Using artificial neural network for computing the development length of MHD channel flows

Abstract In the present study, an artificial neural network (ANN) was applied for computing the development length of laminar magnetohydrodynamics (MHD) flow in the entrance region of a channel. The finite volume method (FVM) was conducted to investigate the laminar MHD channel entrance flow. The investigation was applied for Reynolds number (Re) ranging from 600 to 1200 while Hartmann number ( Ha ) ranging from 4 to 14. 60 datasets were obtained from numerical solution and then, a feed-forward back-propagation neural network contained one hidden layer including was trained and developed to predict the development length. Using ANN, a correlation for predicting the MHD channel flow development length was proposed. Comparison of the ANN with FVM and proposed correlation results revealed that ANN can be employed for modeling the developing MHD channel flow and prediction of the development length. It was found that with the increase of Ha , the velocity profile gets flatten and consequently, the development length becomes shorter. Moreover, by augmentation of Re, the development length increases.