Development of a civil infrastructure resilience assessment framework and its application to a nuclear power plant

Recent catastrophic events highlight the vulnerability of civil infrastructure to natural hazards and the need for resilience-based design. Traditional design codes primarily focus on minimizing loss of life, neglecting the crucial aspect of post-disaster recovery. This paper addresses this gap by proposing a comprehensive framework, CIRAF, for assessing and enhancing the resilience of infrastructure systems. Existing frameworks often have limitations, such as being geographically specific, limited to certain hazard types, or lacking methodologies for correlating interconnected components and quantifying economic losses. Some rely heavily on computationally expensive Monte Carlo simulations. CIRAF aims to overcome these limitations by integrating various modeling techniques, including Bayesian Networks, to efficiently assess the resilience of complex, interconnected systems, considering both direct and indirect losses. The framework’s flexibility accommodates various data sources, from finite element analyses (FEA) to Internet of Things (IoT) sensor data.

The paper reviews existing frameworks for quantifying the resilience of civil infrastructure systems. Several simulation-based approaches are noted, but their reliance on Monte Carlo methods for handling low-probability events is highlighted as a drawback. Other frameworks are criticized for their geographical or hazard-type limitations, or for lacking comprehensive methodologies for component correlation and loss assessment. Bayesian methods are identified as a promising approach for correlating component fragility, offering greater efficiency compared to matrix-based methods, especially for systems with numerous components. However, these existing methods often focus solely on risk assessment without incorporating resilience and economic loss assessments. The reviewed literature reveals a need for a unified framework capable of handling diverse hazard types, complex interdependencies between components, and various types of losses, while remaining computationally efficient.

CIRAF's methodology consists of several key steps: 1) System and hazard definition: Modeling the infrastructure system and identifying potential hazards. 2) Infrastructure component modeling: Developing fragility functions (relating hazard intensity to probability of failure), recovery functions (modeling post-disaster recovery), and upgrade models (analyzing the impact of retrofitting). Demand values for the models are sourced from FEA, IoT devices, or analytical procedures. A tool facilitates exporting results from FEA software into a standard data exchange format. 3) Loss estimation: Using conditional probability tables (CPTs) within a Bayesian Network (BN) framework to model component and subsystem interactions. Loss functions are defined at component and system levels, considering both direct (physical damage) and indirect losses (e.g., flow disruption). The Bayesian approach efficiently handles the probabilistic relationships between components. 4) Performance indicators: Using metrics such as Upgrade Benefit Index (UBI), Damage Consequence Index (DCI), repair time, and repair cost to evaluate system resilience and guide decision-making. The UBI quantifies the benefit of upgrading a specific component, while the DCI assesses the impact of a component's failure. The paper also introduces multiple recovery functions (linear, exponential, cosine, and multi-step), to better account for different system recovery scenarios. 5) Advantages: CIRAF's flexibility in data sources, efficient use of Bayesian Networks, scalability to handle large systems (through the use of 'transfer nodes'), and intuitive web-based implementation are emphasized as key advantages.

A case study focusing on a hypothetical nuclear power plant illustrates CIRAF's application. The plant's seismic fragility and resilience are assessed under earthquake scenarios with a 10,000-year return period. FEA is used to generate demand values for critical components in the reactor building. The study examines four upgrade strategies: (1) as-built (no upgrades), (2) upgrading the turbine building, (3) upgrading the turbine building and switchyard, and (4) upgrading the turbine building, switchyard, generators, and high-pressure heaters. The results show that increased component upgrades lead to reduced repair costs and shorter repair times; however, there is also a trade-off against increasing upgrade costs. The functionality curves for the upgraded cases show varying levels of resilience. The study highlights the framework's capacity to help decision-makers evaluate different mitigation strategies based on cost-benefit trade-offs and overall resilience improvement.

The findings of this case study demonstrate the effectiveness of CIRAF in assessing the resilience of complex infrastructure systems. The integrated use of Bayesian Networks and performance indicators enables stakeholders to make informed decisions about upgrading and risk mitigation. The framework's flexibility in handling different data sources and its capacity to analyze large-scale interconnected systems addresses limitations of existing frameworks. The case study emphasizes the trade-offs between upgrade costs and resilience improvements, highlighting the importance of comprehensive cost-benefit analysis in decision-making. The results underscore the potential of CIRAF to enhance the resilience of critical infrastructure systems and improve community preparedness for extreme events.

CIRAF provides a comprehensive, flexible, and efficient framework for assessing the resilience of civil infrastructure systems. Its ability to integrate various data sources, handle complex system interdependencies, and provide insightful performance indicators makes it a valuable tool for decision-making. Future work should focus on expanding the framework to include more detailed loss estimation, refining recovery models, optimizing computational efficiency, and exploring the integration of socio-economic factors. The web application developed for CIRAF enhances its accessibility and user-friendliness. This study contributes to the development of resilience-based design approaches, moving beyond traditional code-based design to create more resilient and robust infrastructure systems.

The study acknowledges several limitations. The case study uses a simplified model of a nuclear power plant, and the results may not be directly generalizable to other types of systems. The definition of fragility functions, recovery models, and loss functions relies on expert knowledge and judgment, which may introduce some degree of subjectivity. The framework currently focuses primarily on physical losses and could be expanded to include socio-economic impacts. While the Bayesian Network approach is computationally efficient compared to Monte Carlo methods, further improvements in computational efficiency might be beneficial for extremely large systems.