Study On Detection Of Adulterated Sesame Oil And Tea Seed Oil By Near Infrared Spectroscopy

Edible oil plays a key role in daily life,it can be seasoned,and its composition can provide the necessary energy and other nutrients for the human body.The variety of Edible Oil are plentiful,different Edible Oil has its own composition,and different composition has different value.The Sesame Oil is rich in aroma and nutritious,and the ratio of oil to oil of Tea Seed Oil is reasonable,so it can keep fit.The price of Sesame Oil and Tea Seed Oil is higher than most Edible Oil.However,to extract high profits,unscrupulous merchants make shoddy Oil,even add harmful Oils,which would have terrible influence in our human body.Therefore,it is necessary to establish a method to quickly detect the adulteration of Edible Oil.In this dissertation,Sesame Oil and Tea Seed Oil were used as subjects.Near-infrared(NIR)spectroscopy combined with Chemometrics was used to discuss the issue about identification of adulteration in Edible Oil.Binary,ternary,multivariate qualitative and quantitative model of adulteration in Sesame Oil and Tea Seed Oil were established.The main contents are as follows:(1)To study the qualitative and quantitative detection methods of binary adulteration in Sesame Oil and Tea Seed Oil,binary identification model of adulteration in Sesame Oil and Tea Seed Oil were established,qualitative model can recognize authenticity in adulterated Sesame Oil or Tea Seed Oil,and quantitative model can predict the content of adulterated oil in Sesame Oil or Tea Seed Oil.Using NIR spectroscopy combined with Support Vector Machine Classification(SVC),a full-wavelength model and feature-wavelength model which extracted feature-wavelength by competitive adaptive reweighted sampling(CARS)or Backward interval Partial Least Squares(BiPLS)were established.As a result,every model can identify the authenticity of Sesame oil and Tea Seed Oil,what's more,the highest recognition accuracy is SNV-SVC model established by Standard Normal Variate(SNV)pretreatment of sesame oil and this Prediction Set accuracy of the model is 99.4975%.In addition,NIR spectroscopy combined with Support Vector Machine Regression(SVR)can be used to predict the content of adulterated oil in Sesame Oil or Tea Seed Oil.The model's Regression(R)are higher than 99%,and the Mean Squared Error(MSE)are lower than 0.0605.(2)To study the qualitative and quantitative detection methods of ternary adulteration in Sesame Oil and Tea Seed Oil,ternary identification model of adulteration in Sesame Oil and Tea Seed Oil were established,qualitative model could recognize authenticity in adulterated Sesame oil or Tea Seed Oil,and quantitative model could predict the content of Sesame Oil or Tea Seed Oil.Preprocessing by Multiple Scatter Correction(MSC),Standard Normal Variate(SNV)or Standard Normal Variate combined with De-Trending(SNV-DT),then using BiPLS selected the feature-wavelength.The results show that NIR spectroscopy can recognize authenticity in adulterated Sesame Oil and Tea Seed Oil,also can detect the content of Sesame Oil or Tea Seed Oil.The accuracy,identification of ternary adulterated models,can up to 100%(ternary and authentic Tea Seed Oil MSC-BiPLS-GA-SVC),and the Regression of ternary adulterated content models can up to 99.7254% and MSE is 0.0234(the content's prediction model of ternary and authentic Tea Seed Oil SNV-BiPLS-CV-SVR).(3)To study the qualitative and quantitative detection methods of multivariate adulteration in Sesame Oil and Tea Seed Oil,multivariate identification model of adulteration in Sesame Oil and Tea Seed Oil were established,qualitative model could recognize authenticity in adulterated Sesame Oil or Tea Seed Oil,and quantitative model could predict the content of Sesame Oil or Tea Seed Oil.Apply MSC or SNV to preprocessing for removing interference information and improving the model's prediction ability.Then,apply CARS or Synergy interval Partial Least Squares(SiPLS)to select feature-wavelength quickly,the models' variables reduced,and efficiency become higher.The model's consequence indicate that the accuracy of qualitative models are excellent,among which the test model for authenticity of multivariate adulterated Sesame Oil is as high as 100%;in the meanwhile,the established quantitative model can predict the actual content of multivariate adulterated Sesame Oil or Tea Seed Oil samples,and the overall prediction R are about 1,all MSE of the Prediction Set are small and the difference between Prediction Set and Correction Set are small,so the models' prediction ability are stable and the prediction effects are ideal.