A multivariate statistical analysis of chemical composition and physical characteristics of ten diesel fuels

Abstract

Ten different diesel fuels were tested for 26 physical and 45 chemical parameters. Principal components analysis (PCA) was used for statistical analysis of the physical and chemical data separately and combined. Physically, these and similar fuels may be described in a few general terms, such as viscosity, boiling residue, cetane number and density. With three PCs, 88% of the variation in physical terms was accounted for. Only one fuel remained poorly described by the model. The most descriptive chemical components include 1-methylphenanthrene as representative of a major fraction of the polycyclic aromatic hydrocarbons (PAH). With three PCs, all fuels were well accounted for, with 80% of the variation in chemical parameters explained. Although PCA does not provide a means of deriving causal relations, it may be used to find patterns of intercorrelated variables. With both chemical and physical parameters included in a final PCA, relations between chemical contents and physical characteristics were observed for e.g. PAH/olefins/naphthenes and viscosity.