Study On Rapid Detection Of Edible Vegetable Oil

Rapid detection techniques of edible vegetable oil are prospective and attractive subject for quality supervision of edible vegetable oil in market and quality control in manufacturers. Traditional methods based on chromatography are costly, labor intensive and time consuming, require large amounts of solvents and reagents, and are difficult to achieve on-site rapid test. This study focuses on edible vegetable oil using dielectric spectroscopy technology and Raman spectroscopy technology, combined with a variety of chemometrics to research the identification and detection methods for quality and fatty acid composition of edible vegetable oil. The main contents and conclusions are as follows:Dielectric spectroscopy combined with principal component analysis (PCA) was applied to building the classification model to classify camellia oil, peanut oil, canola oil, sunflower seed oil and soybean oil. Dielectric spectroscopy combined with partial least square (PLS) was applied to building the quantitative model to quantitatively predict the adulteration of camellia oil adulterated with peanut oil and sunflower seed oil respectively. Results show that the peanut oil prediction model's root mean square error of prediction (RMSEP) is 0.0247 and its coefficient of determination (R2) is 0.9982; the sunflower seed oil prediction model's RMSEP is 0.0139 and its R2 is 0.9988.Raman spectroscopy combined with least squares support vector machine (LS-SVM) based on multiple iterative optimization was applied to detect the type and content of the oil adulterated in extra virgin olive oil (EVOO). The type of adulteration oil adulterated in EVOO was identified by the identification model based on least squares support vector machine classifier (LS-SCR) and the integrated identification rate was 97%. Based on the above identification, the amount of adulteration was achieved by the quantitative model based on Least squares support vector machine regression (LS-SVR). Results show that the RMSEP of sunflower seed oil, soybean oil and corn oil model are 0.0074,0.0142,0.0121 respectively; and the correlation coefficient R of these model are 0.9998,0.9992, 0.9996 respectively. The same samples were modeled by PLS and artificial neural network (ANN) respectively for comparison. LS-SVR prediction was accuracy than PLS and ANN in this application.Raman spectroscopy combined with LS-SVM and the standard value achieved by gas chromatography(GC) was applied to establish a quantitative model to quantitatively predict the saturated fatty acids (SFA), oleic acid and linoleic acid of edible vegetable oil. Root mean square error of cross validation (RMSECV) of SFA, oleic acid and linoleic acid are 0.5673,0.0987,0.0708 respectively; the R are 0.7839,0.9967,0.9986 respectively; and the RMSEP are 0.5209,0.0795,0.0515 respectively.Accurate, simple, rapid and non-destructive for adulterated olive oil detection and edible oil lipid detection methods were achieved by this research.