Study On The Detection Of Food Additives In Beverages Based On Three-Dimensional Fluorescence Spectroscopy

With the development of science and technology,gradually reflects the chaotic use of many food additives,exposing many food safety problems.Non-compliant businesses use additives in excess of the standard or do not clearly indicate the additives,which undermines the rights and interests of consumers.In this paper,we use three-dimensional fluorescence spectroscopy combined with chemometric methods to detect several representative food additives in beverages,which is a reference value for food additive detection work.The main research work of this paper is as follows:(1)The principles and characteristics of three-dimensional fluorescence spectroscopy techniques and the various influencing factors of fluorescence generation of substances were studied.In order to reduce errors,the spectral data were de-Ramanised using the deduction blank method.For Rayleigh scattering in fluorescence spectra,the defect data reconstruction iterative algorithm combined with principal component analysis(MDR-PCA)is used to repair the data in the scattered region.For the problem of noise interference in fluorescence spectra,three denoising methods were compared and the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise Analysis(CEEMDAN)joint wavelet threshold denoising algorithm was compared based on the derived evaluation index to be more superior.(2)In order to detect the concentration of potassium sorbate in orange juice,a sparrow search algorithm optimised support vector machine(SSA-SVM)model was developed using the regression function of support vector machines,and the prediction of the concentration of potassium sorbate was completed using the CSO-SVM,PSO-SVM and GA-SVM models simultaneously.The results demonstrated that the SSA-SVM model was more suitable for the quantitative detection of potassium sorbate in orange juice.The average recovery of potassium sorbate concentration in the orange juice background was100.39%.And the subsequent amaranthine experiment eventually yielded an average recovery of 100.66% predicted by the SSA-SVM model,which also confirmed the superiority and generalisability of the SSA-SVM model.In order to detect sweeteners in carbonated beverages,a Dung Beetle Algorithm Optimised Support Vector Machine(DBO-SVM)model was developed using the support vector machine classification function.The CSO-SVM,PSO-SVM and GA-SVM models were also used to classify and identify sweetener mixture samples,and the results proved that the DBO-SVM model was more suitable for qualitative analysis of sweeteners in carbonated beverages,and the classification The accuracy of the classification was 100%.The findings were confirmed by subsequent colourant experiments.