Identification Of Edible Oil Species With Multiple Eigenvalues And Detection Of Aflatoxin By Near Infrared Spectroscopy

Edible oil is one of the seasonings that people use daily.At the same time,edible oil can provide the human body with essential nutrients such as unsaturated fatty acids and various vitamins.The safety of edible oil ha s always been a controversial matter.The purchase of inferior edible oil will not only cause economic losses to consumers,but also cause harm to the human body,so the detection of edible oil is particularly important.The iodine value,acid value,saponification value,saturated fatty acid and unsaturated fatty acid in edible oil can be used as characteristic parameters of oil quality analysis.Conducive to rapid detection of edible oil in real time.Therefore,it is extremely important to explore a detection method that can be quickly and effectively detected.After the introduction of near infrared(NIR)spectroscopy,it has blossomed in various industries over time.Near-infrared spectroscopy analysis technology has the advantages of fast,effective,and high efficiency,and has received certain attention in the detection of oil and various foods.Based on near-infrared spectroscopy analysis technology,the subject has carried out research on oil detection.The main research contents and conclusions inc lude:(1)Based on near-infrared spectroscopy technology,collect spectral data of 40 8 types of oil samples,apply four pre-processing methods,two characteristic wavelength extraction algorithms and two parameter optimization methods to establish iodine value,acid value,saponification value,Quantitative prediction model of saturated and unsaturated fatty acids.Analyze and compare the prediction effects of the models established by different algorithm paths.The results show that the correlation coefficients R of the prediction and correction sets of most models are above 0.98.(2)After determining the true content of aflatoxin B1 in the peanut oil sample,quickly collect its spectral data.Three pretreatment methods,two characteristic wavelength extraction algorithms and three parameter optimization methods were used to establish a quantitative prediction model of aflatoxin B1 in peanut oil.The results show that the moving average smoothing method(MAS)combined with particle swarm optimization(PSO)and grid search method(GS)can establish the optimal quantitative prediction model.The correlation coefficient R of the model prediction set is basically 1,and the correction set is related.The coefficients R are all above 0.998.After comparing the p arameters of several optimal quantitative prediction models,the MAS-SPA-GS quantitative model correction set with the smallest average relative error(MRE)and the predicted value of the prediction set are selected as the input value Y of the qualitative identification model.Three different parameter optimization methods are used to establish a qualitative identification model.The accuracy rates of the calibration and prediction sets of the three established identification models are above 96.6%,and e ven the accuracy rates of the GA-SVC and GS-SVC models have reached 100%.(3)Based on five characteristic values,such as iodine value,acid value,saponification value,saturated fatty acid and unsaturated fatty acid,establish a grease type identification model.The prediction value of the quantitative prediction model with the best characteristic values is used as the input function X of the qualitative classification model,and the oil type is used as the output function Y.Two parameter optimization methods are used to establish the oil type identification model.The results show that the accuracy of the GA-SVC model correction set and prediction set are above 93%.There was one misjudgment during training,and only two misjudgments occurred during pr ediction.The research results show that it is feasible to use a variety of characteristic values to quickly identify the type of oil.