Detection Method And Research For The Quality Of Edible Oil Based On Near Infrared Spectroscopy

Edible oil is an important and indispensable component in the human daily diet ,its quality have an important impact on human health. Traditional testing of Edible oil mainly based on chemical methods, it often requires a variety of chemical apparatus and reagents and need for pretreatment of samples,and testing operation is cumbersome and time-consuming. The article is focused on the study of edible vegetable oil using Near Infrared Spectroscopy and chemometrics analysis to research the detection and identification methods for quality of edible vegetable oil. The main findings are as follows:1. Compare and analysis the results of the calibration model for quantitative detection of fatty acids (oleic acid, linoleic acid, linolenic acid)using 10 kinds of pretreatment methods on vegetable oil, we can see the first derivative processing and multiple scattering correction pretreatment were the optimal pretreatment methods for the quantitative detection of oleic acid; second derivative treatment was the optimal pretreatment methods for the quantitative detection of linoleic acid; first derivative treatment and vector normalization were the optimal pretreatment methods for the quantitative detection of linolenic acid.2. Near-infrared quantitative analysis model based on partial least squares (PLS) for the detection of the three kinds of fatty acids(oleic acid, linoleic acid, linolenic acid) in the edible vegetable oil were established . we can see oleic acid, linoleic acid, linolenic acid determination coefficient R~2 of the calibration model were 0.9752,0.937,0.9853, RMSECV were 1.04%, 1.3%, 0.388%,they all have high determination coefficient, determination coefficient R~2 of validation model were 0.9753,0.9655,0.9722; RMSEP were 1.11%, 1.55%, 0.487%, the model can forecast accurately.It shows that the rapid non-destructive testing of oleic acid, linoleic acid, linolenic acid in the edible vegetable oils can be realized by the Quantitative analysis model.3. Near-infrared quantitative analysis model based on partial least squares (PLS) for the simultaneous detection of the three kinds of fatty acids(oleic acid, linoleic acid, linolenic acid) in the edible vegetable oil was established. We can see the determination coefficient R~2 of oleic acid detection calibration model was 0.9693, RMSECV was 1.15%; the determination coefficient R~2 of linoleic acid detection calibration model was 0.9671, the RMSECV was 1.46%; the determination coefficient R~2 of linolenic acid detection calibration model was 0.9792, RMSECV was 0.461%; determination coefficient R~2 of oleic acid validation model was 0.9693, RMSEP was1.3% ; determination coefficient R~2 of linoleic acid validation model was 0.9606,RMSEP was 1.66%;determination coefficient R~2 of linolenic acid validation model was 0.9731, RMSEP was 0.479%,they all have high determination coefficient.It Shows that the model can detect oleic acid, linoleic acid, linolenic acid simultaneously very well.4.using near-infrared spectroscopy qualitative analysis to study the identification of four different varieties of edible vegetable oil, the identification of mixing vegetable oil ,the identification of six different brands of sesame oil.Using hierarchical clustering method, principal component analysis, BP artificial neural network method to identify the four different varieties of edible vegetable oil, results showed that the combination of principal component analysis and BP artificial neural network method is the best method to identify the four different varieties of vegetable oil ,and the recognition rate was 100%; Using hierarchical clustering method, principal component analysis, BP artificial neural network method to identify the mixing vegetable oil, results showed that the combination of principal component analysis and BP artificial neural network method is the best method to identify the mixing vegetable oil ,and the recognition rate was 96.15%; Using hierarchical clustering method, principal component analysis, BP artificial neural network method to identify the six different brands of sesame oil, results showed that the combination of principal component analysis and BP artificial neural network method is the best method to identify the six different brands of sesame oil ,and the recognition rate was 83.33%. The results show that the near-infrared spectroscopy technique can be used as the new rapid identification testing method for the quality of edible vegetable oil.