A hybrid CFD-XGBoost framework for thermal-hydraulic performance prediction of perforated pin-fin heat sinks with different geometries

The thermal-hydraulic performance of perforated pin-fin arrays is critically influenced by fin geometry, perforation configuration, and flow conditions. This study presents an integrated computational fluid dynamics (CFD) and machine learning framework to systematically investigate and predict the heat transfer and pressure drop characteristics of circular, square, and diamond-shaped pin fins with single, double, and triple perforations. High-fidelity CFD simulations provided detailed insights into flow structures, turbulence intensity, and convective heat transfer mechanisms, revealing that fin geometry strongly governs wake development and turbulence production, while perforations effectively moderate pressure losses. The resulting CFD dataset was employed to train Extreme Gradient Boosting (XGBoost) regression models, which accurately predicted pressure drop and heat transfer coefficients, achieving coefficients of determination exceeding 0.93 and 0.99, respectively. The proposed hybrid CFD-AI methodology enables rapid and reliable performance evaluation, offering a powerful tool for the design optimization of perforated pin-fin heat sinks in advanced thermal management applications.