Study On Identifying Peanut Oil Adulteration Based On Molecular Information

In this paper, the detecting methods for adulteration of peanut oil, which is mixed with cottonseed oil, soybean oil, rapeseed oil, palm oil and corn oil, were studied respectively. By chemometric methods, a relatively complete model was also established to identify adulterated peanut oil.Peanut samples which were collected from different regions of China(n=110) were collected. Then the total and 2-position fatty acid composition of oils were tested. The results showed that the variation coefficients of each fatty acid composition and content in peanut oils was small, so the research of peanut oil adulteration was feasible and had wide application.The changes of composition and content of total and 2-position fatty acid were studied for all the pure peanut oils and adulterated peanut oils, and the detection limits of each method were also determined. When cottonseed oil, soybean oil, rapeseed oil, palm oil and corn oil were adulterated into peanut oil respectively, the detection limits of these two methods were 15%, 5%, 5%, 10%, 15% and 20%, 5%, 4%, 20%, 50%.Through the analysis of pattern recognition technologies on the data of total and 2-position fatty acid, the following conclusions were obtained: The linear model- LDA couldn’t correctly identify the adulteration of peanut oil. Cluster analysis for identification of the adulteration of peanut oil was feasible, but the effect was poor when the adulteration was low. Coupled with appropriate kernel function parameters, the LS-SVM model can correctly identify the adulterated sample with accuracy 100% using the total and 2-position fatty acids respectively. These models were used to analysis multi-component-adulteration of peanut oils, the correct rates were 95.45% and 90.91% using the total and 2-position fatty acids composition respectively.The quantitative models of peanut oils adulteration were built by PLS method using the total and 2-position fatty acid composition respectively. The results showed that the prediction accuracy of specific models were much higher than that of global models. Compared with the models using 2-position fatty acid, the prediction accuracy of models using total fatty acid were higher, and its RMSEP values ranged from 3.32% to 4.69%.