Scholarlink

<p>In view of the problem that the high noise and data redundancy in the voiceprint signal of the wind turbine blade lead to insufficient diagnostic accuracy and real-time performance and increase the acquisition cost, this paper combines sparse representation, compressed sensing, and deep learning technology to apply a new wind turbine blade damage detection method, aiming to enhance the accuracy and real-time performance of wind turbine blade damage diagnosis. The sparse representation method is used to effectively encode the voiceprint signal and extract representative signal features; the compressed sensing technology is applied to efficiently reconstruct the signal using a small amount of sampled data, significantly reducing the data collection amount and storage requirements; deep feature learning and damage pattern classification based on convolutional neural network further improve the accuracy and intelligence level of detection.The research results show that the proposed method effectively reduces the computational complexity and greatly improves the detection accuracy. The accuracy is not less than 88% under five damage types: crack, corrosion, deformation, fatigue and impact. It has good adaptability under different computing resources, and the processing delay does not exceed 0.45s under complex environments and large data volumes. It has strong real-time performance and application potential.</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

Research on damage detection technology for wind turbine blade acoustic signals by fusion of sparse representation, compressive sensing and deep learning

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年7月7日

查看全文
  • Liang Wang,
  • Chun Yang,
  • Chao Yuan,
  • Yanan Liu,
  • Yanqing Chen

摘要

In view of the problem that the high noise and data redundancy in the voiceprint signal of the wind turbine blade lead to insufficient diagnostic accuracy and real-time performance and increase the acquisition cost, this paper combines sparse representation, compressed sensing, and deep learning technology to apply a new wind turbine blade damage detection method, aiming to enhance the accuracy and real-time performance of wind turbine blade damage diagnosis. The sparse representation method is used to effectively encode the voiceprint signal and extract representative signal features; the compressed sensing technology is applied to efficiently reconstruct the signal using a small amount of sampled data, significantly reducing the data collection amount and storage requirements; deep feature learning and damage pattern classification based on convolutional neural network further improve the accuracy and intelligence level of detection.The research results show that the proposed method effectively reduces the computational complexity and greatly improves the detection accuracy. The accuracy is not less than 88% under five damage types: crack, corrosion, deformation, fatigue and impact. It has good adaptability under different computing resources, and the processing delay does not exceed 0.45s under complex environments and large data volumes. It has strong real-time performance and application potential.