Estimation of physicochemical parameters of Brazilian commercial gasoline using gas chromatography and partial least square regression

Abstract

The quality of brazilian gasoline is measured by specification tests routinely conducted by monitoring laboratories established by the Government Petroleum Agency (ANP). These tests are usually expensive and time consuming. Thus, their replacement by correlation methods using gas chromatography (GC) is very attractive, particularly if the number of samples is very large. In this way the gasoline properties can be calculated from a single GC analysis and a partial least square (PLS) model. With the modern instrumentation the chromatographic profile can be easily and quickly obtained with high accuracy and related directly to specification parameters. By investigating composition-property relationships, a single GC-FID experiment and a PLS model can be used to estimate several properties of gasoline. In this work, the properties investigated were relative density, distillation curve, octane numbers, and hydrocarbon composition. The results indicated that these parameters can be reasonably predicted in minutes using PLS correlations to GC data.