Civil infrastructure networks, including transportation systems, power grids, and water supply systems, form the essential backbone of modern society. These networks support critical functions of the society by ensuring the continuous and reliable del...
Civil infrastructure networks, including transportation systems, power grids, and water supply systems, form the essential backbone of modern society. These networks support critical functions of the society by ensuring the continuous and reliable delivery of fundamental services. However, as infrastructure networks have become increasingly interconnected and complex, their vulnerability to natural hazards and extreme events has grown to pose significant risks to societal stability and safety. Disruptions in one part of a network can propagate across dependent components, leading to cascading failures that may result in disproportionate impacts on system-level performance. These cascading effects, combined with the challenges of prolonged recovery, pose serious threats to societal resilience. While the significance of this issue has been widely recognized, conventional risk assessment frameworks, which typically focus on individual component reliability, remain limited in capturing the systemic and dynamic nature of infrastructure resilience.
To address these challenges, this dissertation proposes a comprehensive system reliability-based disaster resilience analysis framework tailored for infrastructure networks. The framework integrates multiple dimensions of resilience, i.e., reliability, redundancy, and recoverability, into a unified domain of analyses. It systematically considers the component-level ability to withstand hazards without failure, the network’s capacity to absorb initial disruptions without systemic collapse, and the ability of the network and surrounding society to recover its performance through proper restoration processes. By introducing reliability and redundancy indices, the framework enables probabilistic scenario-based analysis and intuitive visualization of disruption scenarios, supporting comprehensive assessment of network-level resilience.
A distinctive feature of the proposed framework is the incorporation of causality-based analysis, which addresses the limitations of conventional importance measures based on correlation. Causal inference, employing causal diagrams and do-calculus, facilitates the quantification of the true causal effects of component failures on system performance. The proposed causality-based importance measures (CIMs) identify critical components whose disruption or restoration significantly influence overall resilience, thus providing rational guidance for pre-disaster mitigation and post-disaster recovery prioritization.
The framework further extends to the analysis of flow-based infrastructure networks, where the flow redistribution and sequential failures of the components produce highly nonlinear system responses. Cascading failure analysis captures how initial disruptions propagate through the network and interact with its topology and operational characteristics. This enables a realistic evaluation of performance degradation following disasters, including paradoxical phenomena such as Braess' paradox, where increased connectivity may unexpectedly reduce overall network functionality.
Beyond disruption analysis, the proposed framework encompasses a novel approach for modeling the post-disaster recovery process. Recovery simulations account for uncertainties in the recovery time, resource limitations, and recovery strategies, enabling probabilistic evaluation of performance restoration trajectories. The recoverability assessment provides a basis for quantifying the likelihood of the network’s recovery within specified timeframes, thereby extending resilience curve formulation into the recovery phase.
The proposed approaches are demonstrated through numerical case studies involving hypothetical networks, flow-based operational systems, and realistic infrastructure systems such as transportation networks and power grids, with a particular emphasis on seismic hazards. These applications, including earthquake-induced disruptions, show the framework’s capability to characterize disaster scenarios, identify critical components, and support resilience-informed decision making across disruption, cascading, and recovery phases.
By integrating scenario-based resilience analysis, causality-based component importance assessment, cascading failure modeling, and recovery process simulation, this dissertation presents an advanced and unified framework for disaster resilience assessment of infrastructure networks. The proposed framework enhances interpretability and flexibility in resilience analysis, offering valuable tools to support decision-making in both for pre-disaster risk reduction strategies and adaptive post-disaster recovery planning under extreme event conditions.