Predictability evaluation of support vector regression methods for thermophysical properties, heat transfer performance, and pumping power estimation of MWCNT/ZnO–engine oil hybrid nanofluid

Abstract

In this paper, the efficiency of support vector regression (SVR) model to predict the thermophysical properties, heat transfer performance, and pumping power of MWCNT/ZnO–engine oil hybrid nanofluid has been demonstrated. Temperature and solid concentration were considered as inputs in order to train the model to estimate thermophysical properties including thermal conductivity, dynamic viscosity, heat transfer performance, and pumping power in both internal laminar and turbulent flow regimes. The results took the mean of ten runs of fivefold cross-validation of a Gaussian kernel with a combination of the Bayesian optimization technique to find the best tuning parameters. In order to evaluate the performance of the proposed model, three popular statistical indices including root mean square error (RMSE), mean absolute error (MAE), and correlation coefficient (R) have been calculated. The results show that the SVR technique is highly suitable for thermophysical properties prediction as it can map complex relationships between variables. Finally, the studied thermophysical properties have been successfully developed using the proposed model. Regarding the developed results, increasing temperature resulted in increasing the thermal conductivity, dynamic viscosity, and pumping power of MWCNT/ZnO–engine oil hybrid nanofluid. Despite the dynamic viscosity, the thermal conductivity increased as the solid concentration increased. It was revealed that the machine-learning technique proposed in this research could be developed as a very useful predictive tool.