Artificial neural networks to correlate in-tube turbulent forced convection of binary gas mixtures

Turbulent forced convection correlations are documented in the literature for air, gases and vapors (Pr similar to 0.7), for common liquids (Pr > 1) and for liquid metals (Pr < 0.03). In spite of this, there is a small gap in the Pr sub-interval between 0.1 and 1.0, which is occupied by binary gas mixtures. In this paper, data for turbulent forced convection for the in-tube flow have been gathered and a fully connected back-propagation Artificial Neural Network (ANN) is used to learn the pattern of Nu as a double-valued function of Re and Pr. The available data are separated in two subsets to train and test the neural network. A set with 80% of the data is used to train the ANN and the remaining 20% are used for testing. After the neural network is trained, we make use of the excellent nonlinear interpolation capabilities of ANNs to predict Nu for the sought range 0.1 < Pr < 0.7 for Re between 10(4) and 10(6). These predictions are later extended to generate a comprehensive correlation for Re between 10(4) and 10(6) that aptly covers the Complete spectrum of Prandtl numbers. (C) 2008 Elsevier Masson SAS. All rights reserved.