Experimental and explainable machine learning approach on thermal conductivity and viscosity of water based graphene oxide based mono and hybrid nanofluids

Abstract

This study explores the thermal conductivity and viscosity of water-based nanofluids containing silicon dioxide, graphene oxide, titanium dioxide, and their hybrids across various concentrations (0 to 1 vol%) and temperatures (30 to 60 °C). The nanofluids, characterized using multiple methods, exhibited increased viscosity and thermal conductivity compared to water, with hybrid nanofluids showing superior performance. Graphene oxide nanofluids displayed the highest thermal conductivity and viscosity ratios, with increases of 52% and 177% at 60 °C and 30 °C, respectively, for a concentration of 1 vol% compared to base fluid. Similarly, graphene oxide-TiO2 hybrid nanofluids achieved thermal conductivity and viscosity ratios exceeding 43% and 144% compared to the base fluid at similar conditions. This data highlights the significance of nanofluid concentration in influencing thermal conductivity, while temperature was found to have a more pronounced effect on viscosity. To tackle the challenge of modeling the thermophysical properties of these hybrid nanofluids, advanced machine learning models were applied. The Random Forest (RF) model outperformed others (Gradient Boosting and Decision Tree) in both the cases of thermal conductivity and viscosity with greater adaptability to handle fresh data during model testing. Further analysis using shapely additive explanations based on cooperative game theory revealed that relative to temperature, nanofluid concentration contributes more to the predictions of the thermal conductivity ratio model. However, the effect of nanofluid concentration was more dominant in the case of viscosity ratio model.