The flow boiling inside horizontal smooth tubes is studied for mixtures and a new modeling technique is proposed for the heat transfer. In a former work, the flow boiling of pure fluids inside smooth tubes was studied with artificial neural networks (ANN), obtaining individual heat transfer equations in a totally heuristic mode from the direct regression of experimental data. In the present work the method is extended to mixtures. Also in this case the correlation architectures considered for the ANN functions were based on the directly-accessible physical quantities controlling the phenomenon. Two modeling architectures are set up: a first one with the controlling physical quantities as independent variables and a second one where the inputs are the individual heat transfer values of the pure components and the mixture composition. The validation of both the obtained models demonstrate an accurate and substantially equivalent representation of the available heat transfer data with a considerable increase of prediction capability with respect to the conventional methods. The multilayer feedforward neural network equations both in terms of physical variables and in terms of components heat transfer models and composition have obtained very satisfactory accuracies. The modeling technique allows to avoid the need for any thermodynamic and transport property model for the target mixture and, moreover, the method can be used to check the consistency of new data sets before using them for processing.

Mixtures flow boiling: modeling heat transfer through artificial neural networks.

SCALABRIN, GIANCARLO;
2006

Abstract

The flow boiling inside horizontal smooth tubes is studied for mixtures and a new modeling technique is proposed for the heat transfer. In a former work, the flow boiling of pure fluids inside smooth tubes was studied with artificial neural networks (ANN), obtaining individual heat transfer equations in a totally heuristic mode from the direct regression of experimental data. In the present work the method is extended to mixtures. Also in this case the correlation architectures considered for the ANN functions were based on the directly-accessible physical quantities controlling the phenomenon. Two modeling architectures are set up: a first one with the controlling physical quantities as independent variables and a second one where the inputs are the individual heat transfer values of the pure components and the mixture composition. The validation of both the obtained models demonstrate an accurate and substantially equivalent representation of the available heat transfer data with a considerable increase of prediction capability with respect to the conventional methods. The multilayer feedforward neural network equations both in terms of physical variables and in terms of components heat transfer models and composition have obtained very satisfactory accuracies. The modeling technique allows to avoid the need for any thermodynamic and transport property model for the target mixture and, moreover, the method can be used to check the consistency of new data sets before using them for processing.
File in questo prodotto:
Non ci sono file associati a questo prodotto.
Pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11577/106793
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 17
  • ???jsp.display-item.citation.isi??? 15
social impact