Study of rectangular fin heat sink performance and prediction based on artificial neural network

The plate-fin heat sink is a widely used heat dissipation device in thermal management systems, playing a crucial role in maintaining the stable operation of electronic equipment and reducing economic costs. In this study, the wind tunnel experimental device was employed to test the thermal and hydraulic properties of a rectangular fin heat sink. The CFD model was utilized to investigate the Nusselt number and friction factor of the rectangular fin heat sink under varying Reynolds numbers, fin numbers, and fin widths. The results indicate that as Reynolds number increases, the friction factor decreases while the Nusselt number increases. An increase in fin number leads to a decrease in both Nusselt number and friction factor. Moreover, with an increase in fin width, the friction factor increases while changes occur differently for Nusselt number at different Reynolds numbers. Furthermore, a multi-layer feedforward neural network based on a genetic algorithm (GA-BP) is employed to predict both Nusselt number and friction factor; subsequently evaluating these predicted results. The findings demonstrate that GA-BP neural network can accurately and rapidly predict thermal and hydraulic properties of heat sinks under various conditions including different Reynolds numbers, fin numbers, and fin widths.