Gasoline Properties Estimation from Hydrocarbon Clustering Classification

A two-stage approach to properties prediction of gasoline blends is proposed. The first stage is a clustering classification of the hydrocarbon range from C1 to C12, using the physical properties and quantifiable molecular characteristics of each compound as classification variables. This classification divides each of the PONA families into several groups.

In the second stage a sample of Venezuelan gasoline blends is used to build regression models with variables resulting from the clustering classification. The sample analyzed includes different gasolines within a wide composition range, lead antiknock additives and oxygenated compounds.