CFD-ANN-based model for parametric analysis of segmented plate-fin heat sinks: Exploring heat transfer and pressure drop trade-offs

Abstract

Effective air cooling remained critical for high-power electronics as heat fluxes increased. This study integrated experimental measurements, three-dimensional Computational Fluid Dynamics simulations (CFD), and an Artificial Neural Network (ANN) surrogate model to assess air-cooled plate-fin heat sinks under forced convection. Five different modified designs with segmented and staggered fins achieved up to 46.8% reduction in junction-to-ambient thermal resistance relative to the baseline, but the most aggressive layout (HS6) raised the pressure drop to approximately 1000 Pa, about 20x higher. The trained neural network model reproduced CFD temperatures and pressure drops with R2 > 0.99, enabling rapid exploration of the design space. From these data, two closed-form correlations for maximum junction temperature and pressure drop were derived, offering instant estimates during preliminary design without further CFD runs. Overall, staggered-segmented fins delivered substantial thermal gains, yet the associated rise in pumping power must be weighed during early design.