Transferência de calibração de modelos multivariados para previsão de propriedades físico-químicas em petróleo bruto

• Ano de defesa: 2017
• Autor(a) principal: Rodrigues, Rayza Rosa Tavares
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal do Espírito Santo; BR; Mestrado em Química; Centro de Ciências Exatas; UFES; Programa de Pós-Graduação em Química
• Palavras-chave em Português: Calibration transfer; Crude oil; Transferência de calibração; Calibração; Fontes de infravermelho; Petróleo; Química; 54
• Link de acesso: http://repositorio.ufes.br/handle/10/7320

Multivariate calibration combined with infrared technique is an alternative to traditional methods of determination of physicochemical parameters in crude oil. However, a multivariate model can only be applied for the instrument in which the spectra were measured. In case of equipment upkeep or change of instrument, transferring the calibration model is necessary for the new instrumental condition or new instrument. Transfer of calibration of multivariate models have fundamental importance for petroil industry, considering that it increases the aplicability of models and enables quick and unexpensive estimatives of various physico-chemical parameters. This work is divided into two sections: the first one is dedicated to evaluation of transfer models between two mid-infrared instruments (MIR) for PLS (partial least squares) models and OPLS (orthogonal projections to latent structures); and the second aims to apply the PDS (piecewise direct standardization) transfer tecnique to PLS models for prediction of API gravity, WAT (wax appearance temperature, in portuguese TIAC), TAN (total acid number, in portuguese NAT) and NAN (naphtenic acid number). Among the evatuated chemometric methods SBC (slope and bias correction), FR (full recalibration), DS (direct standardization) and PDS, for prediction of API density, the PDS transfer model applied to the OPLS model resulted in the best predictive ability (RMSEP of 1.48). Direct application of PDS to a PLS original model has the advantage of reusing consolidated models, and the results from the first section indicate that this is a possibility. At the second section, the properties API, WAT, TAN and NAN were modeled and validated by PLS. The secondary espectra, other than those of the first section, were subjected to a polynomial interpolation to equalize the variables to those of the primary instrument, prior to the application of the PDS transfer. The transferred and airPL (adaptative iteratively reweighted penalized least squares) processed spectra became indistiguishable from the primary spectra by PCA (principal component analysis). The PDS technique associated to PLS models was shown to be promissing, specially for being capable of estimating with good accuracy the value of API gravity of petroleum samples classified as medium oil