EUVIMEDEUVIMED entdecken
Crossref

Deep learning for performance prediction and inverse design of Helmholtz resonators

Yuechao Zhao, Guifei Yan, Zhangfei Ma, Songhao Tang

Physica Scripta · 20.03.2026 · 101 (11) · 116003

DOI: 10.1088/1402-4896/ae4f2f ↗

Kurz gefasst

Fachartikel in der Zeitschrift Physica Scripta, online erschienen am 19.03.2026. Verfasst von Yuechao Zhao und 3 weiteren Autor:innen. Die Arbeit nennt 57 Literaturangaben. Gefördert von: Basic Scientific Research Operating Fee for Provincial Undergraduate Universities in Heilongjiang Province.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

To address the growing demand for customized Helmholtz acoustic metamaterials, a deep learning driven framework is developed for rapid performance prediction and on demand inverse design. In the first stage, a theoretical model of Helmholtz resonant cavities is established, and its validity is confirmed through finite element simulations and experimental measurements. Based on the theoretical model, a comprehensive dataset is generated, and a deep learning prediction model combining a multilayer perceptron with an attention mechanism is constructed to achieve accurate mapping from geometric parameters to sound absorption spectra. In addition, an inverse design model based on an autoencoder architecture is developed to infer structural parameters from desired sound absorption curves, thereby establishing a bidirectional relationship between acoustic performance and geometry. The proposed deep learning framework shows high prediction accuracy across a broad frequency range, with the inversely derived geometric parameters exhibiting excellent agreement with theoretical calculations. Moreover, the reconstructed absorption spectra closely match the target curves. These findings confirm that the deep learning based methodology provides an efficient, accurate, and generalizable approach for both forward and inverse design of Helmholtz resonant structures, offering a promising route for the rapid development and practical implementation of next generation acoustic metamaterials.

Förderung

  • Basic Scientific Research Operating Fee for Provincial Undergraduate Universities in Heilongjiang Province

Zitieren

APA

Zhao, Y., Yan, G., Ma, Z., & Tang, S. (2026). Deep learning for performance prediction and inverse design of Helmholtz resonators. Physica Scripta, 101(11), 116003. https://doi.org/10.1088/1402-4896/ae4f2f

Vancouver

Zhao Y, Yan G, Ma Z, Tang S. Deep learning for performance prediction and inverse design of Helmholtz resonators. Physica Scripta. 2026;101(11):116003. doi:10.1088/1402-4896/ae4f2f

BibTeX
@article{Zhao2026,
  author = {Zhao, Yuechao and Yan, Guifei and Ma, Zhangfei and Tang, Songhao},
  title = {Deep learning for performance prediction and inverse design of Helmholtz resonators},
  journal = {Physica Scripta},
  year = {2026},
  volume = {101},
  number = {11},
  pages = {116003},
  doi = {10.1088/1402-4896/ae4f2f}
}
Kostenlos starten →