Quimiometria associada a técnica espectroscópica RMN aplicada na classificação de múltiplas propriedades físico-químicas do petróleo

Abstract

Chemometrics associated with spectroscopic techniques are reported in the literature as great tools to study petroleum due to the speed at which a spectrum can be acquired and the speed at which a model can be built, but there is still a shortage in the literature regarding classification models. 147 oil samples (containing an NMR spectrum) were classified according to seven physicochemical properties: sulfur content, total nitrogen content, API grade, UOPk factor, TAN, SARA and pour point, which were previously divided into the appropriate classes. PCA was applied in order to reduce the dimension and observe clusters, and then PLS-DA was applied in order to classify the samples into classes. Variable selection methods such as iPLS and UVE-PLS were also applied to select the region, or eliminate, of variables that best capture the variance for classification. The results indicate that NMR associated with chemometrics was able to classify the samples. The PCA indicated agglomerations between the samples and the iPLS-DA was responsible for the most accurate models. For nitrogen content, SARA, sulfur content, UOPk factor this is the first report of classification models and for TAN the model built in this work proved to be superior to the literature.