Scholarlink

This paper introduces an effective method for damage detection by applying deep learning (DL), which combines the ability to extract the characteristics of the 1D convolutional neural network (1DCNN) with the ability of the bidirectional gated recurrent unit (BiGRU) to process sequential data from both directions. Although 1DCNN has an advantage in extracting characteristics from sequential data, it has limitations in dealing with long-term relationships in time series data. Therefore, BiGRU is employed. To test the effectiveness of the proposed method, termed BiGRU-1DCNN, this study applies a method to determine the damage to the Chuong Duong bridge’s numerical model. The results show that BiGRU-1DCNN is superior to two traditional DL models, 1DCNN and BiGRU, with accuracy achieved at 81.5%, 68.9%, and 67.7% on the validation set, respectively.

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

An Effective Damage Detection Approach for a Truss Bridge Using a Hybrid Deep Learning Model

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Trung-Vu Manh,
  • Hoa-Tran Ngoc,
  • Manh-Tong Duc,
  • Loc-Bui Phuc,
  • Luong-Nguyen Duc

摘要

This paper introduces an effective method for damage detection by applying deep learning (DL), which combines the ability to extract the characteristics of the 1D convolutional neural network (1DCNN) with the ability of the bidirectional gated recurrent unit (BiGRU) to process sequential data from both directions. Although 1DCNN has an advantage in extracting characteristics from sequential data, it has limitations in dealing with long-term relationships in time series data. Therefore, BiGRU is employed. To test the effectiveness of the proposed method, termed BiGRU-1DCNN, this study applies a method to determine the damage to the Chuong Duong bridge’s numerical model. The results show that BiGRU-1DCNN is superior to two traditional DL models, 1DCNN and BiGRU, with accuracy achieved at 81.5%, 68.9%, and 67.7% on the validation set, respectively.