Remote fiber optic Raman analysis of benzene, toulene, and ethylbenzene in mock petroleum fuels using partial least squares regression analysis

Remote fiber optic Raman spectroscopy utilizing CCD detection and near IR diode laser excitation is used to analyze mock petroleum fuel samples with individual benzene, toulene, and ethylbenzene (BTE) concentrations ranging from 0 to 16%. Specific Raman spectral regions which contain BTE vibrational modes, as well as the entire spectrum, are correlated to the known concentrations of BTE in the samples by Partial Least Squares (PLS) regression analysis. The error associated with the models using only specific spectral regions is less than those using the entire spectral regions in the regression analysis. The use of specific spectral regions minimizes both the time of the regression analysis and the amount of noise associated within the most highly correlated regression model factors; thus over fitting of the data is avoided and the reliability of subsequent predictions is improved.