Applying artificial neural network and curve fitting method to predict the viscosity of SAE50/MWCNTs-TiO2 hybrid nanolubricant

In this study, two sets of laboratory data were used to foresee the rheological behavior of the hybrid nanolubricant of SAE50/MWCNTs-TiO2. For this purpose, at the first, power-law and consistency indices were attained by curve-fitting on shear stress-shear rate diagrams. Then, some correlations were proposed to estimate the viscosity, consistency-index and power-law-index as a function of temperature, solid volume fraction and shear rate. After that, an optimal artificial neural network (ANN) was proposed to predict the viscosity and another ANN was designed to predict the consistency-index and power-law-index. One hundred and forty-five viscosity values were used for forecasting viscosity (ANN-1). Since the rheological behavior of the nanolubricant was estimated by the power-law model, the indices were predicted using ANN-2. In this way, thirty experimental data for each index also were considered to supply ANN-2. For both ANNs, the data categories were separated to training and test data sets, which consisted of 80% and 20%, respectively. Results showed that for an exact prediction of power-law-index, it is compulsory to suggest 6 equations that each correspond to a temperature. This problem was solved by providing a neural network for simultaneous prediction of consistency and power-law indices. Comparisons between the proposed correlations and ANNs indicated that ANNs were more useful than correlations. [Display omitted] •2 sets of laboratory data to foresee rheological hybrid nanolubricant behavior.•Power law & consistency indices by curve-fitting on shear stress-shear rate.•New correlations to estimate viscosity, consistency & power law index.•Designing optimal ANNs to predict the viscosity, consistency & power law index.•ANNs were more useful than correlations.