Dynamic viscosity prediction for MWCNT-MgO (10%–90%)-SAE40 oil hybrid nano-lubricant using artificial neural network andmulti-dimensional nonlinear least square curve fitting

Abstract

In this study, the dynamic viscosity of MWCNT-MgO (10%-90%)/SAE40 Oil Hybrid Nanofluid (HNF) is investigated. To this end, two methods are used including artificial neural network (ANN) and least square curve fitting. This nonlinear method uses a multi-dimensional least square strategy to determine coefficients of various parameters in the viscosity function. The viscosity of MWCNT-MgO (10%-90%)/SAE40 Oil HNF is predicted using three input parameters including solid volume fraction (SVF), temperature, and shear rate (SR). According to the obtained data, the optimum ANN to predict the viscosity of HNF as a function of inputs has 14 neurons with a nonlinear activation function resulting in mean square error or MSE=7.74e-4 and R2=0.999226 for all data. Moreover, temperature is the most influential parameter, and the rest of the parameters, have much fewer effects on the viscosity of MWCNT-MgO (10%-90%)/SAE40 Oil HNF. Furthermore, this can be seen in the obtained nonlinear fitted function through coefficients of input parameters. To be more precise, T coefficient is -91.2407 much higher than the solid volume fraction and shear rate with coefficients of 34.0366 and -0.166 respectively. The obtained ANN structure and nonlinear fitted function can be used for optimization purposes in future research. It is worth mentioning that although the fitted function by least square curve fitting is a very handy and symbolic equation, the ANN showed higher accuracy.