Studies On Quality Detection Of Alcohols-gasoline By Near Infrared And Raman Spectroscopy

Automobile industry development put more and more pressure on energy, environment,traffic safety, and so on. Development of energy substitute for petroleum has become a topgovernment priority. The development of alcohols-gasoline as a substitute for automobilefossil fuels had received great attention in many countries. The development of automobiletechnology and gasoline quality are closely correlated, so it’s extremely important to detectalcohols-gasoline quality. With the methanol gasoline and ethanol gasoline as the objects inthis research, near-infrared and Raman spectroscopy and chemometric analysis were used tocarry out the quality detection of alcohol content, density and viscosity of methanol gasolineand ethanol gasoline. The main purpose of this study is to provide reference for alcohol-gasoline quality fast detection in china. Mainly research results were as follows:1. With near-infrared (NIR) spectroscopy and Partial Least Squares(PLS), it wasdiscussed the influence for quantitative model by different spectral pretreatment methods anddifferent effective wavelength optimization.The study results were involved: the spectra ofalcohols-gasoline pretreated by the optimal method were selected effective wavelength for theconstruction of the calibration models by Interval Partial Least Squares (iPLS) and the resultswere generally better.2. NIR quantitative models of alcohol content, density and viscosity of methanolgasoline and ethanol gasoline were established by PLS and LS-SVM respectively. Theoptimization models were evaluated by predicting the prediction set. The results show thatbetter results both were achieved by LS-SVM model for methanol content, density andviscosity with the correlation coefficient(r) of0.992,0.981,0.962and the root mean squareerror of prediction (RMSEP) of0.1%(v/v),0.29g/L,0.026mPa s respectively. Similarly thebest results both were achieved by LS-SVM model for ethanol content, density and viscositywith the r of0.985,0.972,0.962and RMSEP of0.15%(v/v),0.45g/L,0.028mPa srespectively.3. With Raman spectroscopy and Partial Least Squares(PLS), it was discussed theinfluence for quantitative model by different spectral pretreatment methods. Raman quantitative models of alcohol content, density and viscosity of methanol gasoline and ethanolgasoline were established by PCR, PLS and LS-SVM respectively. The optimization modelswere evaluated by predicting the prediction set. The results showed that LS-SVM model bothwas better than the other (PCR and PLS) for methanol content, density and viscosity. Thecorrelation coefficients(r) for methanol content, density and viscosity were0.985,0.984,0.979respectively, while the RMSEP were2.57%(v/v),1g/L,0.064mPa s respectively. TheLS-SVM model both was better than PCR and PLS for ethanol content, density and viscosity.The correlation coefficients(r) for ethanol content, density and viscosity were0.98,0.971,0.941respectively, while the RMSEP were3.22%(v/v),1.1g/L,0.079mPa s respectively.The above results verify the feasibility of quality detection of alcohols-gasoline by nearinfrared and Raman spectroscopy. The results of two kinds of technology are compared, thenear infrared spectroscopy has a high precision and accuracy and portable Raman technologyhas fast and site test advantages.