In order to detect the concentration accurately, Savitzky-Golay polynomial smoothing method is proposed to remove the redundant information of the three-dimensional fluorescence spectral data. The parallel factor analysis(PARAFAC) and alternating penalty trilinear decomposition(APTLD) are respectively adopted to decompose the spectrum data. The detection experiment of polycyclic aromatic hydrocarbons is designed in this paper. Fluorescence spectrum properties of ANA, NAP and their mixture are analyzed. We found that the main characteristic florescence peak of FLU exists in λex/λem=302/322 nm, and there is a continuous side peaks. The two characteristic florescence peaks which exist in ANA are λex/λem=290/322 nm and λex/λem=290/336 nm. When the excitation wavelength is 200~370 nm and the emission wavelength is 240~390 nm, the fluorescence spectra of FLU and ANA overlap seriously.The experimental results show that both algorithms can all distinguish FLU and ANA, both of them have a high average recovery rate. However, the APTLD algorithm has a better detection result. © 2019, Science Press. All right reserved.
