Using artificial intelligence to predict the tribology behavior of MoS2-Al2O3 hybrid nanofluid

Abstract

Abstract Artificial intelligence algorithms including two artificial neural network and two machine learning algorithms were employed to predict the four-ball tribology behavior of MoS2-Al2O3 hybrid nanofluid. MoS2-Al2O3 composite nanoparticles were synthesized using solvothermal method and then dispersed in water-based fluids. 27 groups of tribology tests were conducted according to Box-Behnken experimental design were set as the training groups. The input variables (velocity of friction pairs, test force, test temperature, nanoparticle concentration) and output parameters (friction coefficient, wear scar diameter, wear surface roughness) were selected as the main variables. It was found that the random forest (RF) had better predict accuracy and stability for the four-ball tribology behavior of MoS2-Al2O3 nanofluid than multilayer perceptron (MLP), back propagation (BP) and k-nearest neighbors (KNN) algorithms. Besides, Pearson correlation analysis was carried out to reveal the relationship between input and output as well as different output variables. Through in-depth characterization of worn surface, a tribofilm in the thickness of 15 ∼ 20 nm composed of amorphous phases, ultra-fine nanoparticles and iron compounds was found. Finally, the lubrication mechanism of MoS2-Al2O3 nanofluid were discussed based on analyzing the tribology behavior data and tribofilm structure. Through the above findings, we hope to promote the application and development of artificial intelligence techniques in lubricants design and performance evaluation in the future.