Experimental and Machine Learning-Based Investigation of Heat Transfer Enhancement in Interrupted Minichannels with CuO Nanofluid

Abstract

This study experimentally investigates the influence of nanofluid concentration and minichannel geometry on convective heat transfer and pressure drop characteristics in minichannel heat sinks. A custom-designed test rig, incorporating precision flow control, calibrated thermal sensors, and a differential pressure transducer, was developed to ensure accurate thermal-hydraulic measurements. Six microchannel geometries—comprising both rectangular and V-type interrupted configurations—were fabricated using EDM and evaluated using CuO-water nanofluids prepared via a two-step synthesis method without surfactants.Results demonstrate that the V-type configuration consistently achieves superior heat transfer performance, as evidenced by higher Nusselt numbers, especially at elevated Reynolds numbers and nanoparticle concentrations. The optimal enhancement is observed at a 0.2% CuO concentration and V534 geometry, which strikes a favorable balance between increased heat transfer and manageable pressure drop. Uncertainty analysis confirms the robustness of the experimental data, with errors in heat transfer coefficient and Nusselt number maintained below 1.5%.Complementing the experiments, machine learning models were employed to predict thermal performance across various configurations. Gradient Boosting and Random Forest algorithms demonstrated high accuracy (R² > 0.95) in predicting heat transfer coefficients and Nusselt numbers, validating their applicability for thermal optimization. The findings emphasize the critical role of channel geometry and nanoparticle concentration in designing efficient micro-scale thermal management systems, and highlight the potential of machine learning for predictive modeling and optimization in thermal engineering.