Prospective research on the tribological behavior of graphdiyne nanofluid and its machine learning performance prediction

Graphdiyne (GDY), an emerging type of 2D carbon nanomaterial, was introduced to the field of friction and lubrication in the present study. Water-based GDY nanofluid was prepared and the lubrication performance was studied using pin-on-disk tribometer. Meanwhile, four machine learning algorithms containing linear regression (LR), k-nearest neighbors (KNN), random forest (RF) and XGB (eXtreme Gradient Boosting) were used to predict the tribology behavior of GDY nanofluid. At the optimal concentration (0.3 wt%) of GDY in nanofluid, the COF and wear rate was reduced by about 18.0 % and 49.5 %, respectively, compared to those of base-fluid. The RF algorithm showed the best predict accuracy and stability for the pin-on-disk tribology behavior of GDY nanofluid, followed by XGB that was also acceptable. A dense and uniform tribofilm in the thickness of about 6.4 nm has generated at the friction interface, which ultra-fine GDY nanoparticles, sintered polycrystalline carbonizing compounds and iron oxides. The tribofilm as well as the polishing effect, mending effect and interlayer effect of nanoparticles contributed the superior lubrication performance of GDY nanofluid. Our study extends the application scenario of GDY to the field of friction and lubrication, and artificial intelligence method is promoted in performance prediction and directional design of lubricants.