Reliability analysis and updating of complex infrastructure systems by Bayesian network

生命线网络等民用基础设施系统是现代社会和经济的重要支柱。为了了解和提高它们的抗灾能力，必须使用来自科学和工程多个领域的模型和数据来量化系统级的风险和可靠性。理想情况下，当在系统的生命周期内有新的信息可用时，就会更新这种风险估计。例如，应在灾害事件发生后立即用现有数据更新风险估计，以便能够迅速而审慎地作出关于灾害应对和灾后网络运营的决定。正如最近证明的那样，贝叶斯网络(BN)方法由于其执行可靠性分析和更新的能力，具有近实时评估和传达复杂系统的状态和可靠性的潜力。然而，为了实现该方法在实际基础设施系统中的应用，必须克服建模和计算的局限性，这是本研究项目的目标。为了实现这一目标，以下研究目标被提出：(1)开发生命线系统可靠性的有效表示和计算；(2)将复杂依赖纳入分析并提高BN在处理这些依赖关系时的计算效率；(3)评估大规模现实系统；以及(4)测试和演示生命线系统的应用。相应的研究任务如下。首先，开发适合于系统可靠性分析的精确BN算法，并将其与基于聚类的多尺度系统可靠性分析相结合。之后，我们将研究基于采样的BN算法，以克服精确算法的一些基本局限性，同时仍有利于快速计算。这些算法将有助于合并组件故障之间的复杂相关性。对于大型现实系统，将开发能够处理非包含性模型的代理模型。最后，将所开发的BN模型和方法应用于自然灾害下复杂基础设施系统的可靠性分析和更新。拟议的BN框架将能够将(实时)数据纳入系统可靠性评估，以随时提供对系统状态的最新预测。开发的理论、模型和方法将足够普遍，适用于大多数主要的生命线系统，并应该带来能够启动广泛相关研究工作的智力优势。这项建议是TUM的Daniel Straub教授和首尔国立大学(SNU)的Junho Song教授联合提出的建议的一部分。TOM将专注于BN的开发，SNU将专注于系统的表现。这两个小组相互补充的专业知识以及它们之间的密切合作将确保该项目在这一高度跨学科的领域取得成功。

Civil infrastructure systems such as lifeline networks are critical backbones of modern societies and economies. To understand and enhance their hazard resilience, it is essential to quantify the system-level risk and reliability by use of models and data from multiple fields of science and engineering. Ideally, such risk estimates are updated when new information becomes available during the lifetime of the system. As an example, the risk estimates should be updated with available data immediately after a disaster event such that decisions on hazard response and post-disaster network operations can be made promptly but prudently. As recently demonstrated, the Bayesian Network (BN) methodology has the potential for assessing and communicating the state and reliability of complex systems in near-real time, due to its capability to perform reliability analysis and updating. However, to implement the approach for real-life infrastructure systems, modeling and computational limitations must be overcome, which is the goal of this proposed research project.To achieve this goal, the following research objectives are addressed: (1) develop efficient representation and computation of lifeline system reliability; (2) include complex dependence into the analysis and enhance the computational efficiency of the BN in dealing with those; (3) assess large-scale realistic systems; and (4) test and demonstrate the application to lifeline systems. The corresponding research tasks are as follows. First, exact BN algorithms tailored for system reliability analysis will be developed and coupled with clustering-based multi-scale system reliability analysis. Thereafter, novel sampling-based BN algorithms will be investigated, to overcome some fundamental limitations of exact algorithms, but still facilitate fast computation. These algorithms will facilitate incorporating complex dependence between component failures. For large-scale realistic systems, surrogate models will be developed that can handle non-inclusive models. Finally, the developed BN models and methods will be applied to reliability analysis and updating of complex infrastructure systems under natural hazards. The proposed BN framework will be able to include (real-time) data into system reliability assessments to provide updated predictions of the system state at all times. The theories, models and methods developed will be general enough to be applicable to most major lifeline systems, and should bring about intellectual merits that can initiate a broad range of related research efforts.This proposal is part of a joint proposal between Prof. Daniel Straub at TUM and Prof. Junho Song at Seoul National University (SNU). TUM will focus on the development of the BN, SNU will focus on the system representation. The complimentary expertise of the two groups together with close collaboration between them will ensure the success of the project in this highly interdisciplinary area.