Study On Spectral Response Characteristics And Rapid Detection Method Of Diesel Quality Information

Diesel is one of the most widely used petroleum products,which has many advantages including high thermal efficiency,good power performance.But the existing diesel quality detection method,which is slow,high cost,complex process,hinders the domestic diesel refining technological progress.And the emission of diesel particles with poor quality seriously pollutes the urban environment.Due to serious coupling of hydrocarbon groups in diesel related,highly nonlinear correlation of physical and spectral data,weak spectral absorption,overlapping of characteristic peaks and difficulty in extracting quality information,it greatly affects the stability and universality of the diesel spectral detection model,limiting the popularization and application of on-line detection.The main research results and innovation points can be concluded as follows:(1)Predicted model of diesel cetane number and colloid content based on variable weighted fusion is studied.The spectral response characteristics of hydrocarbon groups and physical parameters associated with diesel is analyzed.The weighting model of characterization factor modified by contribution is established based on nonlinear modeling analysis method of diesel quality group response spectrum and characteristic spectral variable fusion weighted.Then the group characteristics and key characteristics of spectral spectrum weighted fusion are input to nonlinear prediction model.Compared with the traditional model,the accuracy of predicted model of diesel cetane number and colloid content improves significantly.For cetane number,decision coefficient(R~2)is improved from 0.745 to 0.927,for colloid content,prediction(R~2)is increased from 0.610 to 0.821.(2)Non-linear model of least squares support vector machine(LS-SVM)prediction for kinematic viscosity and sulfur content of diesel is studied based on Latent Variables(LVs)and Random Frog(RF)characteristic variable selection,to solve the problem of kinematic viscosity and sulfur content of diesel with near infrared spectrum data of highly nonlinear,and find the accurate correlation peak wavelength of the ownership.The parameters,R~2 and RMSEP,of LVs-LS-SVM and RF-LS-SVM evaluation model were 0.8359,0.317 and 0.97,0.0023 respectively.The number of variables selection are 14,only 3.49%of the original variables,which greatly simplify the model,providing the promising application for the rapid development of online measuring instrument.Grid search method is used to determine the nonlinear kernel function parameters of LS-SVM,the hydrocarbon groups,property index and spectral data of nonlinear coupling relationship is decomposed into linear problem at high altitude space,improving the prediction precision.(3)The classification model of methanol and ethanol diesel with principal component analysisis studied.Limit learning machine(PCA-ELM)and the alcohol content quantitative prediction model with competitive adaptive heavy weight continuous projection and limit learning machine joint algorithm(CARS-SPA-ELM)based on mid infrared spectroscopy is built.When the alcohols are dissolved in diesel,the peak shifts and the NIR spectra overlap seriously.PCA-ELM identified two types of alcohol diesel oil with a precision of 100%.For methanol diesel,alcohol content prediction parameters,R~2 and RMSEP,are 0.9769,1.2450 respectively.For ethanol diesel,alcohol content prediction parameters R~2,RMSEP are 0.9877,1.1906respectively.(4)Measuring instrument for near infrared spectrum of diesel quality is developed,which is composed of STS spectrometer as the core.The influence of the performance parameters of the core spectrometer on the analysis of diesel spectral detection was analyzed.The parameters of prediction model for diesel cetane number and colloid content,R~2 and RMSEP,are 0.902,0.934 and 3.57%,3.42%.To summarize,the response spectral characteristics of diesel quality were analyzed,and a characteristic wavelength selection model was constructed based on the weighting of the characterization factor modified by the contribution degree.The quantitative analysis model of diesel quality parameters and the qualitative discriminant of methyl alcohol was established,and a rapid quantitative analysis instrument based on characteristic wavelengths for diesel quality spectrum was developed.The cetane and sulfur content were verified by the development instrument,which provided a reference basis for the on-line analysis and application of diesel quality.