EUVIMEDEUVIMED entdecken
Crossref

PMSA-DGNet: A Physical Property-Guided Multiscale Spatial Attention Domain Generalization Network for Tool Wear Prediction

Tengfei Jiang, Zisheng Li, Xiaoping Xiao, Xuwei Lai, Honghao Fu, Yu Chen

IEEE Transactions on Instrumentation and Measurement · 2026 · 75 · 1-14

DOI: 10.1109/tim.2026.3661705 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Instrumentation and Measurement, 2026. Verfasst von Tengfei Jiang und 5 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 48 Literaturangaben. Gefördert von: Key Research and Development Program of Sichuan Province, Southwest University of Science and Technology Doctoral Fund Program.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Tool wear monitoring (TWM) is essential for ensuring machining quality and production safety in machine tools. However, data distribution differences arising from inconsistencies in process elements limit the cross-domain generalization (DG) capability of deep learning (DL) models. Existing DG methods can enhance model generalization performance, but they fail to fully account for the multiscale characteristics of monitoring signals. To address this, this article proposes a physical property-guided multiscale spatial attention DG network (PMSA-DGNet) for tool wear prediction. First, a multiscale spatial attention mechanism constrained by physical properties is designed to adaptively enhance the representational capability of cutting signal segments during tool wear stages while effectively suppressing interference from noncutting segments. Second, domain adversarial training combined with maximum mean discrepancy (MMD) is employed to learn domain-invariant feature representations from multiple source domains. Third, an adaptive weight learning strategy is introduced to stabilize the adversarial training process and reduce the cost of hyperparameter tuning. The proposed method is validated on both an open dataset and a self-collected dataset under various working conditions, achieving average $R {^{{2}}}$ values of 0.9398 and 0.9116, respectively. Furthermore, visualizations of the multiscale spatial attention distribution provide interpretability.

Förderung

  • Key Research and Development Program of Sichuan Province
  • Southwest University of Science and Technology Doctoral Fund Program

Zitieren

APA

Jiang, T., Li, Z., Xiao, X., Lai, X., Fu, H., & Chen, Y. (2026). PMSA-DGNet: A Physical Property-Guided Multiscale Spatial Attention Domain Generalization Network for Tool Wear Prediction. IEEE Transactions on Instrumentation and Measurement, 75, 1-14. https://doi.org/10.1109/tim.2026.3661705

Vancouver

Jiang T, Li Z, Xiao X, Lai X, Fu H, Chen Y. PMSA-DGNet: A Physical Property-Guided Multiscale Spatial Attention Domain Generalization Network for Tool Wear Prediction. IEEE Transactions on Instrumentation and Measurement. 2026;75:1-14. doi:10.1109/tim.2026.3661705

BibTeX
@article{Jiang2026,
  author = {Jiang, Tengfei and Li, Zisheng and Xiao, Xiaoping and Lai, Xuwei and Fu, Honghao and Chen, Yu},
  title = {PMSA-DGNet: A Physical Property-Guided Multiscale Spatial Attention Domain Generalization Network for Tool Wear Prediction},
  journal = {IEEE Transactions on Instrumentation and Measurement},
  year = {2026},
  volume = {75},
  pages = {1--14},
  doi = {10.1109/tim.2026.3661705}
}
Kostenlos starten →