Scholarlink

This paper examines the use of autoencoders to identify structural changes in a prestressed hybrid tower in segmental construction. The foundation of concrete steel towers was instrumented with 30 accelerometers spread across 15 levels and exposed to different prestress forces ranging from 0 to 1,000 kN. In contrast to earlier studies, the tower does not show symmetrical bending directions due to the repair of the lowest segment. Ambient vibration data sets were recorded at a sampling rate of 1,000 Hz and analyzed using modal analysis to determine the natural frequencies of the first three vibration modes. For monitoring, sequences of raw time-domain signals serve as the basis for training and testing the autoencoder to reveal changes in stiffness properties. The study establishes a physical basis for detecting structural changes by confirming that changes in stiffness correlate with changes in the autoencoder’s residuals. These results highlight the potential of data-driven structural health monitoring techniques, particularly autoencoder-based approaches, for real-time monitoring and damage detection in wind turbines and other critical infrastructure.

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

Monitoring Different Prestressing Forces of a Concrete Steel Tower Using Neuronal Networks

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Niklas Römgens,
  • Abderrahim Abbassi,
  • Muawiya Talahmeh,
  • Tanja Grießmann,
  • Raimund Rolfes

摘要

This paper examines the use of autoencoders to identify structural changes in a prestressed hybrid tower in segmental construction. The foundation of concrete steel towers was instrumented with 30 accelerometers spread across 15 levels and exposed to different prestress forces ranging from 0 to 1,000 kN. In contrast to earlier studies, the tower does not show symmetrical bending directions due to the repair of the lowest segment. Ambient vibration data sets were recorded at a sampling rate of 1,000 Hz and analyzed using modal analysis to determine the natural frequencies of the first three vibration modes. For monitoring, sequences of raw time-domain signals serve as the basis for training and testing the autoencoder to reveal changes in stiffness properties. The study establishes a physical basis for detecting structural changes by confirming that changes in stiffness correlate with changes in the autoencoder’s residuals. These results highlight the potential of data-driven structural health monitoring techniques, particularly autoencoder-based approaches, for real-time monitoring and damage detection in wind turbines and other critical infrastructure.