Uso de modelos multivariados para a avaliação do valor de dienos em amostras de gasolina de pirólise

Abstract

The quality of gasoline may be affected by the presence of several compounds. Among them, the conjugated dienes play an important role due to the fact that their presence is strongly associated with gum formation, which may clogs the injector of automobiles and may causes damages in the petroleum derivates. The main used method for the determination of total conjugated dienes is the UOP-326 method. However, the long analysis time (from 5 to 6 hours) is the main disadvantages. The available methodologies investigated as alternative to the UOP-326 method do not solve the problem properly. The present work reports the use of multivariate models to assess the diene value (DV) in pyrolysis gasoline samples starting from voltammetric and spectrophotometric data. The calibrations were carried out using the UOP-326 method instead of diene standards. In the voltammetric model, starting from a group of 24 hydrogenated pyrolysis gasoline, the PLS (Partial Least Squares) method was used to predict the DV in an independent group of 7 samples. The deviations observed were lower than 12,2% comparing to the UOP-326 method. In a group of 24 non-hydrogenated pyrolysis gasoline samples, an independent group of 7 samples was also used to predict the DV. The deviations observed were lower than 4,1%. In the spectrophotometric approach, the PLS method was used to predict the DV in an independent group of 5 samples of hydrogenated pyrolysis gasoline (n=21). The deviations observed were lower than 11% by comparing to the UOP-326 method. In a group of 26 non-hydrogenated pyrolysis gasoline samples, an independent group of 10 samples was used to predict the DV. The deviations observed were lower than 5,7%. Comparing the voltammetric and spectrophotometric methods, the former showed more reliable results with lower RMSEP (Root Mean Square Error of Prediction) values.