EUVIMEDEUVIMED entdecken
Crossref

Lightweight LSTM-Optimized Time-Difference Mapping Method for Precise Acoustic-Emission Source Localization in Composite Materials

Xiaoran Wang, Hongyu Li, Zenghua Liu, Jianing Zhang, Yuxuan Zhang, Sebastian Bader

IEEE Transactions on Instrumentation and Measurement · 2026 · 75 · 9703512-9703512

DOI: 10.1109/tim.2026.3711291 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Instrumentation and Measurement, 2026. Verfasst von Xiaoran Wang und 5 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 59 Literaturangaben. Gefördert von: National Key Research and Development Program of China, National Natural Science Foundation of China, Knowledge Foundation.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

A lightweight long short-term memory (LSTM)-optimized time-difference mapping (TDM) framework is presented for acoustic emission (AE) source localization in composite materials, establishing a systematic hardware–software co-design approach that evaluates the tradeoff between localization precision and embedded resource constraints. The TDM-derived features are integrated with an LSTM regression model to compensate for anisotropic wave propagation and multipath dispersion inherent to composite structures. To address the rigid memory boundaries of microcontroller units (MCUs), an orthogonal experimental design is employed to optimize key hyperparameters, balancing deep model capacity against hardware limits. Model-compression techniques, including posttraining int8 quantization, are subsequently applied to produce a lightweight tiny machine learning model. Deployed with TensorFlow Lite for microcontrollers on an ARM Cortex-M33 MCU, the quantized network occupies 78.9 kB of flash and 6.1 kB of RAM. Empirical embedded evaluation demonstrates a viable balance between competing metrics, achieving a mean localization error of 10.89 mm, an inference latency of 2.8 ms, and a per-inference energy footprint of $143.9~\mu $ J. These results validate the feasibility of embedding high-fidelity sequence-regression methods into resource-constrained edge hardware for decentralized structural health monitoring.

Förderung

  • National Key Research and Development Program of China
  • National Natural Science Foundation of China
  • Knowledge Foundation
  • Nanjing Institute of Industry Technology

Zitieren

APA

Wang, X., Li, H., Liu, Z., Zhang, J., Zhang, Y., & Bader, S. (2026). Lightweight LSTM-Optimized Time-Difference Mapping Method for Precise Acoustic-Emission Source Localization in Composite Materials. IEEE Transactions on Instrumentation and Measurement, 75, 9703512-9703512. https://doi.org/10.1109/tim.2026.3711291

Vancouver

Wang X, Li H, Liu Z, Zhang J, Zhang Y, Bader S. Lightweight LSTM-Optimized Time-Difference Mapping Method for Precise Acoustic-Emission Source Localization in Composite Materials. IEEE Transactions on Instrumentation and Measurement. 2026;75:9703512-9703512. doi:10.1109/tim.2026.3711291

BibTeX
@article{Wang2026,
  author = {Wang, Xiaoran and Li, Hongyu and Liu, Zenghua and Zhang, Jianing and Zhang, Yuxuan and Bader, Sebastian},
  title = {Lightweight LSTM-Optimized Time-Difference Mapping Method for Precise Acoustic-Emission Source Localization in Composite Materials},
  journal = {IEEE Transactions on Instrumentation and Measurement},
  year = {2026},
  volume = {75},
  pages = {9703512--9703512},
  doi = {10.1109/tim.2026.3711291}
}
Kostenlos starten →