Nondestructive identification of deformation mechanisms at the microscopic scale is a significant challenge, yet it holds the potential to advance the understanding of damage evolution across many areas of materials mechanics. In this study, we combine spectral analysis of acoustic emission (AE) waveforms with dimensionality reduction to cluster individual AE events occurring during the deformation of a magnesium single crystal. Based on prior knowledge, according to which twinning is preferred at low stresses and slip is expected to become dominant when twinning is exhausted, we link these clusters to their related deformation mechanisms. This enables unsupervised classification of each AE event into the two deformation mechanisms. Furthermore, based on resonance ultrasound spectroscopy and modal calculations, we identify a frequency signature in AE signals, corresponding to one of the sample's resonant modes, and we associate it with slip. This allows us to define a single traceable, physically based classification parameter that performs comparably to the machine learning–based clustering. The results of both classification methods are consistent with each other and align with the expected deformation behavior and the shape of the loading curve. At the same time, they reveal detailed quantitative information at the level of individual AE events, which uncovers the evolution of the transition from twinning-dominant to slip-dominant behavior.

Close

PrevNext

View figure in article

• Anelastic deformation
• Material failure
• Mechanical deformation
• Plastic deformation
• Plasticity