Classification of gasoline by octane number and light gas condensate fractions by origin with using dielectric or gas-chromatographic data and chemometrics tools

The approach for classification of gasoline by octane number and light gas condensate fractions by origin with using dielectric permeability data has been proposed and compared with classification of same samples on the basis of gas-chromatographic data. The precision of dielectric permeability measurements was investigated by using ANOVA. The relative standard deviation of dielectric permeability was in the range from 0.3 to 0.5% for the range of dielectric permeability from 1.8 to 4.4. The application of exploratory chemometrics tools (cluster analysis and principal component analysis) allow to explicitly differentiate the gasoline and light gas condensate fractions into groups of samples related to specific octane number or origin. The neural networks allow to perfectly classifying the gasoline and light gas condensate fractions.