Study On The Detection Technology Of Oilseeds And Oils Based On Near Infrared Spectroscopy

Oilseeds and oil products play an indispensable role in our daily diet,which provide many different nutritious and functional components for human such as fatty acid,sterols and polyphenols.Meanwhile,they are the important basis for food and feed industry.Therefore,their quality and safety have an important influence on human health.Now,the quality assurance methods for edible vegetable oils and the detection methods for nutriment in oilseeds were time-consuming and destructive for samples based on the instrumental analysis,and cannot meet the demand of decentralized mode under China's market-economy.In this paper,the rapid and nondestructive detection technology for quality assurance of edible vegetable oils and nutrients in oil seeds was established based on the near infrared spectrometry.The main research contents are as follows:1.Multivariate adulterated detection method of edible oils was built based on NIRFor the illegal merchants,there is no doubt that they can avoid supervision by mixing all kinds of sources and different proportions of edible oil.Therefore,the research on multivariate adulterated detection based on the simplex linear programming theory was developed.Taking flaxseed oils for example,which possess high risk of being adulterated with other cheaper edible oils.The adulterated flaxseed oils were prepared by mixing other edible oils with pure flaxseed oils.After important variables were selected by orthogonal partial least squares discriminant analysis(Orth-PLSDA),a one-class partial least squares(OCPLS)classifier was used to build an authentication identification model.The results indicate that the model could identify the authentic flaxseed oil samples with accuracy of 100%,for the adulterated oil samples(?5%)with a high accuracy of 95.77%.This method could identify adulteration with 5 kinds of oils mixed at any unknown ratio.Meanwhile,it provides the new idea and technical supporting for the multivariate adulteration detection of edible vegetable oils.2.Prediction models of fatty acids in oil seeds were established by NIRTaking rapeseeds as the example,after acquiring the NIR spectra of 510 rapeseed samples,the first-order derivative(1st Der)and the standard normal variation(SNV)were applied to preprocess the NIR spectrums.After the characteristic wavelength selection based on competitive adaptive reweighted sampling(CARS)algorithm,the predicting models of main fatty acids in rapeseeds were developed by partial least squares method(PLS).In previous study,the models were developed based on the relative contents of fatty acid.However,the relative content cannot confirm the law of Lambert-Beer.In this study,the correcting method was developed to enhance the accuracy of the predicting models.The PLS prediction models were established for 10 kinds of fatty acids in rapeseed.And the independent validation sets were applied to evaluate the predicting capability of models.The precision of models were high fitting and showed great predictive ability(models'Q~2 were all greater than 0.7933,the average of predicting absolute error were less than 13%).Compared with the predicting results by relative content,the results of the maxium of predicting errors,repeatability and reproducibility of the model based on the correcting method were better.The maxium of absolute errors of each fatty acid was decreased 10%.3.Rapid nondestructive detecting method of vitamin E in rapeseeds were developed by NIRAfter acquiring the NIR spectra of 510 rapeseed samples,the second-order derivative(2nd Der)and the standard normal variation(SNV)were applied to preprocess the NIR spectrums.Meanwhile,the total vitamin E in rapeseeds was detected by high performance liquid chromatography(HPLC).After the variable selection by competitive adaptive reweighted sampling(CARS)algorithm,209 important wavelengths were determined.The predicting models of total vitamin E in rapeseeds were established by partial least squares method(PLS).The precision of models were great fitting and showed great predictive ability(models'Q~2 were 0.8599,the average of predicting absolute error was 1.67 mg/100g).The results showed that the predicting results models for total vitamin E based on NIR reached the requrements of detecting results of HPLC method in national standard.