Scholarlink

<p>Accurate prediction of pavement surface friction is crucial for road safety and optimizing maintenance plans. This study presents a data-driven approach to estimate the Friction Number in Continuously Reinforced Concrete Pavement (CRCP). The proposed method uses machine learning models. These models are trained on extensive data from the Long-Term Pavement Performance (LTPP) database. A total of 33 CRCP sections were examined, including structural, climatic, traffic, and performance data. Six prediction models were developed and tested, including Particle Swarm Optimization-tuned Gradient Boosting Machine (PSO-GBM), conventional GBM, Linear Regression, Random Forest, Support Vector Regression (SVR) and Artificial Neural Network (ANN). Under the adopted five-fold cross-validation framework, the PSO-GBM model achieved the best performance among the evaluated configurations. It obtained a mean RMSE of 3.67 and a mean R<sup>2</sup> of 0.83. These results indicate the effectiveness of metaheuristic optimization in improving model accuracy. Feature importance and sensitivity studies found that the most influential variables were Average Annual Daily Traffic (AADT), pavement age, total thickness, and Initial IRI. Partial dependence plots demonstrated that traffic and aging have a detrimental impact on friction, whilst increased structural thickness has a favorable affect. The proposed PSO-GBM model provides a stable and interpretable framework for friction prediction within the available dataset. It offers useful insights into pavement condition. These insights can support data-driven analysis and evaluation.</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

Data-driven prediction of friction number in CRCP using a PSO-optimized gradient boosting machine

  • 期刊论文

  • 发表日期：

    2026年5月19日

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

摘要

Accurate prediction of pavement surface friction is crucial for road safety and optimizing maintenance plans. This study presents a data-driven approach to estimate the Friction Number in Continuously Reinforced Concrete Pavement (CRCP). The proposed method uses machine learning models. These models are trained on extensive data from the Long-Term Pavement Performance (LTPP) database. A total of 33 CRCP sections were examined, including structural, climatic, traffic, and performance data. Six prediction models were developed and tested, including Particle Swarm Optimization-tuned Gradient Boosting Machine (PSO-GBM), conventional GBM, Linear Regression, Random Forest, Support Vector Regression (SVR) and Artificial Neural Network (ANN). Under the adopted five-fold cross-validation framework, the PSO-GBM model achieved the best performance among the evaluated configurations. It obtained a mean RMSE of 3.67 and a mean R2 of 0.83. These results indicate the effectiveness of metaheuristic optimization in improving model accuracy. Feature importance and sensitivity studies found that the most influential variables were Average Annual Daily Traffic (AADT), pavement age, total thickness, and Initial IRI. Partial dependence plots demonstrated that traffic and aging have a detrimental impact on friction, whilst increased structural thickness has a favorable affect. The proposed PSO-GBM model provides a stable and interpretable framework for friction prediction within the available dataset. It offers useful insights into pavement condition. These insights can support data-driven analysis and evaluation.