Research On Identification And Quantitative Detection Of Flour Adulteration Based On Near Infrared Spectroscop

Wheat is a traditional grain crop in China.After thousands of years of cultivation history,my country has become the world’s largest producer and consumer of wheat,as well as a traditional wheat importer.As the main processed product of wheat,flour is one of the main grains in people’s daily diet.However,some unscrupulous merchants mix cheap substances or excessive additives into flour in order to obtain high profits,which will have a negative impact on the economic interests and health of consumers,and has become one of the food safety topics that the public is very concerned about.Therefore,rapid and accurate detection of illegal adulterants in flour is necessary.In view of the shortcomings of slow test speed,low precision and high cost when using traditional detection methods to detect adulterated talcum powder,gypsum powder,and azodicarbonamide(ADA)in flour,a combination of near-infrared spectroscopy(NIRS)and chemometrics was carried out.Methods Research on the classification,identification and quantitative detection of illegally adulterated flour samples,taking spectral feature extraction as the main line,using PLS,SVM,and RBFto build a classification system.The classifier and quantitative detection model makes it possible to quickly and accurately find and identify illegal substances that have been added to flour.Here are the most important parts of the study and what it found:(1)A hybrid genetic algorithm(HGA)was created by combining the genetic algorithm(GA)and the quantum genetic algorithm in order to increase the effectiveness and precision of NIRS detection.(QGA).The goal function was the cross-validation root mean square error of the PLS regression model.Improve the design of GA by expanding the perturbation solution set.Discuss the combination of HGA algorithm and other classic algorithms to optimize the characteristic wavelength,use HGA respectively for the characteristic wavelength optimization of reverse interval partial least squares(Bi PLS)and competitive adaptive reweighting algorithm(CARS),and construct a hybrid genetic algorithm-Inverse interval partial least squares method(Bi PLS-HGA),hybrid genetic algorithm-competitive adaptive reweighting algorithm(CARS-HGA)to optimize the characteristic wavelength for the detection of illegal adulterants in flour.(2)Establish a classification and discrimination model for flour adulteration with talcum powder,gypsum powder,and azodicarbonamide.First,a total of 360 flour adulterated samples with talc powder,gypsum powder and azodicarbonamide were prepared according to a certain adulteration content gradient as research objects.Secondly,the near-infrared spectral data of the flour adulterated samples were collected.After the spectral data were preprocessed by Savitzky-Golay(SG)smoothing and standard normal transformation(SNV),the sample sets were manually divided by selecting one out of three.Third,use PCA,LLE,and Isomap to extract spectral data features,and use the wavelength variable calculated by the dimensionality reduction method as a variable input,and finally establish partial least squares(PLS),support vector machine(SVM),and radial basis neural network respectively.(RBF)model,in which the recognition rate of LLE-PLS,LLE-SVM,and Isomap-RBF prediction set reaches 100%,considering the comprehensive iteration time to consider the LLE-PLS identification effect is the best,it only takes 2.57s to realize the three different Fast,accurate classification of illegally adulterated flour samples.(3)Establish a quantitative detection model for flour adulteration with talcum powder,gypsum powder,and azodicarbonamide.The collected spectral data were respectively subjected to SNV,MSC,SG convolution smoothing and their combination comparison to select the best preprocessing method for each adulterated sample.Secondly,the abnormal sample elimination method is used to remove abnormal samples,and the sample set partition method(SPXY)based on the combination of X-Y distances is used to divide the sample set into a calibration set and a verification set in a ratio of 3:1.Again,explore Bi PLS,CARS,Bi PLS-CARS,Bi PLS-HGA,CARS-HGA 5 spectral The characteristic wavelength,and finally establish the corresponding PLS model.characteristic wavelength optimization methods to select three types of flour adulteration samples After comprehensive analysis,the following conclusions are drawn:the best detection model for talc powder samples is Bi PLS-HGA,and theR~2,RMSE,and RPD of the validation set are 0.929,1.097,and 3.795,respectively.The best detection model for gypsum powder samples is theR~2,RMSE,and RPD of the CARS-HGA verification set,which are0.942,1.290,and 4.574,respectively.The best detection model for azodicarbonamide samples is theR~2,RMSE,and RPD of the Bi PLS-HGA verification set.0.982,0.006,4.367.The findings demonstrate that the optimum selection of the NIRS characteristic wavelength can be achieved by combining the HGA algorithm with the existing characteristic wavelength optimization algorithm.,and can meet the needs of rapid detection of illegally adulterated talcum powder,gypsum powder,and azodicarbonamide in flour.