摘要
针对传统方法难以准确刻画沥青路面损坏状况空间相关性的问题,基于Copula函数构建了PCI空间依赖模型,以提升预测精度与稳健性。以山东省某高速公路2019年—2021年连续3年的PCI检测数据为基础,先通过边缘分布拟合与极大似然估计(MLE)确定PCI的边缘分布类型及参数;后采用Normal、t、Gumbel、Clayton和Frank五类Copula函数建立相邻路段PCI联合分布模型,并通过K-S检验与欧氏距离评估拟合优度;再利用马尔科夫链蒙特卡洛(MCMC)方法对Copula参数进行贝叶斯推断。结果表明:1)Clayton Copula在低PCI区的空间依赖建模能力最优,可有效反映潜在病害聚集特征;2)基于最优Copula模型与MCMC抽样方法得到的PCI空间预测结果,预测区间覆盖率达到1,区间宽度仅为0.021,在保证高覆盖率的同时保持了较窄区间,体现了模型在精度与不确定性刻画上的优势;3)与传统多元线性回归、BP神经网络、支持向量回归及随机森林回归模型相比,Copula-MCMC方法在预测精度、区间覆盖率及区间宽度等指标上均表现更优。所建Copula空间依赖模型能精确刻画PCI的空间相关性,结合MCMC贝叶斯推断具有较高预测精度,可为沥青路面养护决策提供可靠方法支撑。
To address the difficulty in accurately characterizing the spatial correlation of asphalt pavement damage condition using traditional methods,a spatial dependence model of PCI was developed based on the Copula function to enhance prediction accuracy and robustness.On the basis of three consecutive years(2019-2021)of PCI survey data from a highway in Shandong Province,the marginal distribution type and parameters of PCI were first determined through marginal distribution fitting and maximum likelihood estimation(MLE).Five types of Copula functions—Normal,t,Gumbel,Clayton,and Frank—were then employed to establish joint distribution models of PCI for adjacent road sections.The goodness-of-fit of these models was evaluated using the Kolmogorov-Smirnov test and Euclidean distance.The Markov Chain Monte Carlo(MCMC)method was applied for Bayesian inference of the Copula parameters.The results indicate that:1)Clayton Copula provides the best modeling effect for spatial dependence in low-PCI regions,effectively reflecting potential clustering of pavement distress;2)Based on the optimal Copula model and MCMC sampling method,PCI spatial prediction results were obtained with a prediction interval coverage probability(PICP)of 1.0 and an interval width of only 0.021,achieving high coverage rate while maintaining narrow intervals,which reflects advantages in both accuracy and uncertainty quantification;3)Compared with traditional multivariate linear regression(MLR),BP neural networks(BPNN),support vector regression(SVR),and random forest(RF)regression models,the Copula-MCMC approach outperforms in prediction accuracy,interval coverage rate,and interval width.The established Copula spatial dependence model can accurately characterize the spatial correlation of PCI.Combined with MCMC Bayesian inference,it has high predictive precision,and can provide a reliable methodological foundation for asphalt pavement maintenance decision-making.
作者
王川
姚朝林
张宇
闫英俊
田育禾
管延华
孙仁娟
WANG Chuan;YAO Zhaolin;ZHANG Yu;YAN Yingjun;TIAN Yuhe;GUAN Yanhua;SUN Renjuan(Shandong Expressway Group Co.,Ltd.,Ji′nan 250100;School of Qilu Transportation,Shandong University,Ji′nan 250100)
出处
《公路交通技术》
2026年第1期78-84,102,共8页
Technology of Highway and Transport
基金
山东省自然科学基金面上项目(ZR2021ME215)。
关键词
道路工程
COPULA函数
PCI
马尔科夫链蒙特卡洛
贝叶斯推断
road engineering
Copula function
PCI(Pavement Condition Index)
Markov Chain Monte Carlo(MCMC)
Bayesian inference
