Flood-induced scour is by far the leading cause of bridge failures, resulting in loss of lives, traffic disruption and significant economic losses. In Scotland, there are around 2,000 structures, considering both road and railway bridges, susceptible to scour. Scour assessments are currently based on visual inspections, which are expensive, time-consuming, and often the information collected is qualitative and subjective. The two main transport agencies in Scotland, Transport Scotland and Network Rail, spend £2m and £0.4m per annum, respectively, in routine inspections. Nowadays sensor and communication technologies offer the possibility to assess in real time the scour depth at critical bridge locations; yet monitoring an entire infrastructure network is not economically sustainable. A way to overcome this limitation is to install monitoring systems on a limited number of critical locations and use a probabilistic approach to extend this information to the entire population of assets. The state of the bridge stock is represented through a set of random variables and ad-hoc Bayesian networks (BNs) are used to describe their conditional dependencies. The aim of this paper is to develop a probabilistic scour hazard model by building a BN able to estimate the depth of scour in the surrounding of bridge foundations. The BN can estimate, and continuously update, the present and future scour depth using real-time information from monitoring of scour depth and river flow characteristics. In the occurrence of a flood, monitoring observations are used to infer the posterior distribution of the state variables probabilistically, and therefore to give in real-time the best estimate of total scour depth. Bias, systematic and model uncertainties are modelled as nodes of the BN in such a way as the accuracy of predictions can be updated when information from the scour monitoring system is incorporated into the BN. In order to demonstrate the functioning of the BN, bridges managed by TS in South-West Scotland were used to build a small bridge network. They cross the same river (River Nith) and only one of them is instrumented with a scour monitoring system.

中文翻译：

---

使用贝叶斯网络评估公路和​​铁路桥梁的水下冲刷

迄今为止，洪水引起的冲刷是桥梁故障的主要原因，造成人员伤亡、交通中断和重大经济损失。在苏格兰，考虑到公路和铁路桥梁，大约有 2,000 座建筑物容易受到冲刷。Scour 评估目前基于目视检查，这种检查既昂贵又耗时，而且收集的信息通常是定性的和主观的。苏格兰的两个主要运输机构，苏格兰运输局和 Network Rail，每年分别在例行检查中花费 200 万英镑和 40 万英镑。如今，传感器和通信技术提供了在关键桥梁位置实时评估冲刷深度的可能性；然而，监控整个基础设施网络在经济上是不可持续的。克服这一限制的一种方法是在有限数量的关键位置安装监控系统，并使用概率方法将此信息扩展到整个资产群体。桥梁库存的状态通过一组随机变量表示，并使用自组织贝叶斯网络 (BN) 来描述它们的条件依赖关系。本文的目的是通过构建能够估计桥梁基础周围冲刷深度的 BN 来开发概率冲刷危险模型。BN 可以使用来自对冲刷深度和河流流量特征的监测的实时信息来估计并不断更新当前和未来的冲刷深度。在发生洪水时，监测观测被用于概率地推断状态变量的后验分布，从而实时给出总冲刷深度的最佳估计。偏差、系统和模型不确定性被建模为 BN 的节点，这样当来自冲刷监测系统的信息被纳入 BN 时，预测的准确性可以更新。为了展示 BN 的功能，由 TS 在苏格兰西南部管理的桥梁被用来建立一个小型桥梁网络。它们穿过同一条河流（Nith 河），其中只有一条配备了冲刷监测系统。为了展示 BN 的功能，由 TS 在苏格兰西南部管理的桥梁被用来建立一个小型桥梁网络。它们穿过同一条河流（Nith 河），其中只有一条配备了冲刷监测系统。为了展示 BN 的功能，由 TS 在苏格兰西南部管理的桥梁被用来建立一个小型桥梁网络。它们穿过同一条河流（Nith 河），其中只有一条配备了冲刷监测系统。