刘文黎 吴贤国 张文静 王彦玉

摘 要：对城市盾构地铁结构的健康服役和安全可靠性进行评价时，多局限于静态推理，通过构建PCBN（Pair-Copula Bayes Network）网络结构得到评判指标的多元联合概率分布，实现参数相依性的准确识别和高精度重构。基于3种失效模式对盾构地铁结构节点相关性分析及安全状态评价，充分挖掘监测数据的隐藏信息，表征盾构地铁结构的风险状态。以武汉地铁三号线某空推段为工程背景进行PCBN模型建模。结果表明：PCBN模型得出的地铁安全状态与地铁工程中的实际风险信息非常吻合，体现了构建PCBN网络模型的准确性和实用性。

关键词：盾构地铁结构;PCBN模型;相关性分析;可靠度

中图分类号：TU 375.1

文献标志码：A  文章编号：2096-6717（2019）02-0045-08

Abstract：At present， the most common method to evaluate the health service and safety reliability of urban shield metro structures is limited to static analysis. In this paper， by applying the concept of Pair-Copula to Bayesian networks， the multi-joint probability distribution of evaluation index was obtained by constructing PCBN （Pair-Copula Bayes Network） structure， which is capable of  identifying parameters dependencies accurately and high-precision reconstruction. Based on the three kinds of failure modes， the correlation analysis and safety  evaluation of shield metro structure nodes were conducted. The implicit information in monitoring data was exploited to characterize the risk status of the shield metro structure. As an example， an empty pushing section of Wuhan Metro Line 3 is used to build a PCBN model. Results show that the metro safety status obtained by the PCBN model is in good agreement with the actual risk information in the subway project， demon strating the accuracy and practicability of the PCBN network model proposed in this paper.

Keywords：shield subway structure; PCBN model; correlation analysis; reliability

隨着地铁运营网络规模的逐年扩大，盾构地铁结构运营安全问题逐渐凸显，为了对盾构地铁结构的安全状态进行准确评估，许多研究者进行了大量的工作，提出了许多方法，包括模糊层次分析法[1]、事故树分析法[2]、模糊综合评价法[3]、可靠度评价法[4]等。但是，盾构地铁赋存环境因素对地铁承载性能的影响有着复杂的非线性、随机性、不确定性及多时空演化等特征。传统的评估方法多局限于静态的推理过程，难以构建监测指标与结构安全之间的隐性非线性映射关系，进行实时风险评定、动态风险预测以及风险诊断推理。

贝叶斯网络（Bayesian Networks，BN）能够直观地构建两个或多个随机变量因素之间的条件概率分布，且能够将先验信息与样本数据有机结合，实现网络节点因素之间的实时推理[5]。……