Detection Of Synthetic Red Food Colors Using Fluorescence Spectrometry Based On Neural Network

In recent years, food security incidents have been constantly emerging, which caused the food safety supervision to be strengthened and effective detection methods to be established all over the world. Synthetic food colors are a class of food additives which are widely used in food production. Ultraviolet spectrophotometry, high performance liquid chromatography and derivative adsorption voltammetry have been used to detect synthetic food colors.Using fluorescence spectrometry to detect synthetic food colors qualitatively and quantitatively was rarely reported at home and abroad, which has the characteristics of high sensitivity, simplicity of operation, little dosage, easy sampling, high selectivity, rapid analysis, high resolution and non-destructiveness. Fluorescence intensity is closely related to the sample concentration, but the relationship between them is complex and nonlinear. Some mathematic models must be introduced to help the analysis. Artificial neural network, which mimics the structure and function of human brain to process information, is a effective method to process complex spectral information.At present, 10 kinds of synthetic colors could be used legally in food production in China, among them there are two yellow colors sunset yellow and tartrazine; two blue colors brilliant blue and indigotine; six red colors ponceau 4R, amaranth, allura red, erythrosine, acid red and new red. The study of fluorescence spectra of synthetic red food colors has not been systematized.The fluorescence spectra of synthetic red food colors were measured by using Sp-2558 multi-function spectral measurement system. The results showed that these red colors could all emit strong fluorescence induce by UV-light. The ranges of fluorescence were all wide and the spectral characteristics were clear. Meanwhile, they had different fluorescent ranges and peaks.The species identification and concentration determination of synthetic red food colors have been carried out based on artificial neural network and the data compressed by using wavelet transformation. By comparing the different neural network models, probabilistic neural network was chosen to identify the six kinds of food colors qualitatively, the average recognition rate reached 95.80%. And radial basis function neural network was chosen to determine the concentrations of new red solutions quantitatively, the average determination error for four kinds was 2.14%. This method has combined all the advantages that fluorescent spectrometry is rapid, simple and high accurate; wavelet transform is convenient to extract characteristic information and compress data; and artificial neural network is suitable for nonlinear system detection.The research provided a novel method for qualitative and quantitative analysis of synthetic red food colors. Although only synthetic red food colors were taken as research objects, the results could be used in other aspects of food safety testing. For example, we have applied this method to determination of melamine in milk and recognition of alcohol-free beer, and the results turned out to be effective.