[Determination of the Sodium Methylparaben Content Based on Spectrum Fluorescence Spectral Technology and GA-BP Neural Network].

Sodium methylparaben as one kind of preservatives is widely used in our life, but it will do harm to health if it is eaten too much. So there are strict rules on the dosage of sodium methylparaben in every country. The fluorescence spectral properties of sodium methylparaben in aqueous solution and orange juice solution are analyzed with FS920 fluorescence spectrometer. The research result shows that the fluorescence characteristic peak of sodium methylparaben solution is in λ(ex)/λ(em) = 380/5 10 nm, while sodium methylparaben orange juice solution has two fluorescence characteristic peaks which are in λ(ex)/λ(em) = 440/520 nm and 470/530 nm, and its best excitation wavelength is 440 nm. So it can be concluded from the result that there is a significant change between the characteristic peaks of sodium methylparaben in the two solution. Compared with the fluorescence characteristic peak of sodium methylparaben solution, thoses of sodium methylparaben orange juice solution are changed significantly, which are caused by the interference of orange juice fluorescence characteristics. In order to determine the content of sodium methylparaben in the fresh orange juice, a detection model of sodium methylparaben content in orange juice is built based on GA-BP neural network, according to the relationship between fluorescence intensity in λ(ex) = 440 nm and the content of sodium methylparaben orange juice solution. When the accuracy of the mean square error in the process of network training reaches 10(-3), the correlation coefficient of network output and that of the expected is 0.996. At the same time, a better prediction result can be obtained that the average recovery of the forecast samples is 98.67% and the average relative standard deviation is 0.86%. When the concentration ranges from 0.02 to 1.0 g x L(-1), the results testify that detection method based on fluorescence spectroscopy and GA-BP neural network can accurately determine the content of sodium methylparaben in orange juice. This method has the features of novelty and simplicity and it is expected to be applied to the determination of sodium methylparaben in other kinds of drink.