Scholarlink

<p>In order to predict the friction number in Continuously Reinforced Concrete Pavement (CRCP), this study presents an advanced hybrid model that combines Support Vector Regression (SVR) with Particle Swarm Optimization (PSO). PSO was used to adjust the SVR’s hyperparameters for increased predicted accuracy using data from the Long-Term Pavement Performance (LTPP) study. The PSO-SVR model produced better results than traditional models like Linear Regression and Decision Trees, with a mean Root Mean Square Error (RMSE) of 2.7404 and a mean coefficient of determination (R²) of 0.90438. The friction number was shown to be significantly impacted by key influencing parameters, including pavement age, lane number Kilo Equivalent Single Axle Load (KESAL), Subbase Type (L2 type), and Annual Average Daily Truck Traffic (AADTT). This underscores the crucial roles that traffic loading and structural design play in surface friction behavior. Furthermore, 3D surface plots were created to show how age and other input variables interact, offering a more thorough understanding of their combined impact. By facilitating accurate surface condition predictions, the suggested approach shows great promise for assisting with proactive pavement maintenance, which will improve road safety, driver satisfaction, and infrastructure sustainability. By adding more factors and investigating additional advanced machine learning techniques for pavement condition prediction, future work will build on this.</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

Optimizing Support Vector Regression with Particle Swarm Intelligence for Accurate Friction Estimation in CRCP

  • 期刊论文

  • 发表日期：

    2025年10月8日

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

摘要

In order to predict the friction number in Continuously Reinforced Concrete Pavement (CRCP), this study presents an advanced hybrid model that combines Support Vector Regression (SVR) with Particle Swarm Optimization (PSO). PSO was used to adjust the SVR’s hyperparameters for increased predicted accuracy using data from the Long-Term Pavement Performance (LTPP) study. The PSO-SVR model produced better results than traditional models like Linear Regression and Decision Trees, with a mean Root Mean Square Error (RMSE) of 2.7404 and a mean coefficient of determination (R²) of 0.90438. The friction number was shown to be significantly impacted by key influencing parameters, including pavement age, lane number Kilo Equivalent Single Axle Load (KESAL), Subbase Type (L2 type), and Annual Average Daily Truck Traffic (AADTT). This underscores the crucial roles that traffic loading and structural design play in surface friction behavior. Furthermore, 3D surface plots were created to show how age and other input variables interact, offering a more thorough understanding of their combined impact. By facilitating accurate surface condition predictions, the suggested approach shows great promise for assisting with proactive pavement maintenance, which will improve road safety, driver satisfaction, and infrastructure sustainability. By adding more factors and investigating additional advanced machine learning techniques for pavement condition prediction, future work will build on this.