Research On Quantitative Analysis Method Of Emulsified Oil Spill Based On LIF Fluorescence Spectrum

At present,oil spill pollution on the sea surface has become the most serious problem in marine pollution,which not only threatens the marine ecological environment,but also seriously affects people’s health.Laser-induced fluorescence technology is highly sensitive,convenient,flexible,fast and non-destructive.It occupies an important position in the detection of oil spills on the sea surface.However,the research on the detection of oil spills on the sea surface using LIF technology mainly focuses on the inversion of oil film thickness and the classification and identification of oil species,and there are relatively few studies on the quantitative analysis of emulsified oil spills.Therefore,based on the fluorescence spectrum and combined with the regression analysis model,this thesis realizes the efficient and accurate quantitative analysis of the emulsified oil spill.The main research contents are as follows.First,a LIF measurement system was built in the laboratory,taking0~#diesel and3~#kerosene as the research objects,preparing emulsified diesel and emulsified kerosene with different oil contents,and collecting their fluorescence spectra;Collect its fluorescence spectrum.Secondly,the quantitative analysis of emulsified diesel oil and emulsified kerosene is realized.Based on the least squares support vector machine regression model,the parameters of the LSSVM model are optimized by using the hybrid algorithm(GAPSO)combining genetic algorithm,genetic algorithm and particle swarm algorithm.,the wavelength with high information content was screened out by the continuous projection method as the model input,and the oil content was used as the model output,and the GA-LSSVM and GAPSO-LSSVM regression models were established respectively.According to the determination coefficient and root mean square difference and other evaluation criteria,and compared with the SVM,LSSVM and PSO-LSSVM models,verify the performance of the optimized LSSVM model.The results show that the GA-LSSVM and GAPSO-LSSVM regression models have better performance,and both have good prediction accuracy.Finally,the quantitative analysis and research of diesel oil and kerosene oil-in-water emulsions are realized.The radial basis function network(RBFN)algorithm is used to accurately separate the spectral components of the Raman signal of seawater from the original spectral signal,and calculate the Raman peak intensity,Raman spectral parameters such as peak area and half width.Based on this,a multivariate nonlinear regression model of oil-in-water emulsion concentration was established with and without the interaction of each variable,and compared with the multivariate linear regression model.The results show that the multivariate nonlinear model considering the interaction of the respective variables has high accuracy and small error,and can realize the quantitative analysis of oil-in-water emulsions.