Prévision de la performance des tubes capillaires adiabatiques grâce à des approches conventionnelles et des réseaux neuronaux artificiels : une comparaison.

Performance prediction of adiabatic capillary tubes by conventional and ANN approaches: a comparison.

Résumé

An experimental study of adiabatic capillary tubes was conducted to evaluate the flow characteristics of refrigerant HFC-134a. The effect of various input parameters, such as capillary tube diameter, length, and inlet subcooling on the mass flow rate of HFC-134a, were investigated. Moreover, a comparison was made for the mass flow rate of refrigerant HFC-134a in instrumented and noninstrumented capillary tubes. It was found that the provision of taps for pressure measurement on the capillary tube surface has a negligible effect on the mass flow rate of HFC-134a. The data obtained from the experiments were analyzed, and a semi-empirical correlation using a multiple-variable regression analysis was developed. The proposed correlation predicts that more than 86% of the data lies in the error band of plus or minus 10%. Furthermore, an artificial neural network (ANN) model using a feed-forward back-propagation algorithm was developed to predict the mass flow rate from the given set of input parameters. These two approaches were compared, and ANN was found to predict the mass flow rate far more accurately than the conventional empirical correlation developed by regression.

Détails

  • Titre original : Performance prediction of adiabatic capillary tubes by conventional and ANN approaches: a comparison.
  • Identifiant de la fiche : 2009-2343
  • Langues : Anglais
  • Source : ASHRAE Transactions. Papers presented at the 2009 ASHRAE Winter Conference: Chicago, Illinois, January 2009. Volume 115, part 1.
  • Date d'édition : 25/01/2009

Liens


Voir d'autres communications du même compte rendu (20)
Voir le compte rendu de la conférence