Prévisions de la performance des échangeurs de chaleur fondées sur un modèle et des réseaux neuronaux artificiels.

Performance prediction of heat exchangers based on model and artificial neural network.

Auteurs : DING G. L., ZHANG C. L.

Résumé

In order to improve the accuracy of performance prediction and to expedite calculations, a new method is presented which combines mathematical model and artificial neural network (ANN). This method includes three parts. The first part, simplified fundamental modelling of heat exchangers can give a prediction of heat exchanger performance quickly. The second part, ANN for simplification, is to make up the error caused by the model simplification of the first part. The third part, ANN for accuracy improvement, is to automatically modify the simple model according to the difference between the predictions and experimental data. The mean and the maximum error between predictions and experimental data for condensers are 0.63 and 1.72%, respectively. For evaporators, the mean and the maximal error are 1.2 and 10.5%, respectively. The calculation speed with the new method are two orders of magnitude faster than those with the distributed-parameter mathematical model.

Détails

  • Titre original : Performance prediction of heat exchangers based on model and artificial neural network.
  • Identifiant de la fiche : 2006-2323
  • Langues : Anglais
  • Source : Cryogenics and refrigeration. Proceedings of ICCR 2003.
  • Date d'édition : 22/04/2003

Liens


Voir d'autres communications du même compte rendu (166)
Voir le compte rendu de la conférence