Uso da calibração multivariada para a predição de propriedades físico-químicas de misturas de óleo de soja e biodiesel

• Ano de defesa: 2015
• Autor(a) principal: Silva, Juliana Verdan da
• Orientador(a): Araujo, Lucia Regina Raddi de
• Banca de defesa: Scofield, Cynthia Fraga , Langone, Marta Antunes Pereira , Portilho, Marcio de Figueiredo
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade do Estado do Rio de Janeiro
• Programa de Pós-Graduação: Programa de Pós-Graduação em Engenharia Química
• Departamento: Centro de Tecnologia e Ciências::Instituto de Química
• País: BR
• Palavras-chave em Português: Calibração multivariada; mistura biodiesel-óleo de soja; espectros MIR-ATR; massa específica; índice de refração
• Palavras-chave em Inglês: Multivariate calibration; mixing soybean biodiesel-oil; MIR-ATR spectra; density; refractive index
• Área do conhecimento CNPq: CNPQ::ENGENHARIAS::ENGENHARIA QUIMICA::TECNOLOGIA QUIMICA
• Link de acesso: http://www.bdtd.uerj.br/handle/1/12016

In the present work, a model that uses multivariate regression tools was proposed to predict both contents of methyl esters and physical-chemical properties of soybean oil -soybean biodiesel blends. The model was proposed from the correlation of the properties of interest with the attenuated total reflectance infrared spectra of the samples (ATR/MID-FTIR). The composition of the blends (methyl esters content) was determined by high performance liquid chromatography (HPLC), which can be seen as an alternative technique to the standard reference methods based on gas chromatography (EN 14103 and EN 14105). The selected physicochemical properties were refractive index, density, and viscosity. For the study, 11 mixtures were prepared with different proportions of soybean biodiesel and soybean oil (0-100 % wt of soybean biodiesel) in five replications totalizing 55 samples. The infrared spectra were acquired in the range 3801-650 cm-1. The spectra were submitted to the multiplicative signal correction (MSC) and then to mean centering (MC) preprocessing. The properties of interest were submitted to auto scale. It was then applied principal component analysis (PCA) for the purpose of reducing the dimensionality of the data, and detected the presence of outliers. When the outliers were detected, the samples were discarded. The original data were submitted to Kennard-Stone algorithm dividing them into a calibration set to build the model and the validation to verify its reliability. The results showed that the model proposed by PLS2 (Partial Least Squares) was able to adjust well to the refractive index data and density, can be observed a random behavior of errors, indicating the presence of homoscedasticity in residual values, in other words, the constructed model presented a forecast of capacity for specific mass properties and refractive index with 95% confidence. The accuracy of the model was evaluated by estimating the regression parameters which are the slope and the intercept by EJCR (Joint Region Confidence Ellipse). The results confirmed that MIR-developed PLS model was able to predict both the refractive index and specific gravity properties. For contents of methyl esters via liquid chromatography the model showed a lower adjustment of the data. Nevertheless, it was possible to demonstrate that the results obtained from methyl esters by HPLC analytical method can be modeled and correlate them with infrared spectra using multivariate calibration