Determinação do teor de biodiesel de Mafurra e Crambe em misturas com diesel por RMN de 1H e regressão multivariada por OPLS E OPLS-DA

Abstract

The production and the use of biodiesel is being increasingly strengthened in several countries due to the possibility of a shortage of fossil fuels in the medium or long term and the need to reduce the environmental impacts caused by its production process. Brazil is one of the examples of the countries that most produce and use biodiesel, having reached 5,350,036 m³ of biodiesel in 2018 against 4,291,294 m³ of the previous year, thus favoring the evolution of its percentage added to diesel. This evolution represents concerns in relation to its commercialization with content that is not what is required by the standards. For this reason, in this work, four methods were developed to monitor the B10 content of mafurra and crambe biodiesel in mixtures with diesel using hydrogen nuclear magnetic resonance spectroscopy (1H NMR) combined with the multivariate regression by orthogonal projections to the latent structure (OPLS ) and orthogonal projections on the latent structure-discrimination analysis (OPLS-DA). The efficiency of the methods developed based on the multivariate regression by OPLS was analyzed based on the figures of merit and the fit of the models through the correlation of the measured and predicted values of the calibration and forecast sets. The results of the figures of merit were in agreement with the requirements established in the standard ASTM E1655-05. A high correlation between measured and predicted values was evident in all OPLS models, with a correlation coefficient (R2) greater than 0.99. The efficiency of the methods developed based on multivariate regression by OPLS-DA was analyzed based on the parameters of false positive and false negative rates, sensitivity, specificity and Matthew's correlation coefficient, where the presence of false positive and false negative samples was not noticed, consequently, the parameters of sensitivity, specificity and Matthew's correlation coefficient were equal to 1, which means that the models presented a 100% correct classification of the B10 samples (10% biodiesel and 90% pure diesel) and BX (biodiesel content less and greater than B10) used in the training and test sets. The high efficiency demonstrated by the OPLS and OPLS-DA models in monitoring the B10 content of biodiesel from mafurra and crambe mixed with diesel, suggests that the analytical methods developed are ideal, efficient and suitable for use by quality control inspection agencies of that fuel.