The mechanical behavior of most materials involves discrete “avalanche” events, where a local microscopic volume within the bulk undergoes large abrupt deformations. Such events are commonly detected and analyzed based on the acoustic emission (AE) they generate. Many materials display several deformation mechanisms simultaneously, thus calling for analysis tools that can differentiate between the AE signals they produce. We introduce a new classification method where the power-spectral densities of AE signals are mapped onto a two-dimensional space by the $t$-distributed stochastic neighbor embedding (tSNE) algorithm. To study the effectiveness and accuracy of these tools, we produced ground-truth datasets of AE signals generated by dislocation plasticity, deformation twinning, and martensitic transformation. This is achieved by loading three different alloys, each characterized by a single deformation mechanism. We show that each dataset has different distributions of AE features, thus strengthening the necessity to distinguish between AE signals generated by different deformation mechanisms. Moreover, an analysis that does not separate deformation mechanisms, and instead considers the combined datasets, provides distributions that lead to specious conclusions. The two-dimensional representation applied by our data-driven method displays three distinct clusters, which are also identified by the $k$-means clustering algorithm. The clusters match the ground-truth datasets with an accuracy of 98.6%, implying the method's capability to distinguish between different deformation mechanisms that coexist in the same material.

Close

PrevNext

View figure in article

• Acoustic measurements
• Mechanical deformation
• Phase transitions
• Plastic deformation
• Acoustic techniques