Study On Detecting Methods And System Model For The Authentication Of Sesame Oils

In this paper, the detecting methods and identification model for the adulteration of sesameoils, mixed with rapeseed oils, soybean oils, palm oils, sunflower oils and cottonseed oils,respectively, were studied, and the results were as follows:A basic data about the triglyceride, total and2-position fatty acid composition of oils fromsesame seeds collected from different regions of China （n=117） was provided, and theresults showed that the variation coefficient of fatty acid composition and content of sesameoil was small, the research was meaningful and the results had universal applicability.The changes of composition and content of fatty acids and TGs were studied for all thepure sesame oils and adulteration sesame oils, and the detection limits of each method werealso determined. When mixed with rapeseed oils, soybean oils, palm oils, cottonseed oils andsunflower oils, respectively, compared with pure sesame oils, the detection limits were5%,4%,21%;5%,10%,7%;5%,20%,12%;10%,32%,20%;50%,50%,39%in total,2-position fatty acid and triglyceride composition analysis methods.Rancimat analysis showed that pure sesame oils had the best oxidation stability, comparedwith adulterated sesame oils, the adulteration of sesame oil was determined by oxidativestability, but this method can not be used alone. When using qualitative tests, the detectionlimit of cottonseed oil was2%. The detection limit of soybean oil was10%, but the bleachedand deodorized soybean oils can not be detected. Rapid freezing methord was not insensitiveto the palm oil of low melting point, the detection limits were40%and20%when mixedwith24℃and26-28℃palm oils, respectively.Through the conection of pattern recognition technologys and fatty acid composition dataof pure sesame oils and adulterated sesame oils, the following conclusions were obtained: thePCA was an effective method for the identification of pure vegetable oils; The clusteringresults of SOM were significantly better than PCA, but when adulterated with a little otheroils, it was still difficult to separate pure sesame oils and adulterated sesame oils throughSOM methord; The qualitative identification model of sesame oils adulteration was builtthrough PSO-SVM algorithm, using cross validation and test set validation, the classificationaccuracy rate was100%, and the lowest classification limit was5%. When detecting foreignmulti-sesame oil adulterated samples using the model, the classification accuracy rate reached95.45%, this method was effective for the adulteration identification of sesame oil. Several PLS-based models were built using the mixtures with specific adulterant, The RMSEs forprediction sesame oil range from1.19%to4.29%, which meet the requirements of routinefood control. But the predictive capability reduced for multi adulteration.