A well-trained feed-forward perceptron Artificial Neural Network (ANN) for prediction the dynamic viscosity of Al2O3–MWCNT (40:60)-Oil SAE50 hybrid nano-lubricant at different volume fraction of nanoparticles, temperatures, and shear rates

Abstract

In this study, the influence of volume fraction of nanoparticles (φ) and temperature on dynamic viscosity (μnf) of Al2O3 –MWCNT (40:60) - Oil SAE50 hybrid nano-lubricant was analyzed. For this reason, μnf of a new hybrid nano-lubricant has derived for 174 various experiments through a series of experimental tests including a combination of different φ, temperatures and shear rates (γ̇). A well trained Artificial Neural Network (ANN) is created using the trainbr/trainlm algorithm and showed an MSE value of 3.58 along 0.999 as correlation coefficient for prediction of μnf. Variant Error diagrams and error histograms proved the appropriateness of the ANN as a tool for determining the μnf and the capability of the used training algorithms. In this analysis, the most effective factor on the μnf is the temperature and a decrease in temperature value has an impressive growth of μnf value for all φ. The results show that the φ has a negligible effect on the output parameter (μnf) especially in higher temperatures in a manner that by enhancing the φ, the μnfwill increase for all temperatures, while this enhance is more noticeable in lower temperatures. For example, enhancing the φ from 0 to 1%, almost will not change the μnf while this change of φ causes around 40% increase of μnf in T = 60 °C. Also, γ̇has a non-uniform influence on the μnf, in a sense that in lower γ̇, the μnf has much higher values for different combinations of temperature and φ; on the other hand, by increasing the γ̇, the μnfhas much lower values with narrower bandwidth for various temperatures and φ.