Scholarlink

<p>Wind turbine gearbox vibration signals under time-varying speeds suffer severe noise contamination and non-stationarity. To address this, this paper proposes a novel denoising method based on adaptive chirp mode decomposition (ACMD). Firstly, it adaptively optimized ACMD’s bandwidth and smoothing parameters via Bayesian optimization algorithm, using time-frequency kurtosis as the objective function. Secondly, the maximum points in the time-frequency representation are found and connected to obtain the initial instantaneous frequency as the input of ACMD to extract chirp modes (CM) more accurately. Finally, the adaptively extracted CMs are reconstructed to achieve signal denoising. Validated against denoising methods based on traditional decomposition methods, the proposed method overcomes mode aliasing, over-decomposition and parameter dependency. Moreover, ACMD-based denoising significantly outperforms CEEMDAN, VMD, and VNCMD by achieving higher SNR, lower RMSE, and greater CC for simulated signals, while it also yields lower PE for the experimental signals.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Study on denoising of time-varying rotational speed signals of wind turbine gearboxes based on adaptive chirp mode decomposition

  • 期刊论文

  • 发表日期：

    2026年3月9日

查看全文
  • Chenxin Yang,
  • Hongchen Su,
  • Haojie Bian,
  • Longkang Zhao,
  • Yuning Zhang

摘要

Wind turbine gearbox vibration signals under time-varying speeds suffer severe noise contamination and non-stationarity. To address this, this paper proposes a novel denoising method based on adaptive chirp mode decomposition (ACMD). Firstly, it adaptively optimized ACMD’s bandwidth and smoothing parameters via Bayesian optimization algorithm, using time-frequency kurtosis as the objective function. Secondly, the maximum points in the time-frequency representation are found and connected to obtain the initial instantaneous frequency as the input of ACMD to extract chirp modes (CM) more accurately. Finally, the adaptively extracted CMs are reconstructed to achieve signal denoising. Validated against denoising methods based on traditional decomposition methods, the proposed method overcomes mode aliasing, over-decomposition and parameter dependency. Moreover, ACMD-based denoising significantly outperforms CEEMDAN, VMD, and VNCMD by achieving higher SNR, lower RMSE, and greater CC for simulated signals, while it also yields lower PE for the experimental signals.