Thermophysical properties and machine-learning prediction of copper oxide (CuO)-dispersed SAE 10W-30 nano-lubricants at different concentrations

Abstract The paper examines thermophysical characteristics of CuO-packed SAE 10W-30 nano-lubricants and constructs machine learning (ML) predictors for them. Surfactant-aided ultrasonication was used to prepare nano-lubricants at 0, 50, and 100 ppm concentrations, with main properties such as density, thermal conductivity, kinematic viscosity, and specific heat experimentally determined at an engine-relevant range of temperatures. The findings indicate that the introduction of CuO nanoparticles significantly improves thermophysical performance, with thermal conductivity, viscosity retention, and specific heat increasing by 18.1 %, 26.7 %, and 17.2 %, respectively, at 100 ppm relative to the base oil. Density also showed a moderate increase, and all properties expected temperature-related behaviour. R 2 , RMSE, MSE, and MRE measures were used to evaluate and develop machine learning models such as k -Nearest Neighbour, Random Forest, Decision Tree, and XGBoost. With R 2 values close to 0.999, the predictions were high. Experimental analysis combined with ML prediction is shown to provide a solid framework for the efficient design and optimisation of nano-lubricants. The results reveal the potential of low-concentration CuO nano-lubricants to improve heat transfer and lubrication in the engine and thermal management practices. A dataset of experimental data collected at different temperatures and nanoparticle concentrations contained 423 data points. The dataset was split into a training set (80 %) and a testing set (20 %), and model performance was evaluated using R 2 , RMSE, MSE, and MRE to ensure the models were not overfitting and that their performance was robust.