Machine learning-assisted prediction of thermal-hydraulic behaviour in additively manufactured microchannels using hybrid nanofluids and surface roughness correlation

This research explores the thermal and hydraulic behaviour of straight microchannel heat sinks (MCHS) fabricated using Direct Metal Laser Sintering (DMLS) and cooled with hybrid nanofluids containing Al2O3 and CuO nanoparticles. The combined use of these nanoparticles improved the fluid’s thermophysical properties, while the natural rough surfaces formed during additive manufacturing promoted passive convection. Experiments were conducted with nanofluid concentrations of 0.02% and 0.05% for Reynolds numbers ranging from 23 to 125, under heat fluxes of 20 to 40 W/cm². At a 0.05% concentration and a 30 W/cm2 heat flux, the Nusselt number increased by 12.9%, the surface temperature decreased by 13.8%, and the pressure drop rose by a moderate 6.4%. The performance evaluation criteria (PEC > 1) confirmed that the heat transfer benefit outperformed the hydraulic penalty. The experimental data were further applied to machine learning regression models, including polynomial regression, random forest, Gradient boosting and decision tree. Among them, the Gradient boosting model achieved the highest prediction accuracy, with R2 values above 0.94. The findings highlight that integrating hybrid nanofluids, additive manufacturing, and predictive modelling can lead to compact and energy-efficient cooling systems for modern power and electronic devices.