Abonnement à la biblothèque: Guest

ISSN Online: 2377-424X

International Heat Transfer Conference 12
August, 18-23, 2002, Grenoble, France

A new heat transfer correlation in the transition region for a horizontal pipe with a reentrant inlet − using artificial neural network

Get access (open in a dialog) DOI: 10.1615/IHTC12.1280
6 pages

Résumé

Local heat transfer (under uniform wall heat flux condition) was measured along a 6.1 m long stainless steel horizontal circular straight tube with a reentrant inlet configuration. For the heat transfer measurements the Reynolds number varied from about 790 to 49940. In these experiments the test fluid was either distilled water or a mixture of ethylene glycol and distilled water, which gave a Prandtl number range from about 4 to 100. From the experimental data, it can be observed that the transition region for the reentrant inlet ranged from 1700 to 9100. Ghajar and Tam (1994) proposed heat transfer correlations for the laminar and turbulent flow regimes with excellent accuracy. The absolute average deviations are 5.8% and 3.7%, respectively. For the correlation in the transition region, the accuracy is not as accurate since more than 30% of the data were predicted with more than 10% deviation. The reason is due to the abrupt change in heat transfer characteristics in this flow regime. Since the value of heat transfer coefficient has direct impact on the size of heat exchangers, a more accurate correlation is developed using the artificial neural network (ANN). It has been shown that multilayer feedforward ANN's have universal approximation property according to Hornik et al. (1990) and Hornik (1991). The accuracy of the new correlation is excellent with 95.5% of the data points (421 data points) predicted with less than 10% deviation. The ANN method can also be used to establish the most and least important variables according to the form of the correlation.