Multi-objective optimization of a microchannel heat sink with semi-elliptical cavities based on neural network and genetic algorithm

Highly integrated electronic devices generate high heat fluxes, challenging for thermal management. Microchannel heat sinks are effective coolers, performance enhancements often incur high pressure drop (ΔP). This study proposes a microchannel heat sink with semi-elliptical concave cavities to balance thermal and hydraulic performance. We examine the effects of concave cavity geometry and inlet Reynolds number (Re) on the Nusselt number (Nu) and ΔP using numerical simulation. A neural network–genetic algorithm framework was used for multi-objective optimization, with the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) method selecting the optimal design from the Pareto front. The results indicate that cavity structure improves heat transfer without significant increasing ΔP. Increasing the short-axis length b initially enhances Nu, which subsequently declines, while ΔP increases monotonically. In contrast, the long-axis length a exerts a comparatively weaker influence: Nu first increases then decreases, while ΔP drops. Compared to the unoptimized structure, the optimal solution increases Nu by 1.2 % and reduces ΔP by 13.5 %. This study provides a reference for cooling precision sensors and analog chips, aiding in the development of low-power heat dissipation solutions.