基于贝叶斯分层模型的甘肃省国道路面损坏状况指数分析  

Analysis of pavement condition index in Gansu Province based on a Bayesian hierarchical model

作　　者：陈涛 曾铭 余遥 胡潇潇 CHEN Tao;ZENG Ming;YU Yao;HU Xiao-xiao(Road Network Monitoring and Emergency Response Center,Highway Development Center of Gansu Province,Lanzhou 730000,China;School of Mathematics and Statistics,Center for Data Science,Lanzhou University,Lanzhou 730000,China)

机构地区：[1]甘肃省公路事业发展中心交通运行(路网)监测与应急处置中心,兰州730000 [2]兰州大学数学与统计学院,大数据科学研究中心,兰州730000

出　　处：《兰州大学学报（自然科学版）》2024年第6期790-797,共8页Journal of Lanzhou University(Natural Sciences)

基　　金：甘肃省交通运输厅科研项目(2023-03)。

摘　　要：对甘肃省普通国道路面损坏状况指数(P_(CI))的自然衰减规律建立统计模型.考虑到甘肃省不同地区之间存在完全不同的气候环境的情况,根据降水量、日照时长等环境数据对不同路段进行聚类,基于年交通当量(整个年度内的交通当量)与P_(CI)数据,利用吉布斯采样进行贝叶斯推断建立贝叶斯分层模型.结果表明,在不同的评价标准下,贝叶斯分层模型在测试集的整体表现优于非分层模型.年交通量对甘肃省不同气候区域国道路段的P_(CI)的影响没有区别;具有较大P_(CI)初始值的路段其P_(CI)值经历一年期衰减后倾向于增大,同时P_(CI)初始值对P_(CI)的衰减起到缓解作用.不同区域P_(CI)初始值的影响程度不相同,河西地区国道路段的P_(CI)初始值对缓解未来一年P_(CI)的衰减作用最大,其次为陇东南地区,最弱为兰州地区及陇东高纬度地区.A statistical model was established for the natural decay pattern of one of the key performance indicators of Gansu Province's ordinary national road pavements,i.e.the pavement condition index(P_(CI)).Considering the entirely different climate conditions that exist across different regions in Gansu,this study clustered various sections of roads based on environmental data such as rainfall and sunshine duration.Using annual traffic equivalent data(traffic volume for the entire year)and P_(CI)data,a Bayesian hierarchical model was constructed via Gibbs sampling for Bayesian inference.Experimental results demonstrated that,under varying evaluation criteria,the Bayesian hierarchical model outperformed non-hierarchical models in overall performance on the test set.Through model analysis,the following findings were discovered:annual traffic volume had no distinguishable impact on P_(CI)for different climatic zones along national highways in Gansu Province;road sections with larger initial P_(CI)values tended to have a higher P_(CI)value after a year of decay,and the initial P_(CI)value also served to mitigate the rate of P_(CI)decay.The influence of the initial P_(CI)value varied from region to region,with the initial P_(CI)value having the most significant mitigating effect on future P_(CI)decay for national highway sections in the Hexi Corridor,followed by the south-eastern Gansu area,and weakest in the Lanzhou region and high-latitude areas of eastern Gansu.

关 键 词：路面状况指数 年交通当量 贝叶斯分层模型 吉布斯采样 

分 类 号：O213.1[理学—概率论与数理统计] U418.6[理学—数学]