,linear discriminant analysis and quadratic discriminant analysis?were applied to establish discriminant models for distinguishing rapeseed oil grades.The results showed that all the samples with different refining grades were distinguished successfully,and better classification were found by kNN with more samples correctly clustered.The accuracy of the prediction sets of the three pattern recognition methods was more than 94%,demonstrating that GC-IMS technology could be used for judgment of rapeseed oil refining grades successfully.