Predicting yielding in metallic glasses remains challenging because plasticity emerges without clear structural precursors. Here, a physically grounded Bayesian framework is introduced to predict the stress-strain response up to the yield point using plastic strain accumulation already in the nominally elastic regime. Across Cu-Zr(-Al) metallic glasses with varying annealing, two limiting growth laws for plastic strain, power-law and exponential, are identified and linked to distinct microscopic plastic activity patterns. By inferring these growth parameters from stress-strain data below 5% strain, the approach enables early, interpretable predictions of macroscopic deformation and failure.
Mäkinen et al. (Tue,) studied this question.