Answers are grounded in this paper and cite their sources.
This work explores a novel method of characterising an array of cracks using the scatter field of the interacting Rayleigh waves. Numerically simulated line-scans with a stationary transmitter and moving receiver were used to extract scatter field data from defective and intact samples, where the latter provides the baseline response. To mimic experimental data collection, white Gaussian noise was added to create ensembles with different signal-to-noise ratios (SNRs). The crack array was characterised using two parameters: (i) depth of cracks (200−500 µm) and (ii) number of cracks (3−7). The scatter field resulting from different combinations of these parameters was further featurized using the loss of correlation (Locor). The features were trained using different machine learning regression models, which were evaluated for estimating the depth and the number of cracks in an array. A multi-output decision tree regressor using the scatter field features had the best accuracy of >90%, compared to using the traditional transmission coefficient as a feature, which only had an accuracy of 50%. Finally, a feature-importance study revealed that the local resonances in the crack array may contribute to the increased sensitivity.
No takes yet. Share an insight, caveat, or question.
Sharma et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: