AI-Guided Fractal Geometry for Passive Thermal Management in Edge Computing Systems Through Biomimetic and AI-Generated Heat Sink Design

As edge computing technologies advance, efficient thermal management becomes increasingly important. Biomimicry offers a potential approach to passive cooling through geometries inspired by natural branching structures. However, relatively few studies fabricate and experimentally test complex biomimetic heat sinks alongside computational analysis. This study experimentally and computationally compares biomimetic and AI-generated heat sink geometries under controlled natural-convection conditions. A conventional straight-fin control, tree-inspired fractal, lung-inspired fractal, and AI-generated fractal heat sink were manufactured using AlSi10Mg and tested under a uniform 7.5 W heat input and controlled ambient conditions. The same geometries and thermal conditions were modeled using computational fluid dynamics (CFD). Thermal performance was evaluated using 25-minute endpoint temperature, endpoint thermal resistance, and mass-normalized thermal conductance. The AI-generated fractal produced the lowest endpoint temperature and thermal resistance; however, its substantially greater mass and 10 mm base thickness limited direct attribution of this performance to geometry alone. The tree-inspired fractal exhibited thermal behavior comparable to the conventional control while using 51.1% less material and achieved the highest mass-normalized thermal conductance of the four designs. CFD simulations consistently predicted higher temperatures than experimental measurements, highlighting differences between computational and physical testing conditions. These findings demonstrate that biomimetic branching geometries can improve material efficiency while maintaining competitive thermal performance, supporting further investigation of lightweight passive thermal management for compact computing systems.