Scholarlink

Purpose <p>Structural cracks critically affect the durability and safety of concrete structures, making accurate crack detection vital for maintenance and design. This review aims to analyze over 100 studies (2005–2024) on beam crack detection using various machine learning algorithms.</p> Method <p>This study provides a comprehensive review of research employing machine learning techniques, including SVM, ANN, DNN, and ANFIS, to detect cracks in beam structures. The analysis evaluates the performance and effectiveness of both single and hybrid metaheuristic–machine learning models.</p> Results <p>The results demonstrate that hybrid metaheuristic–machine learning models, such as PSO–SVM and GA–ANN, achieve high accuracy levels of 97–99%. In comparison, single models (SVM, ANN, ANFIS) achieve accuracy levels of 93–97%, depending on the complexity of the datasets used.</p> Conclusion <p>The review highlights the strengths of various algorithms, addresses challenges related to data quality, and emphasizes the necessity of physics-informed learning to enhance model generalization. Consequently, machine learning serves as a promising tool for reliable and efficient crack detection in beam structures.</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

Detecting Beam Damage Using Machine Learning Algorithms—A Comparative Study

  • 期刊论文

  • 发表日期：

    2026年5月30日

查看全文
  • Saeedeh Ghaemifard,
  • Amin Ghannadiasl

摘要

Purpose

Structural cracks critically affect the durability and safety of concrete structures, making accurate crack detection vital for maintenance and design. This review aims to analyze over 100 studies (2005–2024) on beam crack detection using various machine learning algorithms.

Method

This study provides a comprehensive review of research employing machine learning techniques, including SVM, ANN, DNN, and ANFIS, to detect cracks in beam structures. The analysis evaluates the performance and effectiveness of both single and hybrid metaheuristic–machine learning models.

Results

The results demonstrate that hybrid metaheuristic–machine learning models, such as PSO–SVM and GA–ANN, achieve high accuracy levels of 97–99%. In comparison, single models (SVM, ANN, ANFIS) achieve accuracy levels of 93–97%, depending on the complexity of the datasets used.

Conclusion

The review highlights the strengths of various algorithms, addresses challenges related to data quality, and emphasizes the necessity of physics-informed learning to enhance model generalization. Consequently, machine learning serves as a promising tool for reliable and efficient crack detection in beam structures.