Optimal configuration of discrete heat sources in a vertical duct under conjugate mixed convection using artificial neural networks

This paper reports the results of a numerical investigation of the problem of finding the optimum configuration for five discrete heat sources, mounted on a wall of a three-dimensional vertical duct under mixed convection heat transfer, using artificial neural networks (ANN). The objective is to locate the positions for the five heat sources in such a way that the maximum temperature of any of the heat sources in a given configuration is a minimum. The three-dimensional governing equations of mass, momentum and energy equations for the fluid flow and the energy equation for the solid regime have been solved by using FLUENT 6.3 and a database of temperature versus configuration was generated. The temperature database developed from CFD simulations is used to train the neural network. The trained neural network predicts the temperature of the heat sources very accurately and much faster than the CFD software. With the use of this network, an exhaustive search for all possible configurations was done that resulted in a global optimum for the problem.