Harder, Better, Faster, Stronger: Predicting Magnesium Alloy Hardness with Machine Learning
Description
Material hardness is crucial in engineering and manufacturing, affecting component performance and durability. This study examines magnesium (Mg) alloys, valued for their lightweight and stiffness properties, with applications in automotive, aerospace, electronics, and medical industries. We conducted Vickers Hardness (VH) testing to analyze the effects of alloy composition, microstructure (phase composition and grain size), and rolling on hardness. The study focused on Mg-6Al (6 wt% aluminum) in both cast and rolled conditions. Results showed that rolling increased VH, demonstrating how structural modifications impact hardness. Additionally, machine learning (ML) was employed to predict the hardness of Mg-6Al alloys. Hardness, defined as resistance to localized permanent deformation, depends on grain size, precipitates, dislocation density, and phase composition. The Mg alloy was cast and rolled at NC A&T, with 257 VH data points collected. Using MATLAB’s Regression Learner, Gaussian Process Regression (GPR) showed the best predictive performance. ML integration aims to improve hardness prediction accuracy, reduce testing errors, and enhance efficiency and cost-effectiveness in material evaluation.