Optimization of accuracy in estimating the dynamic viscosity of MWCNT-CuO/oil 10W40 nano-lubricants

Abstract

Artificial neural network (ANN) is one of the best models with good performance for predicting laboratory data, Due to its high accuracy, this design can be a suitable alternative to frequent and costly testing. In this study, the viscosity (μnf) of MWCNT-CuO (10-90)/Oil 10W40 nano-lubricant is modeled by ANNs by experimental data. μnf is measured in φ=SVF=(0.05-1% and temperature range T=5 to 55°C to train the ANNs. To check the precision of predicted data by ANN, mean square error (MSE), regression coefficient, and also margin of deviation (MOD) are used. The optimal structure was selected from among 400 ANN samples for MWCNT-CuO (10:90)/Oil 10W40 nano-lubricant, which has two hidden layers and the number of 4 and 8 neurons, as well as tansig and logsig transfer functions. The inputs of the ANN model are solid volume fraction (SVF or φ), temperature (T), and shear rate (SR), and the output of the ANN is the μnf. A comparison shows that the ANN calculates the laboratory data more accurately.

Keywords

• Dynamic viscosity
• Nano-lubricant
• Correlation
• ANN