Tensile damage identification of high-strength CrMoV steel based on acoustic emission and machine learning

The accurate decoding and classification of acoustic emission (AE) signals associated with different damage mechanisms are critical for reliable structural integrity assessment. In this study, damage identification in high-strength CrMoV steel under tensile deformation at varying strain rates was investigated through the integration of AE analysis with machine learning techniques. The results showed that four distinct damage stages were identified through multivariate AE parameters: elastic deformation with microplasticity (stage 1), yielding and strain hardening (stage 2), localized necking (stage 3), and crack growth (stage 4). Furthermore, machine learning models incorporating multivariate AE parameters exhibited good performance in classifying AE signals across different damage stages and achieved high accuracy in identifying crack growth signals. Notably, the K-Nearest Neighbors (KNN) and Random Forests (RF) models attained the highest classification accuracy of 79.59 % for crack growth identification among all tested algorithms. Additionally, the influence of strain rate on tensile properties and AE behavior was examined through microstructural characterization. A pronounced correlation was observed between average kernel average misorientation (KAM) values and key AE parameters, with higher KAM values corresponding to elevated AE parameter levels. This correlation confirmed that the progression of plastic deformation primarily governed AE signal characteristics throughout deformation stages 1–3. This work will provide an approach for the precise identification of damage mechanisms by integrating AE monitoring with machine learning, offering potential applications in structural health monitoring and failure prediction.