Rheological behavior of MWCNT-SnO2/SAE50 hybrid nanolubricant: Experimental evaluation and viscosity prediction using optimized machine learning model

Abstract

The present paper investigated the rheological behavior and dynamic viscosity of SAE 50 lubricant in the presence of a novel mixture of MWCNT(25 %)/SnO2(75 %) nanopowders. In the first part of the research, three essential parameters, including shear rate (SR = 1333.0–2932.6 s−1), temperature (T = 25–67 °C), and volume fraction (VF = 0–1.50 %), were considered to evaluate the rheological behavior of the hybrid nanolubricant. In the second part of this paper, highly accurate models for predicting the viscosity of the nanolubricant were developed by proposing a whole search-based strategy for the structural/training optimization of the multilayer perceptron neural network (MLPNN). The experimental results demonstrated that the nanolubricant behaved as a non-Newtonian fluid at various concentrations and temperatures. Dynamic viscosity increased by a minimum of 84.07 % (at VF = 1.50 % and SR = 1333.0 s−1) and a maximum of 88.72 % (at VF = 0.25 % and SR = 2132.8 s−1) when the temperature decreased from 67 °C to 25 °C. Additionally, the maximum augmentation in dynamic viscosity (103.22 %) occurred at T = 67 °C and SR = 1333.0 s−1. The proposed MLPNN optimization strategy, through the optimal selection of parameters such as the number of neurons of hidden layers (HLs), transfer functions of HLs, the transfer function of the output layer, and the training algorithm, provided significant efficiency in developing single HL (R2 = 0.99979) and double HLs (R2 = 0.99996) MLPNN models.

Keywords

• Hybrid nanofluid
• Dynamic viscosity
• Rheological behavior
• Nanolubricant
• Artificial Neural Network
• Machine learning