Flow geometry optimization of channels with baffles using neural networks and second law of thermodynamics

Artificial Neural Networks results from an experimental study on the heat transfer and flow characteristics in channels with baffles are presented. Nine different types of channels with baffles were used in order to test the effects of baffle spacing to and the influence of baffle geometries and their positions on heat transfer, and their flow characteristics. Experiments were performed for laminar and turbulent flows, and Prandtl number of 0.7. The convective heat transfer coefficients and pressure drops provided by the experimental studies and artificially generated data were examined. Finally, the geometric features of the proposed flow geometry to improve heat transfer can be selected in order to yield the maximum opposite reduction in heat exchange channel irreversibility by using entropy generation minimization method. The experimental results for different design constraints show that optimum baffle position angle for laminar flow is 90° and for turbulent flow is 45° and baffle spacing modulo length ratio is 3 for laminar flow and two for turbulent flow, respectively. Then, some new data is generated using artificial neural networks for succession parameters. Also these generated data are compared with experimental results. New flow geometries are presented for the future applications.