Determinação do número de acidez total em resíduo de destilação atmosférica e de vácuo do petróleo empregando a espectroscopia no infravermelho (ATR-FTIR) e calibração multivariada

Detalhes bibliográficos
Ano de defesa: 2007
Autor(a) principal: Parisotto, Graciele
Orientador(a): Não Informado pela instituição
Banca de defesa: Não Informado pela instituição
Tipo de documento: Dissertação
Tipo de acesso: Acesso aberto
Idioma: por
Instituição de defesa: Universidade Federal de Santa Maria
BR
Química
UFSM
Programa de Pós-Graduação em Química
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://repositorio.ufsm.br/handle/1/10427
Resumo: The crude oil might contain a series of contaminants that can to produce undesired properties. Naphthenic acids (NA), for example, perhaps promotes corrosion in refineries. The determination of total acid number (TAN) in petroleum products is recommended by ASTM standard D 664-04, based on potentiometric titration in non aqueous media. As an alternative to official methodology, in this work, TAN in atmospheric residue and vacuum residue samples of petroleum distillation was determined using middle infrared spectroscopy with attenuated total reflection (ATR-FTIR) in association to chemometric methods. For the development of calibration models, at first, was made the detection of outliers samples using principal component analysis (PCA). After, the selection of samples for calibration and prediction set was made by hierarchical cluster analysis (HCA). Calibration set was consisted of 44 samples and prediction one by 13, summarizing 16 samples of atmospheric residue and 41 samples of vacuum residue. Calibration models were developed using three variable selection models: method of interval partial least squares (iPLS), method of synergy partial least squares (siPLS) and method of backward interval partial least squares (biPLS). Different treatments and preprocessing were evaluated for models development also. The treatment based on first derivative with Savitzky-Golay filter and the data centered in the media produced the best models using biPLS. Spectra were divided in 20 intervals and, finally, 5 intervals were combined (2992 till 2826, 1823 till 1657, 1656 till 1490, 1489 till 1323 and 821 till 655 cm-1). This model produced a root mean square error of cross-validation (RMSECV) of 0.1649 mg KOH g-1 and a root mean square error of prediction (RMSEP) of 0.1642 mg KOH g-1, showing a coefficient of determination of 0.9819 and a medium error of 22.5%. The analytical method for TAN determination allows fast analysis and relatively low cost in these samples, being of easy application in industry environment.