Scholarlink

<p>This paper introduces a non-destructive inspection paradigm for reinforced concrete that integrates Magnetic Resonance Eddy-Current Testing (MRECT) with a bidirectional Gated Recurrent Unit (GRU) dual-target regressor. The MRECT probe excites resonant magnetic fields whose near-field coupling with embedded steel yields an order-of-magnitude increase in sensitivity to rebar presence and geometry. By spatially scanning resonant-voltage signatures, the system synthesizes a high-resolution map of the hidden reinforcement grid without surface preparation. Concurrently, the GRU-based post-processor lifts the resonant-voltage signatures acquired directly above each rebar into a pre-trained, learned feature space whose coordinates seamlessly encode continuous variations in bar diameter and embedment depth, enabling single-shot, high-fidelity prediction of both parameters. Extensive validation on 120 concrete slabs (cover 2–5&#xa0;cm, diameters 6–25&#xa0;mm) delivers Mean Absolute Percentage Errors (MAPE) of 2.41% for diameter and 2.44% for depth—outperforming ISO 1920–7 benchmarks for cover meters and ground-penetrating radar. The resulting rapid, single-sided, contactless protocol provides a laboratory-validated technical prototype, offering promising potential for future application in construction quality assurance and long-term structural health monitoring of concrete infrastructure.</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

Magnetic-Resonance Eddy-Current Test Enhanced by a Bidirectional GRU Network for Simultaneous Prediction of Rebar Diameter and Embedment Depth in Concrete

  • 期刊论文

  • 发表日期：

    2026年5月13日

查看全文
  • Jie Zhu,
  • Leng Liao,
  • Jinming Zhang

摘要

This paper introduces a non-destructive inspection paradigm for reinforced concrete that integrates Magnetic Resonance Eddy-Current Testing (MRECT) with a bidirectional Gated Recurrent Unit (GRU) dual-target regressor. The MRECT probe excites resonant magnetic fields whose near-field coupling with embedded steel yields an order-of-magnitude increase in sensitivity to rebar presence and geometry. By spatially scanning resonant-voltage signatures, the system synthesizes a high-resolution map of the hidden reinforcement grid without surface preparation. Concurrently, the GRU-based post-processor lifts the resonant-voltage signatures acquired directly above each rebar into a pre-trained, learned feature space whose coordinates seamlessly encode continuous variations in bar diameter and embedment depth, enabling single-shot, high-fidelity prediction of both parameters. Extensive validation on 120 concrete slabs (cover 2–5 cm, diameters 6–25 mm) delivers Mean Absolute Percentage Errors (MAPE) of 2.41% for diameter and 2.44% for depth—outperforming ISO 1920–7 benchmarks for cover meters and ground-penetrating radar. The resulting rapid, single-sided, contactless protocol provides a laboratory-validated technical prototype, offering promising potential for future application in construction quality assurance and long-term structural health monitoring of concrete infrastructure.