Study On Rapid And Quantitative Determination Of Edible Blend Oil Based On Raman Spectroscopy

The characteristics of the spectral data and the samples are nonlinear in the process of rapid quantitative detection of edible oil.Because there are many factors such as the large range of the tested samples,the interaction of the components and the external environment in the actual process.In this paper,a method for the quantitative determination of edible blend oil by combining Raman spectroscopy with chemometrics is presented.This method solves the problem of quantitative detection of edible oil(three components).Firstly,the domestic and foreign research present situation based on the qualitative and quantitative detection of edible blend oil has carried on the thorough research in this paper.And the paper clarified the significance of the study.The mechanism of Raman scattering is introduced.The basic principle of qualitative and quantitative detection based on Raman spectroscopy is introduced.Secondly,the nonlinear mathematical model was used to analyze the content of edible oil in this paper.The nonlinear algorithms of Support Vector Machine(SVM),Least Squares Support Vector Machine(LSSVM),Particle Swarm Optimization(PSO)and Particle Swarm Optimization Least Squares Support Vector Machine(PSO-LSSVM)are compared in this paper.The regression forecasting method of LSSVM and PSO-LSSVM are proposed.Again,when the proportion of edible oil is different,the change of functional group content will lead to the change of the characteristic peak strength of edible oil.Therefore,a rapid quantitative detection system based on the relationship between the content of edible oils and the strength of characteristic peaks is established.It can realize the identification of mixed oil content.The Raman spectra of 66 samples were detected and pretreated.The quantitative prediction model system of LSSVM and PSO-LSSVM is designed and the prediction model is validated.Finally,the quantitative prediction models of LSSVM and PSO-LSSVM multi input single output are compared in this paper.Then the LS-SVM model parameters areoptimized by PSO algorithm.The predictive ability of the model was analyzed by the correlation coefficient and mean square error in the test samples.