Uso de espectroscopia no Infravermelho Médio e análise discriminante por quadrados mínimos parciais no controle de qualidade de biodiesel de Pinhão Manso e Crambe

• Autor(a) principal: Buiatte, José Eduardo
• Data de Publicação: 2019
• Tipo de documento: Tese
• Idioma: por
• Título da fonte: Repositório Institucional da UFU
• Texto Completo: https://repositorio.ufu.br/handle/123456789/27927; http://dx.doi.org/10.14393/ufu.te.2019.2541
• topic: Biocombustível; Biofuel; FT-MIR; PLS-DA; Crambe abissynica H.; Jartropha curcas L.; Quimiometria; Chemometrics; CNPQ::CIENCIAS EXATAS E DA TERRA

Biodiesel is a renewable fuel derived from vegetable oils or animal fats that has stood out in the Brazilian energy matrix and in several countries as an energy option and counter points the dependence on fossil fuels. The biodiesel was included into the Brazilian energy matrix in 2005.It is usually marketed in mixture with the mineral diesel in proportions legally established by the National Agency of Petroleum, Natural Gas and Biofuels (ANP). Although there are regulatory and oversight organizations that assure the quality of marketed fuels, certain forms of adulterations and contaminations that affect the quality of this fuel do sometimes occur. In this work the use of medium infrared spectroscopy (MIR)data analyzed with the chemometric method of Discriminating Analysis by Partial Least Squares discriminating analysis, according to the partial least squares (PLS-DA) criterion, was used in an attempt to develop an efficient method to classify biodiesels derived from oils from mansonia pinus and Crambe in mixtures with the diesel, both in relation to the biodiesel in the mixture with the diesel along with common adulterants as like: soybean oil, waste oil, used automotive oil and gasoline. Between 58 and 60 samples were used to check the models, with contents varying between 1.00 and 30.00 mass% biodiesel and adulterants. The test sets, that is, for the validation of the model, were constructed with approximately 30 samples each. The efficiency of these models was assessed on the basis of figures of merit, sensitivity and specificity. The threshold values were chosen based on the Bayes', whereby the ypredicted values result in a similar distribution to the values that will be observed for future samples, theorem, in order to minimize the occurrence of false positives and false negatives for the classified samples. The relation between specificity and sensitivity were also evaluated graphically using the receiver operating characteristic (ROC) curves. These parameters presented values ranging from 0.93 to 1.00, which represents correct classification of the calibration and forecast samples above 93.30%. From the good results of these models, it was prove it that, in fact, the analytical method is indeed reliable and that the qualitative control of these fuels is efficient enough.