Prediction of the Dynamic Viscosity of MXene/palm Oil Nanofluid Using Support Vector Regression

Abstract

Dyanamic viscosity values of Mxene/palm oil nanofluid as availed from the experiments at varying temperatures and concentrations are used to create a vector regression model. The range of nanoparticle concentration was taken in the range of 0.01–0.1% and the mass percentage of the nanofluid was chosen as 0.2% with the range of temperature varying between 20 and 100 ℃. In order to create the model, concentrations and temperature were chosen as the inputs, whereas dynamic viscosity was chosen as the output parameter. 80% of data was considered, while training the model, whereas the rest of the data was used to validate the performance of the model. To develop the most optimum model, performance of the model was determined for several kernels, and the radial basis function kernel was found to be giving the best performance. Further, grid search technique was used to determine the most optimum model. The developed support vector regression model showed good performance with the values of coefficient of determination (R²) and mean average error (MAE) being 0.99987 and 7.9E–03, respectively. The maximum deviation between the experimental values and model predicted values was only −5.18%. It was also found that 99.33% of values predicted by the model were in the deviation range of ±4%. From the obtained values of these parameters, it can be concluded that support vector regression models are able to anticipate the values of dynamic viscosity with appreciable precision.