Scholarlink

<p>Damage detection in steel structures requires robust and intelligent techniques to ensure reliability and efficiency. Traditional damage detection techniques rely on manual inspections, which can be labour-intensive, time-consuming, and prone to human mistakes. To address these difficulties, this paper develops an ensemble deep learning technique that combines Convolutional Neural Networks (CNN) with Long Short-Term Memory (LSTM) networks for joint damage detection in a steel frame structure. Moreover, the CNN part of the network pulls out spatial features from the data, and the LSTM part finds temporal relationships, which makes damage detection more accurate. For this purpose, a steel 3D frame structure is considered. We use an impact hammer to vibrate the frame and collect time-history acceleration data under both healthy and unhealthy configurations. Frequency-domain scalogram images are generated and used as input to the CNN-LSTM network. The training, validation, and testing accuracies are found to be 94.50%, 92.89%, and 91.25%., respectively. The outcomes show that the developed technique can easily distinguish healthy and unhealthy configurations. This research can enhance predictive maintenance, reduce downtime, and make it more accurate to identify joint damages in steel frame 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

A combined CNN-LSTM-based technique for joint damage detection in steel frame structures

  • 期刊论文

  • 发表日期：

    2025年3月18日

查看全文
  • Maloth Naresh,
  • Ashish Balavant Jadhav,
  • Yogesh S. Patil

摘要

Damage detection in steel structures requires robust and intelligent techniques to ensure reliability and efficiency. Traditional damage detection techniques rely on manual inspections, which can be labour-intensive, time-consuming, and prone to human mistakes. To address these difficulties, this paper develops an ensemble deep learning technique that combines Convolutional Neural Networks (CNN) with Long Short-Term Memory (LSTM) networks for joint damage detection in a steel frame structure. Moreover, the CNN part of the network pulls out spatial features from the data, and the LSTM part finds temporal relationships, which makes damage detection more accurate. For this purpose, a steel 3D frame structure is considered. We use an impact hammer to vibrate the frame and collect time-history acceleration data under both healthy and unhealthy configurations. Frequency-domain scalogram images are generated and used as input to the CNN-LSTM network. The training, validation, and testing accuracies are found to be 94.50%, 92.89%, and 91.25%., respectively. The outcomes show that the developed technique can easily distinguish healthy and unhealthy configurations. This research can enhance predictive maintenance, reduce downtime, and make it more accurate to identify joint damages in steel frame structures.