Scholarlink

<p>Continuously Reinforced Concrete Pavement’s (CRCP) long-term performance and durability are significantly impacted by longitudinal cracking. In order to accurately predict the extent of longitudinal cracking using data from the Long-Term Pavement Performance (LTPP) database, this study proposes a hybrid machine learning approach that combines Genetic Algorithm (GA) optimization with Gradient Boosting Machine (GBM). Twenty input variables related to structure, traffic, climate, and initial pavement condition were incorporated into a total of 385 observations from 33 CRCP sections. Key GBM hyperparameters were optimized using the GA, which improved the model's functionality and capacity for generalization. With a mean RMSE of 4.55 and an R2 of 0.954 across five-fold cross-validation, the hybrid GA-GBM model outperformed conventional models including Linear Regression, Random Forest, Support Vector Regression (SVR), and Artificial Neural Networks (ANN). Age, Layer #3 Thickness, AADTT, KESAL, and Initial IRI were found to be the most significant predictors by feature importance analysis. The model's greatest responsiveness to Layer #3 Thickness, AADT, AADTT, Initial IRI, and Age was validated by sensitivity analysis, highlighting the importance of structural and traffic-related factors. Climate variables had a smaller influence, indicating a secondary contribution in this dataset. This study demonstrates the efficacy of hybrid machine learning algorithms for predicting pavement degradation and provides useful insights for prioritizing data factors in design and maintenance planning. The suggested GA-GBM framework is a viable and scalable tool for promoting proactive pavement management and enhancing infrastructure sustainability.</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

Optimized prediction of longitudinal cracking in concrete pavements using hybrid GA-GBM models

  • 期刊论文

  • 发表日期：

    2025年7月15日

查看全文
  • Ali Alnaqbi,
  • Ghazi G. Al-Khateeb,
  • Waleed Zeiada

摘要

Continuously Reinforced Concrete Pavement’s (CRCP) long-term performance and durability are significantly impacted by longitudinal cracking. In order to accurately predict the extent of longitudinal cracking using data from the Long-Term Pavement Performance (LTPP) database, this study proposes a hybrid machine learning approach that combines Genetic Algorithm (GA) optimization with Gradient Boosting Machine (GBM). Twenty input variables related to structure, traffic, climate, and initial pavement condition were incorporated into a total of 385 observations from 33 CRCP sections. Key GBM hyperparameters were optimized using the GA, which improved the model's functionality and capacity for generalization. With a mean RMSE of 4.55 and an R2 of 0.954 across five-fold cross-validation, the hybrid GA-GBM model outperformed conventional models including Linear Regression, Random Forest, Support Vector Regression (SVR), and Artificial Neural Networks (ANN). Age, Layer #3 Thickness, AADTT, KESAL, and Initial IRI were found to be the most significant predictors by feature importance analysis. The model's greatest responsiveness to Layer #3 Thickness, AADT, AADTT, Initial IRI, and Age was validated by sensitivity analysis, highlighting the importance of structural and traffic-related factors. Climate variables had a smaller influence, indicating a secondary contribution in this dataset. This study demonstrates the efficacy of hybrid machine learning algorithms for predicting pavement degradation and provides useful insights for prioritizing data factors in design and maintenance planning. The suggested GA-GBM framework is a viable and scalable tool for promoting proactive pavement management and enhancing infrastructure sustainability.