Research of Identification Method for the Oil Spills Species Based on Fluorescence Excitation-Emission Matrix and Parallel Factor Analysis
Zhou Yan-lei,Zhou Fei-fei,Jiang Cong-cong,Shi Xiao-yong,Su Rong-guo
DOI: https://doi.org/10.3964/j.issn.1000-0593(2018)02-0475-06
2018-01-01
Spectroscopy and spectral analysis
Abstract:Accidental oil spills occur frequently around the world and they can have serious influence on the human health and the ecosystem due to the compounds in oil. Thus, there is an urgent need for an accurate method for determining the source of spills. In order to meet the rapidly demand identifying methods of oil spills, this paper utilized parallel factor analysis to develop an identification method for the crude oils and fuel oils based on their fluorescence excitation-emission matrixes. Firstly, the fluorescence excitation-emission matrixes of six kinds of crude oils (Roncador crude oil, Basra crude oil, Russian crude oil, Saudi crude oil (heavy crude oil), Upper Zakum crude oil, sea two station crude oil) and three kinds of fuel oils (380 CST fuel oil, Fuel oils No. 5-No. 7, LanShan fuel oil) were normalized after removing scatter by Blannany triangulation with linear interpolation. Afterwards, the parallel factor analysis was used for the fluorescence excitation-emission matrixes. Seven factors can be reliably extracted from the data set, and then seven fluorescence components were obtained, which made up the characteristic spectrum. By clustering analysis and Bayesian Discrimination of the characteristic spectrum from the samples that had experienced weathered for 3, 15 and 45 days and not experienced weathered, the capacity of the oil fluorescence spectrum analysis was determined and the fluorescence feature spectra library consisted of 12 standard fluorescence spectra of crude oils and 6 standard fluorescence spectra of fuel oils was established. In the end, multiple linear regression was used to recognize the samples that had experienced weathered for 0, 7 and 30 days by the oil standard fluorescence spectra. The results showed that, except for Russian crude oil, the method could clearly classify 5 kinds of crude oils and 3 kinds of fuel oils which had experienced weathered and had not experienced. Furthermore, the accuracy of the identification for crude oils and fuel oils were 100. 0%, and the overall accuracy of the identification for crude oils and fuel oils were 87. 5% and 100%, respectively.