中文核心期刊
CSCD来源期刊
中国科技核心期刊
RCCSE中国核心学术期刊
重庆交通大学学报（自然科学版）

重庆交通大学学报（自然科学版） ›› 2025, Vol. 44 ›› Issue (6): 35-44.DOI: 10.3969/j.issn.1674-0696.2025.06.04

• 道路与铁道工程 • 上一篇    

基于BOA-XGBoost的沥青路面抗滑性能预测方法

许新权1，户媛姣2，翁宇涵3，何伟杰1   

  1. (1. 广东华路交通科技有限公司， 广东 广州 510400； 2. 西安石油大学 计算机学院， 陕西 西安 710065； 3. 长安大学 信息工程学院，陕西 西安 710064)
  • 收稿日期:2024-07-01 修回日期:2024-10-17 发布日期:2025-06-30
  • 作者简介:许新权（1980—），男，江苏徐州人，教授级高工，博士，主要从事路面结构与材料方面的研究。E-mail：xuxinquan998@126.com 通信作者：翁宇涵（1997—），男，陕西安康人，博士，主要从事人工智能算法和沥青路面性能方面的研究。E-mail：wengyuhan97@163.com
  • 基金资助:
    国家自然科学基金项目(52178407)

Road surface texture is a key factor affecting the skid resistance performance. To further study its influence mechanism and to solve the problem of limited accuracy of traditional prediction methods under multi-feature data conditions, a pavement skid resistance performance assessment model based on the fusion of Bayesian optimization algorithm (BOA) and extreme gradient boosting (XGBoost) was proposed. Firstly, specimens phalt mixtures of different gradation types were prepared. Friction data and 3D texture data of the specimen surface were respectively obtained by using a pendulum friction meter and a 3D laser scanning device. Secondly, the height, the wavelength and shape parameters were extracted to describe the texture structure, and the texture feature importance analysis was carried out to clarify the factors that significantly affected the skid resistance. Then, the optimal key parameters were improved by introducing of the model BOA-XGBoost and the prediction model of skid resistance performance was established. The research results show that compared with the comparative models, the proposed model has higher accuracy, with a correlation coefficient R2 of 0.890 6, which is 25.2%, 13.0%, and 15.1% higher than those of the comparative models, respectively. The proposed model can effectively correlate texture features with pavement skid resistance performance.

XU Xinquan1, HU Yuanjiao2, WENG Yuhan3, HE Weijie1   

  1. (1. Guangdong Hualu Transportation Technology Co., Ltd., Guangzhou 510400, Guangdong, China; 2. School of Computer Science, Xian Shiyou University, Xian 710065, Shaanxi, China; 3. School of Information Engineering, Changan University, Xian 710064, Shaanxi, China)
  • Received:2024-07-01 Revised:2024-10-17 Published:2025-06-30

PDF

摘要： 道路表面纹理是影响抗滑性能的关键因素。为深入研究其影响机理，解决多特征数据条件下传统预测方法精度受限的问题，提出了一种基于贝叶斯优化(BOA)和极端梯度提升(XGBoost)融合的路面抗滑性能评估模型。制备了不同级配类型的沥青混合料试件，基于摆式摩擦仪和三维激光扫描设备分别获取试件表面的摩擦数据和三维纹理数据；提取高度、波长、形状参数用以描述纹理结构，并进行纹理特征重要性分析，明确显著影响抗滑性能因子；引入贝叶斯优化算法的搜索极端梯度来提升模型的最优关键参数，并构建了抗滑性能预估模型。研究结果表明：所提出的模型与对比模型相比，其精度更高，相关系数R2=0.890 6，分别比对比模型提升了25.2%、13.0%、15.1%，能有效地关联纹理特征与路面抗滑性能。

关键词: 道路工程；路面抗滑性能；三维纹理；特征重要性分析；贝叶斯优化算法；极端梯度提升

Abstract: Road surface texture is a key factor affecting the skid resistance performance. To further study its influence mechanism and to solve the problem of limited accuracy of traditional prediction methods under multi-feature data conditions, a pavement skid resistance performance assessment model based on the fusion of Bayesian optimization algorithm (BOA) and extreme gradient boosting (XGBoost) was proposed. Firstly, specimens phalt mixtures of different gradation types were prepared. Friction data and 3D texture data of the specimen surface were respectively obtained by using a pendulum friction meter and a 3D laser scanning device. Secondly, the height, the wavelength and shape parameters were extracted to describe the texture structure, and the texture feature importance analysis was carried out to clarify the factors that significantly affected the skid resistance. Then, the optimal key parameters were improved by introducing of the model BOA-XGBoost and the prediction model of skid resistance performance was established. The research results show that compared with the comparative models, the proposed model has higher accuracy, with a correlation coefficient R2 of 0.890 6, which is 25.2%, 13.0%, and 15.1% higher than those of the comparative models, respectively. The proposed model can effectively correlate texture features with pavement skid resistance performance.

Key words: road engineering; pavement skid resistance performance; three-dimensional texture; feature importance analysis; Bayesian optimization algorithm; extreme gradient boosting

中图分类号: