Collaborative Research: Bayesian Methods for the Data-Driven Recovery of Networks: Measuring Impact and Building Resilience in Infrastructures and Communities

美国政府越来越强调关键基础设施的弹性规划，其中抵御和从破坏中恢复的结合，加剧了我们老化和脆弱的基础设施系统，构成了弹性。根据国土安全部的说法，关键基础设施的弹性运行“对国家安全、公共卫生和安全、经济活力和生活方式至关重要”。最近特别关注的是对破坏性事件后社区恢复力的强调，承认基础设施不是为了自身而存在，而是服务于社会（例如，公民，工业），并且在某些情况下，弹性社区有助于保护建筑环境。理想情况下，弹性社区将能够利用物理基础设施有效地沟通风险，协调恢复策略，以应对中断并从中断中恢复，并最终适应变化并从过去的中断中学习。这项工作的目标是开发一个新的数据驱动优化框架，以提高(i)基础设施网络性能建模的能力，以及（ii）在中断后为这些网络的恢复进行规划的能力，重点是社区恢复能力和经济生产力。研究方法由三个部分组成。第一部分发展了一种新的统计技术，即层次贝叶斯核方法，该方法结合了贝叶斯在动态获取数据时提高预测精度的特性、核函数为模型增加特异性并使非线性数据更易于管理的特性，以及在破坏性事件场景中常见的稀疏和多样化数据情况下从不同来源借鉴信息的层次特性。第二个组件开发了一个基础设施网络恢复优化公式，该公式使用数据驱动（和动态更新）的基础设施恢复的分层贝叶斯核参数，以及考虑模型参数的大小和动态性质的解决方案技术，最大限度地减少基础设施网络性能的较大影响。将前两个集成组件应用于电力网络（对社区的安全和弹性进行测量）和内陆水道（对多个行业的经济生产力进行测量），构成第三个组件，为基础设施网络弹性和恢复的影响提供了两个应用视角。

The US government has increasingly emphasized resilience planning for critical infrastructure, where the combination of withstanding and recovering from disruptions that exacerbate our aging and vulnerable infrastructure systems, constitutes resilience. According to the Department of Homeland Security, the resilient operation of critical infrastructures is "essential to the Nation's security, public health and safety, economic vitality, and way of life." Of particular interest recently is an emphasis on the resilience of communities after a disruptive event, acknowledging that infrastructures do not exist for their own sake but serve society (e.g., citizens, industries), and in some cases, resilient communities assist in protecting the built environment. A resilient community would ideally be able to use the physical infrastructure to effectively communicate risk and coordinate recovery strategies to respond to and recover from disruptions, and ultimately adapt to change and learn from past disruptions. The objective of this work is to develop a new data-driven optimization framework to improve (i) the ability to model the performance of infrastructure networks, and (ii) the ability to plan for the recovery of these networks after a disruption, with an emphasis on community resilience and economic productivity. The research approach is composed of three components. The first component develops a new statistical technique, the hierarchical Bayesian kernel method, which integrates the Bayesian property of improving predictive accuracy as data are dynamically obtained, the kernel function that adds specificity to the model and can make nonlinear data more manageable, and the hierarchical property of borrowing information from different sources in sparse and diverse data situations which are common in disruptive events scenarios. The second component develops an infrastructure network recovery optimization formulation that minimizes the larger impact of infrastructure network performance with data-driven (and dynamically updated) hierarchical Bayesian kernel parameters of infrastructure recovery, along with solution techniques that account for the size and dynamic nature of model parameters. The application of the first two integrated components to electric power networks (where impact is measured on the safety and resilience of the community) and inland waterways (where impact is measured on economic productivity across multiple industries), constitutes the third component, offering two application perspectives on the impact of infrastructure network resilience and recovery.