Optimization of a microchannel heat sink with surrogate model and genetic algorithm

Abstract

Flow maldistribution significantly impacts the performance and operation of microchannel heat sinks, which are extensively utilized across various industries. This study develops a surrogate model to address the flow maldistribution in microchannel heat sinks, employing a combination of genetic algorithm backpropagation neural networks and numerical simulations. Genetic algorithms optimize the microchannel heat sink manifold, involving five variables. The optimized manifold reduces the flow maldistribution factor by 38 % at a coolant inlet velocity of 0.1 m/s compared to that of the conventional structure. The mechanisms of this improvement are explored through comparative analyses of the manifold shape and pressure distribution. However, addressing the flow maldistribution alone proves insufficient in eliminating local hot zones within the microchannel heat sink. Consequently, this study introduces the concept of a corrected maldistribution factor and establishes a new structure guided by this factor as the optimization index. The results indicate a 47 % reduction in the maximum-to-minimum channel flow velocity ratio compared to that of the conventional structure at a coolant inlet velocity of 2 m/s. Furthermore, the maximum temperature drops by 20 K, significantly enhancing the flow and temperature distribution within the microchannel heat sink.