Share

Flow boiling of pure fluids: local heat transfer and flow pattern modelling through artificial neural networks.

Author(s) : SCALABRIN G., CONDOSTA M., MARCHI P.

Type of article: Article

Summary

A new modelling technique based on neural networks as a universal function approximator has been applied to study the local representation of flow boiling heat transfer at varying fluid dynamics conditions along the tube, i.e., moving through different flow pattern regions. After subdivision of the experimental data into subsets homogeneous for flow conditions, specific heat transfer equations have been heuristically got. Such specific equations have been validated and compared with the global MLFN heat transfer coefficient representation for the same target fluid as from a former work. The study has been extended to four fluids for which suitable data were available from the literature. The representation accuracies of the specialized heat transfer models are very high, even if such a method is not in general suggested due to the greater complication for both the equations development and for their practical use. It was furthermore shown that a global high accuracy flow boiling heat transfer equation can be developed also without any relation with local fluid dynamics conditions. A new modelling technique has been furthermore proposed for the fluid dynamics conditions along the tube. The linking of the picture of the actual flow condition directly with a conventional continuous real number, the so-called SF factor, allows to convert the flow condition into a numerical variable which can be turned into a continuous fluid specific function through a further MLFN technique. Limitedly to the few available literature SF data the method presents high performance and coherence for the whole SF surface. The present SF MLFN equation and the more advanced literature flow pattern model get comparable results, in spite of the limited data base. It is also shown that a flow boiling heat transfer model is not intrinsically flow pattern dependent, because a flow condition representation is never required for a global heuristic heat transfer coefficient modelling. [Reprinted with permission from Elsevier. Copyright, 2005].

Details

  • Original title: Flow boiling of pure fluids: local heat transfer and flow pattern modelling through artificial neural networks.
  • Record ID : 2007-1018
  • Languages: English
  • Source: International Journal of thermal Sciences - vol. 45 - n. 8
  • Publication date: 2006/08

Links


See other articles in this issue (2)
See the source

Indexing