Scholarlink

<p>Deep-hole radial steel gates pose challenges for in-situ health monitoring. We develop a multi-sensor, image-based status monitoring and fault diagnosis framework that converts vibration signals into symmetric dot pattern (SDP) images and classifies them with a hybrid CNN-GRU. Using a full-scale gate from a hydraulic project, we conducted flow-induced vibration tests to reassess safety and to collect normal/abnormal data; a complementary damage dataset was built via hydraulic model testing, numerical simulation, and controlled defects. These datasets validate the approach. Despite strong noise, the method requires no pre-denoising and yields high accuracy in both tasks (status monitoring 99.41 %, fault diagnosis 99.14 %), outperforming constituent sub-models in classification accuracy and noise robustness. The proposed framework enables real-time health monitoring and intelligent fault diagnosis of deep-hole radial steel gates, offering practical support for the safe operation of large hydraulic structures.</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

Deep learning-based health monitoring of deep-hole radial steel gates using multi-sensor fusion and image transformation

  • 期刊论文

  • 发表日期：

    2026年7月10日

查看全文
  • Chen Wang,
  • Yakun Liu,
  • Di Zhang

摘要

Deep-hole radial steel gates pose challenges for in-situ health monitoring. We develop a multi-sensor, image-based status monitoring and fault diagnosis framework that converts vibration signals into symmetric dot pattern (SDP) images and classifies them with a hybrid CNN-GRU. Using a full-scale gate from a hydraulic project, we conducted flow-induced vibration tests to reassess safety and to collect normal/abnormal data; a complementary damage dataset was built via hydraulic model testing, numerical simulation, and controlled defects. These datasets validate the approach. Despite strong noise, the method requires no pre-denoising and yields high accuracy in both tasks (status monitoring 99.41 %, fault diagnosis 99.14 %), outperforming constituent sub-models in classification accuracy and noise robustness. The proposed framework enables real-time health monitoring and intelligent fault diagnosis of deep-hole radial steel gates, offering practical support for the safe operation of large hydraulic structures.