Scholarlink

This study introduces an innovative, field-oriented framework for Forensic Concrete Analysis (FCA) applicable to both historical and modern structures. Our approach accurately segments gray-toned coarse aggregates using advanced encoder-decoder models, while relying solely on standard camera images without color enhancements or microscopy. Specifically, we employ a Vision Transformer (SwinTV2 + SegFormer) and a convolutional Neural Network (CNN) (EfficientNetV2 + DeepLabV3+) to achieve high-precision segmentation. This enables detailed quantification of aggregate properties, such as area, size, and spatial distribution through Image Processing (IP) techniques, allowing analysis of concrete segregation influenced by varying water-cement ratios and Supplementary Cementitious Materials (SCMs). A strategy for evaluating structural health is proposed using aggregate proportions as indicators of mix consistency to enhance the FCA. A comparative study of three mix ratios (1:3:6, 1:2:4, and 1:1.5:3) assesses aggregate distribution using area aggregate ratio (AAR) from 2D image data. Results show a strong correlation between mean AAR the theoretical mix ratio values. The method provides a Nondestructive alternative to traditional techniques like ASTM C856, offering particular advantages for preserving historical structures. This robust, image-based framework supports effective structural health monitoring and maintenance planning.

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 Novel Image-Based Forensic Framework for Concrete in Historical and Modern Structures

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Afaq Ahmad,
  • Mati Ullah,
  • Vagelis Plevris,
  • Junaid Mir,
  • Syed Sameed Husain

摘要

This study introduces an innovative, field-oriented framework for Forensic Concrete Analysis (FCA) applicable to both historical and modern structures. Our approach accurately segments gray-toned coarse aggregates using advanced encoder-decoder models, while relying solely on standard camera images without color enhancements or microscopy. Specifically, we employ a Vision Transformer (SwinTV2 + SegFormer) and a convolutional Neural Network (CNN) (EfficientNetV2 + DeepLabV3+) to achieve high-precision segmentation. This enables detailed quantification of aggregate properties, such as area, size, and spatial distribution through Image Processing (IP) techniques, allowing analysis of concrete segregation influenced by varying water-cement ratios and Supplementary Cementitious Materials (SCMs). A strategy for evaluating structural health is proposed using aggregate proportions as indicators of mix consistency to enhance the FCA. A comparative study of three mix ratios (1:3:6, 1:2:4, and 1:1.5:3) assesses aggregate distribution using area aggregate ratio (AAR) from 2D image data. Results show a strong correlation between mean AAR the theoretical mix ratio values. The method provides a Nondestructive alternative to traditional techniques like ASTM C856, offering particular advantages for preserving historical structures. This robust, image-based framework supports effective structural health monitoring and maintenance planning.