Advanced Machine Learning Techniques for Predicting Wear Performance in Graphene Oxide Particulate Interpenetrating Polymer Network Composites

Abstract

This research investigates the wear behavior of hybrid polymeric composites made from synthetic glass and natural cotton fibers, reinforced with varying proportions of Graphene Oxide (GO) (0 %, 1 %, 3 %, 5 %, 7 %, 9 %). The effect of fiber arrangement and Graphene Oxide (GO) incorporation on wear rate and Coefficient of Friction (CoF) was evaluated using the Pin-On-Disk method, with analysis based on Taguchi's L32 Orthogonal Array. The optimal parameters were found at 6 min, 5 % GO, 300 revolutions per minute (rpm) speed, 20 mm (mm) track diameter, and 10 N (N) load, achieving a minimum wear rate of 0.612 × 10−4 cubic millimeters per newton-meter (mm3/N-m) and a CoF of 0.151. Predictive modeling was performed to predict the wear rate and coefficient of friction using supervised machine learning algorithms, including Linear Regression, Decision Tree, and Random Forest, to forecast material behavior. Performance evaluation using Confusion Matrix, Distribution Analysis, and various metrics showed that the Decision Tree model excelled, achieving near-perfect predictive power with a Mean Squared Error (MSE) of 0 and an R-squared value of 0.9999. The model demonstrated 100 % accuracy, with precision, recall, and F1-scores all equal to 1. This research demonstrates the effectiveness of combining natural and synthetic fibers with GO, along with the predictive power of machine learning in optimizing material properties.