Tribological performance of graphene oxide reinforced PEEK nanocomposites with machine learning approach

Abstract

• PEEK-GO nanocomposite developed using Sol-Gel with varying GO concentrations. • Higher GO content improves wear resistance under lower loads and track diameters. • Friction coefficient decreases with larger track diameters and higher loads. • Extra tree and XGBoost models predict wear and friction with high accuracy. Nano composite materials exhibit a range of mechanical, chemical, electrical, optical, and catalytic properties, with nanoparticles enhancing characteristics such as wear resistance, corrosion resistance, specific strength/stiffness, friction coefficient, and high temperature strength. Tribology research focuses on the friction, wear, and lubrication of contacting surfaces. In this study, a polymer-based Nano composite reinforced with graphene oxide (GO) was developed to improve wear resistance and friction coefficient. Polymer specimens were produced by incorporating varying concentrations of graphene oxide (GO) (1, 3, and 5 wt percent) into the PEEK matrix using the Sol-Gel process. Wear rate and friction coefficient of the polymer Nano composite (PEEK-GO) were evaluated using a pin-on-disk machine at room temperature under different loads (20, 30, and 60) and track diameters (60, 90, and 100). Results indicate that increasing filler content (GO concentration) led to lower wear rates with decreasing loads and track diameters. Conversely, increasing track diameter and loads while reducing reinforcement contents (GO concentration) resulted in decreased friction coefficients. Additionally, the predictive performance of Extra Tree and XGBoost regression models in estimating wear and friction force was investigated using performance metrics such as MAE and R-squared, incorporating confidence intervals to quantify prediction uncertainty. Predictive modeling with Extra Tree and XGBoost regression techniques yielded MAE values of 5.21 and 8.03, respectively, for wear prediction with a 0.2 test size. For friction force prediction, MAE values were 0.77 (Extra Tree) and 0.76 (XGBoost). DFFITS analysis further indicated all wear and friction force data points were influential, remaining within a narrow interval (+0.57735 to -0.57735). Both models exhibited promising predictive capabilities across different test sizes for wear and friction force prediction, highlighting the significance of feature selection in improving model accuracy.

Topics & Concepts