Influence of fluid flow models on the topology optimization of a passively cooled heatsink

Topology optimization of passively cooled heat sinks for electronics components remains a relatively unexplored area in comparison to other topology optimization problems. Most works use a full high-fidelity model based on the Navier–Stokes equations, but this can be prohibitively expensive without access to high-performance computing (HPC). If the methodology is to gain traction in industry, less computationally expensive models must be explored and developed. The influence of fluid flow modeling on the topology optimization of heatsinks is studied, specifically comparing a low-fidelity model, utilizing a potential-like flow model, and a high-fidelity model, utilizing the Navier–Stokes equations. This work introduces three key innovations: (1) a hybrid optimization strategy that combines low- and high-fidelity fluid models to reduce computational costs by 59–86% while retaining 70–85% of the performance gains; (2) an evaluation model using Heaviside interpolation to rapidly analyze designs from density-based results, enabling practical industrial adoption; and (3) an improved tuning procedure for the proposed low-fidelity model. Compared to traditional finned heatsinks for the included case study, this framework achieves up to 6.7% lower thermal compliance with 44% less volume, which was achieved using the high-fidelity model exclusively along with the highest computational time of 1287 core-hours. The exclusive design of the low-fidelity model outperforms the reference design with 4.4% lower thermal compliance with 46% less volume, and had the lowest computational time of only 86 core-hours. The proposed hybrid approaches outperform the reference design by 4.8% in thermal compliance with 47% less volume with 184 core-hours, and 5.7% in thermal compliance with 43% less volume with 527 core-hours.