Answers are grounded in this paper and cite their sources.
Ultrasonic guided waves have become a prominent research focus in structural health monitoring (SHM) for their high defect sensitivity and suitability for large-scale inspection. However, conventional guided wave imaging localization methods mostly adopt a two-stage “feature extraction-imaging computation” architecture, lacking deep synergy between multi-scale features and elliptical probability mapping, which restricts defect localization accuracy.To address this limitation, this paper proposes an improved elliptical probability imaging localization method assisted by Gaussian physical calibration and Siamese networks. A Gaussian path weighting mechanism based on spatial wave path difference attenuation is established to convert damage identification into a continuous probability regression problem.Ablation experiments on a public dataset verify the positive contribution of each core module. Results show the proposed method achieves an average central localization error of 10.45 mm, outperforming both conventional imaging algorithms (14.8 mm, 12.59 mm) and state-of-the-art deep learning models (12.43 mm, 11.31 mm, 10.71 mm). After parameter convergence calibration via offline supervised training, the method can automatically output path damage weights and generate localisation results during the inference phase without manual intervention or prior damage information. This approach provides a feasible technical pathway for online health monitoring of high-end composite structures.
No takes yet. Share an insight, caveat, or question.
Liu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: