Quantitative Prediction of Synthetic Food Colors by Fluorescence Spectroscopy and Radial Basis Function Neural Networks

In this paper, quantitative prediction of synthetic food colors by fluorescence spectroscopy and radial basis function neural networks is introduced. Taking Amaranth as an example, for sample solution of Amaranth with different concentrations, the fluorescence spectroscopy excited by the light with the wavelength of 400 nm is measured. The peak wavelength of Amaranth solution fluorescence spectroscopy is about 640 nm, and a non-linear relationship is observed between fluorescence relative intensity and concentration of solution. For each sample, 10 emission wavelength values were selected. The fluorescence intensity corresponding to the selected wavelength was used as the network characteristic parameters, a radial basis function neural network was trained and constructed. The trained radial basis function neural network was employed to predict the three kinds of samples Amaranth solution concentration .The relative error of prediction were 4.60%, 4.71%, 5.94%. The results show that the method is convenient, fast, of high accuracy. It is a new technique for the detection of synthetic food color in food safety supervision and management.