Desenvolvimento, validação, avaliação da incerteza de medição e análise de custos de método baseado em espectroscopia no infravermelho e análise multivariada para previsão de parâmetros de qualidade de amostras de E85

• Ano de defesa: 2015
• Autor(a) principal: Hatanaka, Rafael Rodrigues [UNESP]
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual Paulista (Unesp)
• Palavras-chave em Português: Espectroscopia de infravermelho; Incerteza de medição (Estatística); Quimiometria; Estudos de validação; Análise espectral; Chemometrics
• Link de acesso: http://hdl.handle.net/11449/126560

The use of ethanol as fuel can reduce emissions of greenhouse gases and dependence on fossil fuels, as it can be obtained from renewable sources. However, this biofuel can present problems of ignition and driving in areas with cold climates due to lower vapor pressure in relation to gasoline. To avoid such problems, ethanol can be mixed with in various proportions (50-85% in Europe and the United States). These mixtures are commercially known as E85. Other parameters (such as methanol and water content) are also important to ensure the quality of the fuel and are fixed by international standards. Thus, the present study shows the development, validation and estimation of the uncertainty of analytical methods using spectroscopy in the infrared region (middle and near) and multivariate regression, which allow quantifying levels of ethanol, methanol, oil and water in E85 samples prepared in laboratory. The validation was made by calculating the figures of merit using the concept of net analytical signal (NAS). Thus, the proposed methods were able to quantify the levels of ethanol, methanol, oil and water linearly, free of systematic errors, with accuracy of 0.2% (m/m), 0.03% (m/m), 0.2% (m/m) and 0.03% (m/m), accuracy of 0.3% (m/m), 0.02% (m/m), 0.2% (m/m) and 0.02% (m/m), respectively, and measuring intervals covering whole range of parameters established by the standards. Furthermore, pseudounivariate forms of multivariate regression models (using NAS) were used to estimate uncertainty of measuring methods, allowing calculations in accordance with the Guide to the Expression of Uncertainty in Measurement and also Monte Carlo simulation.