Experimental Investigation and Machine Learning Modeling of Tribological Characteristics of AZ31/B₄C/GNPs Hybrid Composites

In this study, the AZ31 hybrid composites reinforced with boron carbide (B₄C) and graphene nano-platelets (GNPs) are prepared by the stir casting method. The main aim of the study is to study the effect of various wear parameters (reinforcement percentage (R), applied load (L), sliding distance (D), and velocity (V)) on the wear characteristics (wear rate (WR)) of the AZ91/B₄C/GNP composites. Experiments are designed using the Taguchi technique, and it was determined that load (L) is the most significant parameter affecting WR, followed by D, R, and V. The wear mechanisms under conditions of maximum and minimum wear rates are examined using SEM analysis of the worn-out surfaces of the specimens. From the result analysis on the WR, the ideal conditions for achieving the lowest WR are R = 4 wt.%, L = 15 N, V = 3 m/s, and D = 500 m. Machine learning (ML) models, including linear regression (LR), polynomial regression (PR), random forest (RF), and Gaussian process regression (GPR), are implemented to develop a reliable prediction model that forecasts output responses in accordance with input variables. A total of 90% of the experimental data points were used to train and 10% to evaluate the models. The PR model exceeded the accuracy of other models in predicting WR, with R² = 0.953, MSE = 0.011, RMSE = 0.103, and COF with R² = 0.937, MSE = 0.013, and RMSE = 0.114, respectively.