An artificial neural network (ANN) was applied successfully to predict laminar free convection heat transfer coefficient from an isothermal horizontal cylinder of elliptical cross section confined between two adiabatic walls. Neural networks were used since they constitute a general, powerful function-approximator tool proving able to represent a convectional heat transfer coefficient precisely in the present case. The input database for the network includes 171 experimental data points. The experiment is carried out using Mach-Zehnder Interferometry. Tube axis ratio, wall spacing to miner axis ratio of tube and Rayleigh number are variable parameters or the experimental study. The values of the average Nusselt numbers predicted by the network are in very good agreement with the available experimental data. Therefore the network is used to predict the unavailable data points within the range of our experimental results.

This content is only available via PDF.
You do not currently have access to this content.