EUVIMEDEUVIMED entdecken
Crossref

Multi-Scale Decomposition and Feature Fusion of Sugarcane Harvester Cutter-Disk Vibration Signals for Operating-Condition Recognition

Chunming Wen, Bingxu Hou, Qingquan Liang, Shangping Li, Leilei Liu, Yang Cheng, Youzong Huang, Kaihua Li

Journal of the ASABE · 2026 · 69 (6) · 909-924

DOI: 10.13031/ja.16625 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of the ASABE, 2026. Verfasst von Chunming Wen und 7 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Highlights Five cutter-disk operating conditions were recognized from a single vibration sensor within one field campaign. Sample entropy merged nine IMFs and one residual into three Co-IMFs, cutting recurrent inputs by 70%. In within-campaign validation, CEEMDAN-SE-BiGRU reached 97.2 ± 0.3% accuracy and a Cohen's kappa of 0.965 ± 0.004 over 10 random-seed training runs. One 256-point window was processed in 125.4 ms, below the 0.64-s window update interval. ABSTRACT. Accurate recognition of cutter-disk operating conditions supports cutter-height adjustment and protection of sugarcane ratoons during mechanical harvesting. This study developed a five-class operating-condition recognition framework with vibration-signal reconstruction as an auxiliary learning task. Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) decomposed each vibration window into nine intrinsic mode functions (IMFs) and one residual, and a sample-entropy (SE) criterion derived from the training set grouped these components into three composite intrinsic mode functions (Co-IMFs). Three parallel bidirectional gated recurrent unit (BiGRU) branches processed the Co-IMFs; the fused branch features were used for five-class recognition, and an auxiliary head reconstructed the vibration velocity. The combined model is denoted CEEMDAN-SE-BiGRU. Field data were collected in Wuming District, Nanning, China, in December 2024, with one push-down whole-stalk sugarcane harvester operating on one cultivar and one soil type in two plots over two harvesting days. Complete acquisition runs were assigned to the training, validation, and test sets before windowing. In within-campaign validation, the model reached a mean accuracy of Keywords: Adaptive method, Condition monitoring, Subsurface cutting, Sugarcane harvesting, Vibration signal.

Zitieren

APA

Wen, C., Hou, B., Liang, Q., Li, S., Liu, L., Cheng, Y., Huang, Y., & Li, K. (2026). Multi-Scale Decomposition and Feature Fusion of Sugarcane Harvester Cutter-Disk Vibration Signals for Operating-Condition Recognition. Journal of the ASABE, 69(6), 909-924. https://doi.org/10.13031/ja.16625

Vancouver

Wen C, Hou B, Liang Q, Li S, Liu L, Cheng Y, et al. Multi-Scale Decomposition and Feature Fusion of Sugarcane Harvester Cutter-Disk Vibration Signals for Operating-Condition Recognition. Journal of the ASABE. 2026;69(6):909-924. doi:10.13031/ja.16625

BibTeX
@article{Wen2026,
  author = {Wen, Chunming and Hou, Bingxu and Liang, Qingquan and Li, Shangping and Liu, Leilei and Cheng, Yang and Huang, Youzong and Li, Kaihua},
  title = {Multi-Scale Decomposition and Feature Fusion of Sugarcane Harvester Cutter-Disk Vibration Signals for Operating-Condition Recognition},
  journal = {Journal of the ASABE},
  year = {2026},
  volume = {69},
  number = {6},
  pages = {909--924},
  doi = {10.13031/ja.16625}
}
Kostenlos starten →