EUVIMEDEUVIMED entdecken
Crossref

MMDRN: a multi-modal multi-scale dual residual network for ultrasonic guided wave signal denoising

Jiayang Yu, Qiang Wang, Gaocheng Chen, Ziling Xu, Sirong Tong, Qiuhan Liu

Measurement Science and Technology · 23.01.2026 · 37 (3) · 036105

DOI: 10.1088/1361-6501/ae309e ↗

Kurz gefasst

Fachartikel in der Zeitschrift Measurement Science and Technology, online erschienen am 07.01.2026. Verfasst von Jiayang Yu und 5 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 36 Literaturangaben. Gefördert von: China Postdoctoral Science Foundation.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Ultrasonic guided wave monitoring signals contain critical information for structural damage assessment. However, these signals are inevitably contaminated by noise in real-world acquisition environments, significantly compromising monitoring reliability. Therefore, this paper proposes a multi-modal multi-scale dual residual network for ultrasonic guided wave signal denoising. First, the original ultrasonic guided wave monitoring signal and its gramian angular difference field image are utilized as multi-modal inputs to realize complementary feature learning among different modalities. Second, the channel and spatial features of the signal are extracted by the sequential modality branch and the spatial modality branch, respectively. And the multi-scale feature extraction module and the multiple residual module in two branches are designed to enhance the feature extraction capability. In addition, a loss function is integrated by root mean square error, Jensen–Shannon divergence and extreme difference to enhance the robustness, interpretability and generalization ability of the network. Experimental results show that the proposed method improves denoising performance by 29.87%, 22.23%, 12.80%, and 5.38% compared to the second-best method for noise signals with 3 dB, 6 dB, 9 dB and 12 dB levels, which demonstrates the effectiveness and advancement of our proposed method.

Förderung

  • China Postdoctoral Science Foundation

Zitieren

APA

Yu, J., Wang, Q., Chen, G., Xu, Z., Tong, S., & Liu, Q. (2026). MMDRN: a multi-modal multi-scale dual residual network for ultrasonic guided wave signal denoising. Measurement Science and Technology, 37(3), 036105. https://doi.org/10.1088/1361-6501/ae309e

Vancouver

Yu J, Wang Q, Chen G, Xu Z, Tong S, Liu Q. MMDRN: a multi-modal multi-scale dual residual network for ultrasonic guided wave signal denoising. Measurement Science and Technology. 2026;37(3):036105. doi:10.1088/1361-6501/ae309e

BibTeX
@article{Yu2026,
  author = {Yu, Jiayang and Wang, Qiang and Chen, Gaocheng and Xu, Ziling and Tong, Sirong and Liu, Qiuhan},
  title = {MMDRN: a multi-modal multi-scale dual residual network for ultrasonic guided wave signal denoising},
  journal = {Measurement Science and Technology},
  year = {2026},
  volume = {37},
  number = {3},
  pages = {036105},
  doi = {10.1088/1361-6501/ae309e}
}
Kostenlos starten →