Desenvolvimento e aplicaÃÃo de modelos matemÃticos baseados em redes neurais artificiais e modelos estatÃsticos na prediÃÃo de propriedades fÃsico-quÃmicas da gasolina petroquÃmica

AUTOR(ES)
DATA DE PUBLICAÇÃO

2003

RESUMO

The production of petrochemical gasoline is simply carried out by the mixture of industrial hydrocarbon streams of different units of petrochemical processing. These streams must present a chemical composition that allows to reach a blend with a composition spectra like the gasoline. That is, the main difficult consists of determining of the best set of volumetric fractions of these streams that must be used to obtain the properties in accordance with the quality parameters of the gasoline. To reach these objective, itâs necessary to develop mathematical models for the prediction of the physical-chemical properties of the streams blends. Here it is presented the development of empirical models, based on artificial neural networks and statistical models that, starting of the knowledge of the volumetric fractions of five petrochemical streams that are used for the formulation of gasoline, were able to carry out efficient predictions of important properties that are used for its quality characterization: vapor pressure, distillation curve and density. As the models are empirical, it was necessary the carrying out of essays of mixture of the petrochemical streams to formulate several blends of gasoline. That is, essays of physical-chemical characterization of these blends were carried out. Therefore, a big experimental data bank was constructed. Based on this data bank, the empirical models were development and tested

ASSUNTO(S)

distillation curve pressÃo de vapor gasolina artificial neural networks modelos estatÃsticos gasoline redes neurais artificiais vapor pressure statistical models engenharia quimica curva de destilaÃÃo

ACESSO AO ARTIGO

Documentos Relacionados