Recent studies increasingly adopt simulation-based machine learning (ML) models to analyse critical infrastructure system resilience. For realistic applications, these ML models consider the component-level characteristics that influence the network response during emergencies. However, such an approach could result in a large number of features and cause ML models to suffer from the ’curse of dimensionality’. A clustering-based method is presented that simultaneously minimises the problem of high-dimensionality and improves the prediction accuracy of ML models developed for resilience analysis in large-scale interdependent infrastructure networks. The methodology has three parts: (a) generation of simulation dataset, (b) network component clustering, and (c) dimensionality reduction and development of prediction models. First, an interdependent infrastructure simulation model simulates the network-wide consequences of various disruptive events. The component-level features are extracted from the simulated data. Next, clustering algorithms are used to derive the cluster-level features by grouping component-level features based on their topological and functional characteristics. Finally, ML algorithms are used to develop models that predict the network-wide impacts of disruptive events using the cluster-level features. The applicability of the method is demonstrated using an interdependent power-water-transport testbed. The proposed method can be used to develop decision-support tools for post-disaster recovery of infrastructure networks.@en

In this study, we propose a methodological framework to identify and evaluate cost-effective pathways for enhancing resilience in large-scale interdependent infrastructure systems, considering decision-makers’ risk preferences. We focus on understanding how decision-makers with varying risk preferences perceive the benefits from infrastructure resilience investments and compare them with upfront costs in the context of high-impact low-probability (HILP) events. First, we compute the costs of interventions as the sum of their capital costs and maintenance costs. The benefits of the interventions include the reduction in physical damage costs and business disruption losses resulting from the improved resilience of the network. In the final stage, we develop statistical models to predict the perceived net benefits of different network resilience configurations in power, water, and transport networks. These models are employed in an optimization framework to identify optimal resilience investment pathways. By incorporating Cumulative Prospect Theory (CPT) in the optimization framework, we show that decision-makers who assign higher weights to low probability events tend to allocate more resources towards post-disaster recovery strategies leading to increased resilience against HILP events, like earthquakes. We illustrate the methodology using a case study of the interdependent infrastructure network in Shelby County, Tennessee.

@en

Enhancing the resilience of critical infrastructure systems requires substantial investment and entails trade-offs between environmental and economic benefits. To this aim, we propose a methodological framework that combines resilience and economic analyses and assesses the economic viability of alternative resilience designs for a Water Distribution System (WDS) and its interdependent power and transportation systems. Flow-based network models simulate the interdependent infrastructure systems and Global Resilience Analysis (GRA) quantifies three resilience metrics under various disruption scenarios. The economic analysis monetizes the three metrics and compares two resilience strategies involving the installation of remotely controlled shutoff valves. Using the Micropolis synthetic interdependent water-transportation network as an example, we demonstrate how our framework can guide infrastructure stakeholders and utility operators in measuring the value of resilience investments. Overall, our approach highlights the importance of economic analysis in designing resilient infrastructure systems.@en

Integrated simulation models are emerging as an alternative for analyzing large-scale interdependent infrastructure networks due to their modeling advantages over traditional interdependency models. This paper presents an open-source integrated simulation package for the component-level analysis of interdependent power-, water-, transport networks. The simulation platform, named ’InfraRisk’ and developed in Python, can simulate network-wide effects of disaster-induced infrastructure failures and subsequent post-disaster restoration. InfraRisk consists of an infrastructure module, a hazard module, a recovery module, a simulation module, and a resilience quantification module. The infrastructure module integrates existing infrastructure network packages (wntr for water distribution systems, pandapower for power systems, and a static traffic assignment model for road transport systems) through an interface that facilitates the network-level simulation of infrastructure failures. The hazard module generates infrastructure component failures based on various disaster characteristics. The recovery module determines repair sequences and assigns repair crews based on predefined heuristics-based recovery strategies or model predictive control (MPC) based optimization. Based on the schedule, the simulation module simulates the consequences of the disaster impacts and the recovery actions on the performance of the interdependent network. The resilience quantification module offers system-level and consumer-level metrics to quantify both the risks and resilience of the integrated infrastructure networks against disaster events. InfraRisk provides a virtual platform for decision-makers to experiment and develop region-specific pre-disaster and post-disaster policies to enhance the overall resilience of interdependent urban infrastructure networks.@en

Recent studies increasingly adopt simulation-based machine learning (ML) models to analyze critical infrastructure system resilience. For realistic applications, these ML models consider the component-level characteristics that influence the network response during emergencies. However, such an approach could result in a large number of features and cause ML models to suffer from the `curse of dimensionality'. We present a clustering-based method that simultaneously minimizes the problem of high-dimensionality and improves the prediction accuracy of ML models developed for resilience analysis in large-scale interdependent infrastructure networks. The methodology has three parts: (a) generation of simulation dataset, (b) network component clustering, and (c) dimensionality reduction and development of prediction models. First, an interdependent infrastructure simulation model simulates the network-wide consequences of various disruptive events. The component-level features are extracted from the simulated data. Next, clustering algorithms are used to derive the cluster-level features by grouping component-level features based on their topological and functional characteristics. Finally, ML algorithms are used to develop models that predict the network-wide impacts of disruptive events using the cluster-level features. The applicability of the method is demonstrated using an interdependent power-water-transport testbed. The proposed method can be used to develop decision-support tools for post-disaster recovery of infrastructure networks.@en

Climate change is increasing the frequency and the intensity of weather events, leading to large-scale disruptions to critical infrastructure systems. The high level of interdependence among these systems further aggravates the extent of disruptions. To mitigate these impacts, models and methods are needed to support rapid decision-making for optimal resource allocation in the aftermath of a disruption and to substantiate investment decisions for the structural reconfiguration of these systems. In this paper, we leverage infrastructure simulation models and Machine Learning (ML) algorithms to develop resilience prediction models. First, we employ an interdependent infrastructure simulation model to generate infrastructure disruption and recovery scenarios and compute the resilience value for each scenario. The infrastructure-, disruption-, and recovery-related attributes are recorded for each scenario and ML algorithms are employed on the synthetic dataset to develop accurate resilience prediction models. The results of the prediction models are analyzed and possible design strategies suggested based on the resilience enhancement attributes. The proposed methodology can support infrastructure agencies in the resource-allocation process for pre- and post-disaster interventions.@en

The resilience of water distribution systems (WDSs) has gained increasing attention in recent years. Various performance loss and recovery behaviors have been observed for WDSs subject to disruptions. However, a model for their characterization, which could provide further insight for resilience assessment and enhancement, is still lacking. Here, the authors develop a recovery function to model WDS performance over time following a disruption. This function is useful to compare system responses under different disruption and recovery scenarios and supports the identification of areas for improvement within various aspects of the resilience of a WDS. The proposed model was applied to two benchmark networks. Different scenarios were analyzed in which one node at a time was disrupted and two recovery strategies were implemented. It was found that the developed model supports the implementation of tailored strategies to improve WDS resilience according to the location of the disruption, therefore enhancing the efficient allocation of resources.

@en

Water Distribution Systems (WDSs) are critical infrastructures for providing water to sustain life and human activities. Some recent trends, such as climate change, urbanization, and increasing system interdependence, have led to more frequent threats, with detrimental effects on WDSs. In recent years, resilience has been considered an effective approach to address those threats, for which it is difficult to estimate the likelihood and consequences (Henry and Ramirez-Marquez, 2012). In the literature, two approaches were used to assess WDS resilience. On the one hand, performance-based metrics were used to quantify the impacts of disruptions on the WDS. Specifically, recovery functions were developed to model the time-continuous system response following a disruption, during periods of loss and restoration of performance (Cassottana et al., 2019). On the other hand, indexes based on system attributes were developed to classify WDSs and identify structural vulnerabilities. For example, algebraic connectivity, clustering coefficient, and average path length were used as proxies for robustness, redundancy, and efficiency, respectively (Yazdani et al., 2011). However, those approaches were applied separately, and the relationship between the performance of WDSs and their attributes is still unknown. Hence, the question arises from the above analysis on how to identify the key structural factors determining the resilience of a WDS. The goal of this research is to understand how and to what extent different network topologies determine different performance losses and recovery behaviors given increasing magnitude of disruption, i.e., pipe breakdown. To this end, we consider different network topologies as case studies and vary their structural attributes, e.g., water source head and tank capacity. We then simulate disruption scenarios of increasing magnitude and model the resulting system performance by means of recovery functions for the assessment of resilience. The estimated parameters of these functions are useful for characterizing different system responses, including severe or limited performance losses and fast or slow recoveries. By systematically varying the network topologies and the structural attributes, the function parameters could be in turn associated with key structural factors. We find that, while increasing the WDS supply capacity results in limited performance loss in terms of satisfied demand for water, increasing the reserve capacity improves the robustness of the system by delaying the loss of performance. This analysis will inform the design of resilient water networks based on their topology and unique attributes.@en