Taguchi and machine learning integration for tribological analysis of polyurethane/nanodiamond nanocomposites

Abstract

This study investigates the tribological behavior of thermoplastic polyurethane (PU) reinforced with nanodiamonds (ND) at concentrations ranging from 0.1 to 0.5 wt.%. The wear rate and coefficient of friction (COF) of PU/ND nanocomposites were evaluated using experimental, Taguchi, and machine learning methods. Wear tests were performed using a Pin-On-Disc apparatus under varying conditions, including sliding distances (500–1500 m), applied loads (10–30 N), and sliding speeds (100–300 rpm), designed according to the Taguchi L9 orthogonal array. The experimental results demonstrated that incorporating nanodiamonds (ND) into the PU matrix significantly reduced the wear rate. Among the parameters analyzed, Taguchi analysis identified sliding distance as the most critical factor influencing the wear rate and the coefficient of friction (COF), followed by applied load, ND content, and sliding speed. Machine learning models further validated and extended the experimental findings, showcasing robust predictive capabilities. The Mean Squared Error (MSE) for wear rate prediction ranged from 0.0005 (1.21%) using Linear Regression to 0.0004 (0.79%) with Random Forest, while for COF, the Linear Regression model yielded an MSE of 0.0161 (5.01%) and the Random Forest model achieved an even lower MSE of 0.0054 (1.68%). Integrating Taguchi methods with machine learning improved the precision of wear mechanism analysis, enabling accurate predictions. This approach provided a robust framework for optimizing the tribological performance of PU-ND composites. It holds great promise for advanced material design and practical applications.