A population-adaptive genetic algorithm for topology optimization of liquid-cooled heat sinks

Genetic algorithm (GA)-based topology optimization (TO) has emerged as a promising approach for addressing conjugate heat transfer problems. While the GATO method offers numerous advantages due to its explicit and gradient-free character, the high computational cost remains a major limitation. This study aims to enhance the efficiency and practical applicability of GATO in the design and structural optimization of liquid-cooled heat sinks under heterogeneously heating conditions in electronic device thermal management. First, a systematic parametric investigation examines the influence of GA parameters (elite number, mutation rate, crossover scheme, and population size) on TO outcomes, identifying population size as the most influential factor. Motivated by this finding, we propose a Population-Adaptive Genetic Algorithm (PAGA), which dynamically reduces the population size in response to the objective function values during TO convergence. This strategy substantially reduces computational cost by over 50% without compromising solution quality. The effectiveness, robustness, and versatility of the PAGA-based TO framework are further demonstrated through optimization studies under diverse boundary and operating conditions, including varied inlet–outlet layouts, heat source distributions, and liquid coolants with distinct thermophysical properties. In all examined cases, the optimized flow channel configurations are physically consistent and thermally efficient due to their ability to spatially match cooling capacity with demand. For all coolants tested, TO-derived heat sinks consistently outperform the conventional straight-channel design in mitigating temperature hotspots, confirming the effectiveness and broad applicability of the proposed method.