Spectral analysis and optimization of time-resolved oil fluorescence based on PCA

Laser-induced Fluorescence (LIF) technique can be widely used in oil pollution monitoring. However, ordinary oil fluorescence spectra can only achieve cursory oil classification, which was disabled to distinguish crude and fuel oils. Herein, time-resolved fluorescence spectra classification method based on Principle Component Analysis (PCA) was investigated and employed to analyze the spectral features of 20 kinds of oils, of which the fluorescence lifetimes and the spectral timing characteristics were obtained. Then referring to fluorescence lifetimes of oils (less than 10 ns commonly), three-dimensional spectra of samples within this time range were used for obtaining a vector space which was composed of first three principal components and was considered as a three-dimensional coordinate system. In this coordinate system, correlation distances of position vectors at difference delay time of fluorescence acquisition were analyzed for spectral clustering of time-resolved oil fluorescence. To reflect timing characteristics of correlation distances, dispersion parameters were introduced into the PCA optimization method. The experimental result indicates that the method based on time-resolved fluorescence spectroscopy can discriminate between crude oils and fuel oils with a higher recognition rate. © 2017, Science Press. All right reserved.