Monitoramento da estabilidade oxidativa de biodiesel empregando espectroscopia vibracional associada a ferramentas quimiométricas
Ano de defesa: | 2022 |
---|---|
Autor(a) principal: | |
Orientador(a): | |
Banca de defesa: | |
Tipo de documento: | Dissertação |
Tipo de acesso: | Acesso aberto |
Idioma: | por |
Instituição de defesa: |
Universidade Federal de Minas Gerais
Brasil ICX - DEPARTAMENTO DE QUÍMICA Programa de Pós-Graduação em Química UFMG |
Programa de Pós-Graduação: |
Não Informado pela instituição
|
Departamento: |
Não Informado pela instituição
|
País: |
Não Informado pela instituição
|
Palavras-chave em Português: | |
Link de acesso: | http://hdl.handle.net/1843/50680 |
Resumo: | Petroleum-derived fuels are the primary source of energy worldwide, nevertheless, the growing environmental awareness makes it necessary to establish new renewable sources of fuel, cheaper to produce and environmentally friendly. Biodiesel gains emphasis among these sources, as it has been blended with mineral diesel. One of the biggest problems in using of biodiesel is the formation of solids from its oxidation, which is increasingly common as the biodiesel content increases in the blend. Currently, there are standardized methods for fuel analysis, however, such methods are expensive. Thus, it is urgently necessary to develop new analysis methodologies that are simpler, more robust, portable, faster, and less costly, allowing for monitoring of the production chain. In this context, the objective of the work was to develop classification models to characterize pure biodiesel in two categories: compliant and non-compliant, regarding oxidative stability, employing spectroscopic techniques (infrared and Raman). Samples were analyzed by attenuated total reflectance mid-infrared spectroscopy (ATR-FTIR), near-infrared spectroscopy (NIR) with portable equipment, and Raman spectroscopy. Two types of models were built and evaluated, a linear model employing linear discriminant analysis (LDA) and a non-linear model using the Random Forest method. In addition, a resampling strategy employing the Adasyn method was used to correct for class unbalance. The performance of the models was evaluated by the confusion matrix and the parameters efficiency and Matthews correlation coefficient were calculated. The models here constructed were able to classify biodiesel into compliant and non-compliant, and the Matthews correlation coefficient was greater than 0.8 for all models. The ATR-FTIR technique was the most promising for linear models and Raman spectroscopy for non-linear methods, both with Matthews correlation coefficient of 0.97. Finally, the use of spectroscopic techniques associated with machine learning methods for the classification of biodiesel, according to ANP standards, for oxidative stability is promising, allowing the inspection of B100 biodiesel samples directly, without any preparation, quickly and with on-site detection using portable equipment, commercially available for the three techniques employed in this work. |