Optimization on microchannel structures made of typical materials based on machine learning

With the trend toward miniaturization of functional devices, material preparation and thermal management processes are also limited to small spaces. Microchannels have emerged as an optimal solution for these challenges. Microchannel-based reactors can generate hybrid materials, and the integration of microchannel heat sinks and substrates can control the temperature of high-power devices. The microstructure within microchannels significantly influences fluid flow and heat transfer, impacting the efficiency of both reaction and heat dissipation processes. Pin-fins are widely used microstructures due to their ability to increase heat transfer area and enhance fluid mixing. In order to find the optimal structure of the fins, it is essential to explore a vast parameter space. In this paper, artificial neural network and genetic algorithm are combined to optimize the copper irregular pin–fin microchannels. Initially, a large number of numerical simulations are performed, focusing on adjustable parameters such as fin radii in various directions, while monitoring the heating surface temperature and the pressure drop of the fin section. Then, nearly 2000 sets of accumulated data are used to train the neural network, establishing the relationship between structural and performance parameters. Finally, a genetic algorithm is employed for multi-objective optimization, yielding a Pareto front. The findings reveal that the newly obtained optimized microchannels exhibit superior thermal–hydraulic performance compared to traditional microchannels. The mechanism of heat transfer enhancement in the optimized microchannel has been revealed: the arrangement of asymmetric fins allows for more thorough contact between the fluid and the fins. Based on this rule, the newly designed multi-fin microchannels exhibit better performance under both fixed heat flux and fixed temperature conditions. In addition, doping high thermal conductivity materials into the substrate to form composite materials can significantly improve the heat transfer performance of microchannels, and using materials with different doping ratios in different parts of the microchannel can effectively improve the temperature uniformity of the heating surface. Thus, uniform-temperature microchannels are designed by combining metal materials (such as copper and aluminum) with non-metal materials (like diamond and graphite).

Graphical Abstract