Experimental and Data-Driven Modeling of Thermo-Hydraulic Performance in Carbon Nanofluids at Ultra-Low Concentrations

• Experimental thermo-hydraulic assessment of carbon-based nanofluids at ultra-low concentrations (0.005–0.01 wt.%). • Graphene nanofluid (0.005 wt.%) achieved >20% heat transfer enhancement with <15% pumping power increase. • Integrated thermal–hydraulic performance evaluation balancing heat transfer gain and pressure penalties. • Explicit machine-learning-based correlations developed with R² up to 0.999. • OLS, Theil–Sen and Elastic Net outperformed ensemble methods for physically interpretable modeling. Nanofluids at ultra-low nanoparticle concentrations have emerged as promising alternatives for heat transfer enhancement while minimizing hydraulic penalties. This study presents a data-driven modeling framework supported by experimental data to evaluate the thermo-hydraulic performance of carbon-based nanofluids (MWCNT, FMWCNT, and graphene) dispersed in H 2 O - EG (50:50 wt.%) at 0.005 and 0.01 wt.%, based on measurements in a horizontal smooth copper tube (L = 1700 mm, Di = 14.29 mm) under constant heat flux (5 - 11 kW/m²) and Reynolds numbers from 250 to 5400. The dataset was used to develop machine-learning-based predictive correlations using OLS, Theil–Sen, Elastic Net, Random Forest, and XGBoost models, achieving high predictive accuracy (R 2 = 0.995 - 0.999) and capturing nonlinear interactions between flow conditions and nanoparticle concentration. Experimentally, maximum heat transfer enhancements reached ∼19.97% for FMWCNT and ∼18.15% for graphene, while pressure drop penalties were ∼31 % for MWCNT in laminar flow. Thermohydraulic analysis indicated that graphene at 0.005 wt.% provides the most favorable performance trade-off, delivering > 20% heat transfer enhancement with <15 % increase in pumping power. These findings demonstrate that data-driven models can predict nanofluid performance and support the design of compact heat exchangers, electronic cooling systems, and energy-efficient thermal management applications.