A machine learning framework to predict thermoplastic deformation behavior and rheological regime transition in bulk metallic glass composites

Abstract

Understanding the deformation behavior of materials is critical for the effective manufacturing of formed parts. The thermoplastic deformation behavior (TDB) of bulk metallic glass composites (BMGCs) is influenced by numerous process parameters, and the complex nonlinear interactions present significant challenges for accurate description using phenomenological or physics-based constitutive models. The success of machine learning (ML) in various fields suggests its potential to address such complexities in materials science. In this study, we apply four ML regression algorithms to predict the TDB of (Zr₅₅Cu₃₀Al₁₀Ni₅)₉₄Ta₆ BMGCs. Our results demonstrate that the Extra Trees algorithm provides the most accurate predictions, with an analysis of its performance based on its underlying principles and the material's deformation behavior. Using the well-trained model, we generate a strain rate sensitivity index contour plot, revealing the transition between Newtonian and non-Newtonian rheological regime.