A reduced-order model-based genetic algorithm for optimizing pin–fin arrays in multi-source electronics cooling

The thermal management of densely packed electronic devices is often challenged by multiple heat sources with uneven power dissipation, which can lead to severe temperature non-uniformity and hot spots. To address this issue, we develop an efficient optimization framework for pin–fin heat sinks that integrates a genetic algorithm (GA) with a reduced-order model (ROM). The ROM enables rapid generation of large GA populations, eliminating the heavy cost of CFD- or experiment-based approaches, and its predictions are validated against high-order finite element simulations. In a case study with two heat sources of unequal intensity, the optimized fin distribution places larger fins near high-flux regions and smaller fins in low-flux areas, improving temperature uniformity by 22.2% while limiting the overall pressure drop increase to just 1.5%. The results demonstrate the potential of the ROM–GA framework as a scalable strategy for thermal optimization of pin–fin arrays, with applicability to more complex systems featuring multiple and irregularly shaped heat sources.