The Component Analysis And Research Of The Mixture Of Oils Based On The Three Dimensional Fluorescence Spectrum

With the development of the society,the demand for oil is increasing,but a large number of petroleum products are released into the environment in the process of oil prospection,exploration,transportation and process.Water pollution caused by oil is a serious threat to human survival environment,so the qualitative and quantitative detection of oil pollutant in water is of great significance to the problem of environmental monitoring and management.By analyzing photoluminescence,fluorescence spectroscopy is pionted out in this paper.The characteristics of fluorescent material,such as the fluorescence spectrum,the efficiency of fluorescence and the external factors affecting the fluorescence,are analyzed.This paper discusses the composition of petroleum pollutants,and proves the feasibility by using fluorescence method to detect petroleum substances.With diesel,gasoline and kerosene acting as examples,the excitation and emission scan range is selected considering of substances composition.Three dimensional fluorescence spectrum and projection spectra of each dimension of samples are measured and spectral features are analyzed.The multivariate calibration analysis theory is researched and the second-order correction advantage is pointed out.The calibration model of samples are set up and the concentration of diesel oil,gasoline and kerosene in the mixture are accurately predicted by using the parallel factor analysis and partial least-squares respectively.Gasoline and kerosene are set as the calibration substance,diesel is set as the interfering substance to make up samples.The Savitzky-Golay polynomial is used to expand second-order data to fourth-order,and the five-way parallel factor analysis and unfolded partial least-squares are designed to deal with five-way data.Results show that fourth-order data has a more comprehensive sample information than the second-order data.It is more stable for the interference and matrix effects.Under the condition of reaction,it is not easy to change,and the prediction accuracy is improved.Unfolded partial least-squares has more superiority than the parallel factor analysis.