Scholarlink

<p>Carbon fiber reinforced polymer (CFRP) cables are increasingly used in bridges, cableways, and other engineering structures due to their lightweight, high strength, and excellent corrosion resistance. However, accurately identifying and monitoring their damage modes under cyclic loading remains a critical challenge for structural health monitoring. In this study, progressive reciprocating tensile tests were conducted on CFRP cables and their constituent materials, while acoustic emission (AE) technique was employed to capture damage‑related signals. To achieve precise classification of damage modes, a Kepler Optimization Algorithm (KOA) enhanced k‑medoids clustering was first used to preliminarily identify the characteristic frequency ranges of matrix cracking and fiber breakage. Subsequently, a hybrid deep learning model integrating convolutional neural networks (CNN), bidirectional long short‑term memory networks (BiLSTM), and an attention mechanism, with hyperparameters optimized by KOA, was developed for accurate damage classification. The results demonstrate that the proposed KOA‑CNN‑BiLSTM‑Attention model achieves a classification accuracy exceeding 99.71% (validated by tenfold cross‑validation) and effectively distinguishes three damage modes: matrix cracking, debonding, and fiber fracture. The model’s predicted damage evolution shows excellent agreement with experimental observations of final failure behaviors, such as matrix explosion and fiber fracture. This work provides a reliable deep‑learning‑based tool for in‑situ damage monitoring of CFRP cables under service‑like cyclic loading.</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

Damage Classification of Carbon Fiber Reinforced Polymer Cables under Progressive Reciprocating Tensile Loading using a Hybrid Deep Learning Approach

  • 期刊论文

  • 发表日期：

    2026年4月29日

查看全文
  • Peng-Fei Zhang,
  • Lian-Hua Ma,
  • Ran Liu,
  • Wei Zhou

摘要

Carbon fiber reinforced polymer (CFRP) cables are increasingly used in bridges, cableways, and other engineering structures due to their lightweight, high strength, and excellent corrosion resistance. However, accurately identifying and monitoring their damage modes under cyclic loading remains a critical challenge for structural health monitoring. In this study, progressive reciprocating tensile tests were conducted on CFRP cables and their constituent materials, while acoustic emission (AE) technique was employed to capture damage‑related signals. To achieve precise classification of damage modes, a Kepler Optimization Algorithm (KOA) enhanced k‑medoids clustering was first used to preliminarily identify the characteristic frequency ranges of matrix cracking and fiber breakage. Subsequently, a hybrid deep learning model integrating convolutional neural networks (CNN), bidirectional long short‑term memory networks (BiLSTM), and an attention mechanism, with hyperparameters optimized by KOA, was developed for accurate damage classification. The results demonstrate that the proposed KOA‑CNN‑BiLSTM‑Attention model achieves a classification accuracy exceeding 99.71% (validated by tenfold cross‑validation) and effectively distinguishes three damage modes: matrix cracking, debonding, and fiber fracture. The model’s predicted damage evolution shows excellent agreement with experimental observations of final failure behaviors, such as matrix explosion and fiber fracture. This work provides a reliable deep‑learning‑based tool for in‑situ damage monitoring of CFRP cables under service‑like cyclic loading.