Scholarlink

<p>Corrosion-induced cracking in reinforced concrete (RC) structures significantly compromises structural integrity, leading to costly repairs and safety hazards. In the context of developing sustainable cities, ensuring the longevity of existing infrastructure is vital to minimize the environmental footprint of frequent reconstruction. Accurately predicting internal cracks remains challenging due to the complex interplay of material properties, environmental factors, and corrosion dynamics. This work proposes an innovative approach based on machine learning (ML) to analyse the relationship between surface indicators and the state of internal degradation. Unlike conventional studies, we rigorously compared four distinct architectures (ML) models such as Gaussian Process Regression (GPR), Random Forest (RF), Extra Trees Regressor (ETR), and Long Short-Term Memory (LSTM) to predict internal crack propagation patterns. Our methodology involves collecting experimental datasets, selecting critical features (e.g., current density, duration and maximum external crack width), training the models, and validating their performance using RMSE, MAE, and R² metrics. Results demonstrate that GPR outperforms others, achieving [MSE = 0.1111, R² = 0.7745]. This work provides a data-driven framework for proactive structural health monitoring, enabling timely maintenance and enhanced durability of RC 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

Prediction of internal cracking maps in corroded reinforced concrete structures using a machine learning approach

  • 期刊论文

  • 发表日期：

    2026年5月12日

查看全文
  • O. Loukil,
  • N. Affes,
  • A. Talha,
  • A. Daoud

摘要

Corrosion-induced cracking in reinforced concrete (RC) structures significantly compromises structural integrity, leading to costly repairs and safety hazards. In the context of developing sustainable cities, ensuring the longevity of existing infrastructure is vital to minimize the environmental footprint of frequent reconstruction. Accurately predicting internal cracks remains challenging due to the complex interplay of material properties, environmental factors, and corrosion dynamics. This work proposes an innovative approach based on machine learning (ML) to analyse the relationship between surface indicators and the state of internal degradation. Unlike conventional studies, we rigorously compared four distinct architectures (ML) models such as Gaussian Process Regression (GPR), Random Forest (RF), Extra Trees Regressor (ETR), and Long Short-Term Memory (LSTM) to predict internal crack propagation patterns. Our methodology involves collecting experimental datasets, selecting critical features (e.g., current density, duration and maximum external crack width), training the models, and validating their performance using RMSE, MAE, and R² metrics. Results demonstrate that GPR outperforms others, achieving [MSE = 0.1111, R² = 0.7745]. This work provides a data-driven framework for proactive structural health monitoring, enabling timely maintenance and enhanced durability of RC infrastructure.