Pesticide residue detection by fluorescence spectral analysis based on BP neural network

At present, it is a difficulty to achieve rapid and accurate detection of pesticide residues. In the paper, an overlapped spectrum in fluorescence spectroscopy measurements of acetamiprid residues was separated using artificial neural networks,and a fluorescence spectrum measaring system of solid surface acetamiprid residues was designed. By means of artificial neural network principle and back-propagation training algorithm, acetamiprid concentration was determined in mixed components of residues and filter paper with overlapped fluorescence spectrum. In the range of 340nm～400nm, the fluorescence intensities corresponding to 20 wavelengths were used as character parameters, and the neural network was trained and tested. The mean recoveries of 40mg/kg and 90mg/kg acetamiprid were 102% and 97% respectively. The RSDs of the results were 1.4% and 1.9%. The results have shown that the method to using BP network in fluorescence spectral analysis of acetamiprid residue has some advangtages such as shorter measurement cycle, faster training speed and higher accuracy.