摘要
针对传统养护决策依赖人工巡查、主观性强、数据融合不足等问题,本文以沈海高速A段为对象,提出一套基于多源数据融合的路面养护决策方法。首先,采用滑动窗口均值聚合算法实现20 m与100 m等多尺度病害数据的空间统一化,并通过Box-Cox变换和归一化消除指标量纲差异。其次,构建结合熵权法与AHP的组合权重模型,兼顾数据离散性与专家经验。然后,基于组合权重与TOPSIS方法得到路段优劣排序,并引入GMM聚类进行养护等级划分。为克服GMM对离散点敏感的不足,提出TOPSIS-GMM联合分级策略,设置人工阈值动态修正等级边界。实例验证表明,该方法可有效识别重点养护路段,与实际工程需求高度契合,减少不必要养护约41%,显著提升决策科学性与资源利用效率。
To address the limitations of traditional maintenance decision-making—characterized by reliance on manual inspection,strong subjectivity,and insufficient data integration—this study takes Section A of the Shenhai Expressway as a case and proposes a pavement maintenance decision-making framework based on multi-source data fusion.First,a sliding-window mean aggregation algorithm is applied to unify multi-scale pavement distress data(20 m and 100 m)into a 100 m evaluation grid,and Box-Cox transformation combined with normalization is used to eliminate dimensional inconsistencies.Second,a combined weighting model integrating the entropy weight method and analytic hierarchy process(AHP)is established to balance objective data characteristics with expert knowledge.Subsequently,the Technique for Order Preference by Similarity to Ideal Solution(TOPSIS)is employed to rank maintenance priorities,and Gaussian Mixture Model(GMM)clustering is used to classify maintenance levels.To overcome GMM’s sensitivity to outliers,a TOPSIS-GMM joint classification strategy with adaptive threshold adjustment is proposed.Case verification shows that the proposed framework effectively identifies priority maintenance segments,aligns well with engineering requirements,and reduces unnecessary maintenance by approximately 41%,significantly improving decision-making scientificity and resource allocation efficiency.
作者
黄开青
HUANG Kaiqing(Fujian Fuquan Expressway Co.,Ltd.,Fuzhou Fujian 350000,China)
出处
《交通节能与环保》
2025年第6期34-38,56,共6页
Transport Energy Conservation & Environmental Protection
