Determination of the physical properties of gasoline mixtures is a difficult and usually costly undertaking both economically and in the time it takes to analyze them. Fuels have a range of possible compositions consisting of a complex mixture of hydrocarbons and fuel additives. This makes the determination of specific properties difficult to accomplish. Raman spectroscopy is a useful tool in analyzing the composition of gasoline and other petroleum products. A Raman spectrum is able to be taken quickly and does not consume the sample in the process. Raman spectra are able to clearly detect highly symmetric, polarizable molecules, which makes up the majority of the species of molecules found in gasoline. Spectra were taken of one hundred-seventy gasoline samples using five lasers: 1064, 830, 785, 638 and 532 nm. Raman excitation wavelengths can excite different vibrational modes not seen at other wavelengths. The higher energy wavelengths can cause the sample to fluoresce adding variance to the data set that is not representative of Raman excitation. Lower energy wavelengths are capable of degrading the samples which can add artificial noise. The artificial variance introduced by these phenomena makes normalization of the spectra difficult. The samples originated from twenty countries spanning six continents making this one of the most comprehensive global fuel sample sets studied to date. The spectra collected were analyzed by principal component analysis (PCA) and soft independent model classification analogy (SIMCA) to find correlations between the spectra taken using the five excitation wavelengths and the octane numbers of the fuel samples. The Mahalanobis distances of the data were calculated and used to identify potential outliers from the bulk of the data set. The models are compared to those created through the use of multi-bounce ATR using the same training sets.

Multivariate analysis of global fuel spectral data sets

2019 Spring National Meeting

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.