Abstract: In order to achieve the rapid field detection of gasoline brand and research octane number (RON), the spectral signals of 113 gasoline samples were collected by a portable Raman spectrometer. Then the gasoline brand model was respectively established by principal component analysis and partial least squares discriminant analysis, and the gasoline RON model was established by partial least squares method. The results show that the gasoline brand model based on the baseline-corrected spectral data treated by principal component analysis and derivation, the accuracy of sample classification can reach 92.92%, while the positive rate of the PLS-DA model is above 95%, which is better for distinguishing 92# and 95# gasoline. The gasoline RON rapid prediction model based on partial least squares, the prediction set correlation coefficient is 0.8927, and the root mean square error of prediction is 0.6096, which shows that the predicted value has a good correlation with the actual value.

Key words: gasoline, Raman spectrum, principal component analysis, partial least squares method, research octane number