Scholarlink

This study investigated the mechanical performance of concrete produced with recycled aggregates derived from demolished lime-mortar masonry, a widely used construction material in Vietnam from the 1980s to the early 2000s. Three groups of specimens were prepared using recycled coarse and fine aggregates with 0%, 50%, and 100% replacement levels, and tested for compressive strength at 7 and 28 days. Additionally, a machine learning-based method was employed to analyze the fracture surfaces of the 28-day compressive specimens. The K-Nearest Neighbors (KNN) algorithm was used to segment cross-sectional images into coarse aggregate and mortar regions, enabling the calculation of the aggregate-to-mortar area ratio as an indirect indicator of material strength. The experimental and image-based results showed consistent trends, highlighting both the viability of using lime-based recycled aggregates in concrete and the potential of lightweight machine learning models to support quality assessment. This approach is particularly useful for rapid classification of input construction waste materials, aiding sustainable construction practices.

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

Machine Learning-Assisted Mechanical Characterization of Cementitious Composites with Recycled Lime-Mortar Aggregates

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Khai-Hoan Tran,
  • Duy-Hung Mac,
  • Nhu-Khoa Ngo

摘要

This study investigated the mechanical performance of concrete produced with recycled aggregates derived from demolished lime-mortar masonry, a widely used construction material in Vietnam from the 1980s to the early 2000s. Three groups of specimens were prepared using recycled coarse and fine aggregates with 0%, 50%, and 100% replacement levels, and tested for compressive strength at 7 and 28 days. Additionally, a machine learning-based method was employed to analyze the fracture surfaces of the 28-day compressive specimens. The K-Nearest Neighbors (KNN) algorithm was used to segment cross-sectional images into coarse aggregate and mortar regions, enabling the calculation of the aggregate-to-mortar area ratio as an indirect indicator of material strength. The experimental and image-based results showed consistent trends, highlighting both the viability of using lime-based recycled aggregates in concrete and the potential of lightweight machine learning models to support quality assessment. This approach is particularly useful for rapid classification of input construction waste materials, aiding sustainable construction practices.