Scholarlink

<p>Utility tunnels are essential components of modern urban infrastructure, and ensuring their safe operation requires real-time, comprehensive monitoring. Utility tunnels generate distinctive acoustic signals during operation and in the event of faults, and the enclosed structure facilitates efficient acoustic signal acquisition. Therefore, we proposed an intelligent monitoring method for utility tunnels based on voiceprint recognition in the paper by taking advantage of non-contact and real-time monitoring. First, the acoustic characteristics present in utility tunnels were analyzed, and an experimental platform was built to collect the voiceprint data. Then multiple deep learning models were designed and evaluated for their ability to recognize the voiceprint data, with the highest recognition accuracy exceeding 99%. Furthermore, the robustness of the proposed method under noisy conditions was verified, achieving recognition accuracies above 90% in the interference scenarios. Finally, the method was applied in a pilot-scale utility tunnel. Testing results show that the recognition accuracy exceeds 95%. These results confirm the feasibility and practical applicability of voiceprint-based intelligent monitoring for utility tunnel operation and fault detection.</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

Intelligent acoustic monitoring and fault recognition in utility tunnels: a hybrid E-VMD and CNN-based approach

  • 期刊论文

  • 发表日期：

    2026年7月8日

查看全文
  • Qi-wen Tian,
  • Hui-qing Lan,
  • Nan Lin,
  • Fei Wang,
  • Xin Zhao

摘要

Utility tunnels are essential components of modern urban infrastructure, and ensuring their safe operation requires real-time, comprehensive monitoring. Utility tunnels generate distinctive acoustic signals during operation and in the event of faults, and the enclosed structure facilitates efficient acoustic signal acquisition. Therefore, we proposed an intelligent monitoring method for utility tunnels based on voiceprint recognition in the paper by taking advantage of non-contact and real-time monitoring. First, the acoustic characteristics present in utility tunnels were analyzed, and an experimental platform was built to collect the voiceprint data. Then multiple deep learning models were designed and evaluated for their ability to recognize the voiceprint data, with the highest recognition accuracy exceeding 99%. Furthermore, the robustness of the proposed method under noisy conditions was verified, achieving recognition accuracies above 90% in the interference scenarios. Finally, the method was applied in a pilot-scale utility tunnel. Testing results show that the recognition accuracy exceeds 95%. These results confirm the feasibility and practical applicability of voiceprint-based intelligent monitoring for utility tunnel operation and fault detection.