Experimental analysis of aerothermal relations in a heat sink with novel 3D-printed turbulators

Machine learning (ML) has rapidly emerged as a powerful tool that could transform the heat transfer area in the next three decades. Nevertheless, there is a lack of high-spatial-resolution experimental data for training the ML models. This study presents detailed experimental analyses of thermal-fluidic relations in a heat sink with innovative 3D-printed turbulators using non-intrusive measurement techniques such as infrared thermography (heat transfer) and particle image velocimetry (velocity). The heat sink consists of a U-shaped square channel whose side length and hydraulic diameter are equal to 45.5 mm. Featuring a curved shape, the turbulators are designed according to the heat transfer enhancement mechanism of many previous turbulators and manufactured with the additive manufacturing (3D printing) owing to their geometric complexity. The convergence angle, divergence angle, clearance ratio, and pitch ratio of the turbulators are 35 degrees, 40 degrees, 0.56, and 1.2-infinity, respectively, while the Reynolds number (Re) is kept at Re = 104. It is observed that the 3D-printed turbulators direct flow towards the heated wall and accelerate the near-wall flow, leading to impingement and strong convective cooling, respectively. Additionally, the highly accelerated core flow aids the Dean vortices in the 180-deg sharp turn, resulting in significant augmentation in heat transfer. Further, the Pearson correlation analysis shows that heat transfer Nu/Nu(infinity) is highly correlated with the mean spanwise velocity |W|/U-b and streamwise velocity |U|/U-b (Pearson correlation coefficient R >= 0.8) in the mid-turn and the first and second halves of the turn, respectively, revealing the spanwise and streamwise convection as the dominant heat transfer mechanism. Finally, detailed R data are provided in this study, which fills the void of database for machine learning of heat transfer in heat sinks.

Automated summary

Machine learning is a powerful tool that can transform the heat transfer field. This study presents detailed experimental analyses of a heat sink with innovative 3D-printed turbulators, providing high-spatial-resolution experimental data for training machine learning models.