Spectral Quantitative Detection Of Acid Value And Aflatoxin In Edible Oil And Identification Of Oil Species Characteristic Value

With the continuous improvement of people's living standards,the hot topic of today's society has changed from being full to eating.Especially in the current situation of frequent safety of edible oils,it is an urgent problem for every consumer to be able to effectively distinguish and manage the quality of edible oil.At present,the detection of the quality of edible oil mainly uses physical and chemical methods.The traditional method of operation is cumbersome,time consuming,and expensive to test instruments.Therefore,it is extremely important to explore a fast and efficient measurement method.Based on laser near-infrared spectroscopy,this paper studies two main components affecting the quality of edible oil,and studies a method for rapid quantitative prediction of acid value of edible oil and aflatoxin,and studies on the identification of various characteristic components of edible oil.The feasibility of the type of edible oil was finally established,and the quantitative identification model of acid value and aflatoxin and the qualitative identification model of edible oil were established.The main contents are as follows:(1)Research and exploration on the basis of near-infrared spectroscopy technology,the near-infrared spectrum of 42 oil samples was collected,and the quantitative value prediction model of acid value and the qualitative identification model of edible oil were established by using the true values of spectra and sample acid values.Four different preprocessing methods,multiplicative scatter correction(MSC),standard normal variate transformation de-trending(SNV-DT),moving average smoothing(MAS)and savitzky-golay(SG),combined with competitive adaptive reweighted sampling and partial least squares(CARS-PLS),interval partial least squares(iPLS)and succession projection algorithm(SPA)three feature variable extraction methods,through particle swarm optimization(PSO)and genetic algorithm(GA)Two parameter optimization algorithms were optimized.The support vector machine(SVM)was used to establish the support vector character(SVC)and support vector ration(SVR)model of edible oil,and the quantitative prediction and type characterization of the optimal edible oil acid value were studied.Modeling techniques and methods for identifying models.The experimental results show that the quantitative prediction of the acid value of edible oil using the three pretreatments of MSC,SNV-DT and MAS can predict the performance of the model,and can realize the quantitative prediction of the acid value of edible oil.The correlation coefficient R of the prediction set of MSC-CARS-PLS-SVR model reaches 100%,which indicates that the model can theoretically realize the qualitative identification function of edible oil species.(2)Study and explore the technical methods for qualitative identification and quantitative prediction of aflatoxin in peanut oil,and establish a qualitative identification and quantitative prediction model for aflatoxin in peanut oil.Two kinds of preprocessing methods,SNV-DT and MSC,iPLS and CARS-PLS are used to extract two characteristic variables,grid search algorithm(GS)and PSO two parameters optimization algorithm,combined with support vector machine to establish peanut oil aflatoxin Qualitative identification and quantitative prediction models were used to study the modeling techniques and methods for determining the qualitative identification and quantitative prediction models of the optimal peanut oil aflatoxin.The experimental results show that the accuracy of MAS-CARS-PLS-PSOSVC prediction set model of peanut oil aflatoxin is 100%,and the accuracy of SNVDT-CARS-PLC-GS prediction set model is predicted by peanut oil aflatoxin.The highest reached 98.9818%,meeting the expected goals.(3)Research and explore methods for identifying edible oils based on nine characteristic components of edible oil.Three optimization parameters of GA,PSO and GS were used to optimize the algorithm.The support vector machine was used to establish three kinds of characteristic components to identify the qualitative model of edible oil.The experimental results show that the method of parameter optimization has different effects on the accuracy of model identification.The correlation coefficient R of the correction set of PSO-SVC model is 98.5294%,the correlation coefficient R of prediction set is 100%,and the correlation coefficient is higher than other two types of models.It is concluded that PSO-SVC is the most effective in the qualitative identification models of various eigenvalues.