Use of neural networks for the prediction of multicomponent vapor liquid equilibrium / Predição de equilibrio liquido-vapor de sistemas multicomponentes atraves de redes neurais

AUTOR(ES)
DATA DE PUBLICAÇÃO

2005

RESUMO

Many thermodynamic models for the data correlation of multicomponent liquid-vapor equilibrium (LVE) can be found in the literature. However, due the difficulty of these thermodynamic models to interpolate data at pressures where experimental data is not available, the use of Artificial Neural Networks was considered. Initially the resolution of liquid-vapour equilibrium equations was made through calculations of the bubble-point T for the ternary system 2-butanol / 2-butanone / water in order to get a reasonable amount of data to be used in the training of the networks. The thermodynamic model used in the representation of the liquid phase was NRTL (Non-Random-Two-Liquid). These data were then used to train and test neural network models, and we verified that the neural nerworks were capable of describing the equilibrium behavior with small deviations in predicted vapor composition, for isobaric systems. A neural model was then developed in MATLAB to make predictions of thermodynamic properties for the 2-butanol / 2-butanone / water system, using data at different pressures to train the network, and predict vapor composition and temperature at pressures nor used to train the network. As expected, a very poor result was obtained when two isobaric sets of data were used to predict LVE behavior at an intermediate pressure... Note: The complete abstract is available with the full electronic digital thesis or dissertations

ASSUNTO(S)

sistema ternario vapor-liquid equilibrium equilibrio liquido-vapor neural networks ternary systems redes neurais (computação)

ACESSO AO ARTIGO

Documentos Relacionados