Theoretical Study On Fluorescence Detection Technology Of Preservative Based On Optimized BP Neural Network

Preservatives as a kind of common food additives are used in food production,transportation and storage widely. In recent years, food safety events happened frequently,which seriously endanger human health. Therefore, the detection of food preservative caused common concern of food and testing industry. In view of the strong point of fluorescence detection technology, high sensitivity, good selectivity, simple operation,fluorescence method is used to detect Potassium Sorbate and Sodium Methylparaben in orange juice respectively.Firstly, we analyze the basic principle of fluorescence detection and the feature of fluorescence. Then according to fluorescence characteristics of potassium sorbate, sodium methylparaben and orange juice, we set the scanning range of excitation and emission wavelength. We detect the fluorescence spectrum of the two preservatives used FS920 fluorescence spectrometer.Secondly, based on the theory analysis of wavelet threshold denoising and wavelet packet denoising, two-dimensional wavelet threshold denoising and wavelet packet denoising are used to process the fluorescence spectra. In order to evaluate the effect of wavelet threshold denoising and wavelet packet denoising Root Mean Square Error,Signal to Noise Ratio and Normalized Correlation Coefficient are used. Comparing the two denoising effects the results show that the denoising effect of wavelet packet is better than that of wavelet denoising method.Thirdly, to solve the defects of local minimum and slow convergence of BP neural network, the optimized methods based on Genetic Algorithm are proposed. On the basis of thorough theoretical analysis, the measurement models of GA-BP neural network are established and the two preservatives are detected respectively.Finally, basic principle of Particle Swarm Optimization Algorithm was analyzed, and a BP neural network optimization method based on Particle Swarm Optimization algorithm was designed. The detection of potassium potassium and sodium benzoate in orange juice is completed, respectively. Then we compare the two kinds of optimizationmethods, and results show that the detection results of PSO-BP neural network are best followed by GA-BP neural network.