Lecturer University of Glasgow
Assessing risk across regional infrastructure networks at scale
Co-Author: Hyeuk Ryu (Geoscience Australia)
Abstract: Regional-scale network risk assessment is a central challenge in infrastructure resilience analysis, requiring integration of component-level damage models with system-level performance evaluation across large, interconnected networks. Existing regional damage assessment tools, such as SimCenter R2D, provide robust component-level damage estimation, yet system-level risk evaluation remains computationally constrained when network performance depends on complex, nonlinear interactions among thousands of components. Exhaustive evaluation of performance functions, such as connectivity or maximum flow, quickly becomes intractable due to the exponential growth of possible component state combinations.
This work highlights the potential for scalable computational software for regional network risk assessment by leveraging a recently proposed tensor-based framework. The framework reformulates system reliability analysis by decomposing the event space into minimal failure and survival rules and decoupling rule discovery from probabilistic inference. Once extracted, these rules can be reused across evaluations under evolving hazard scenarios or damage probabilities, avoiding repeated system-level simulations. A stochastic rule exploration strategy, combined with tensor-based representations and GPU acceleration, enables efficient analysis of networks with thousands of components.
By integrating this framework with regional damage assessment pipelines such as R2D, component-level damage outputs can be translated into scalable, system-level risk metrics that capture not only asset-level failure likelihoods but also broader impacts on network performance. This integration also raises an important question of how multiple, often competing, network objectives—such as connectivity across multiple origin–destination pairs—can be coherently represented and communicated through user-facing platforms.
PhD Student Stanford University
Surrogate Modeling of Highway Bridges for Regional Earthquake Simulation of Transportation Networks
Co-Author: Gregory Deierlein (Stanford University)
Abstract: Hazus is widely used for regional earthquake loss estimation, yet its reliance on generic fragility functions for standardized bridge classes limits its ability to represent bridge-specific behavior. This study develops and evaluates a surrogate-model-based approach that integrates bridge- and site-specific parameters to improve regional bridge damage simulations. Surrogate models, implemented using the QuoFEM backend, provide a flexible and efficient means to incorporate uncertainty and variability in input parameters. Variability in these inputs can be propagated through the surrogate models to generate bridge-specific damage predictions, enabling differentiation among structures that Hazus would otherwise treat identically.
The surrogate model approach is applied in a case study of pre-1971, two-span, single-column concrete bridges in the San Francisco Bay Area and compared against Hazus predictions. R2D is used to obtain ground motion intensity measures using the Earthquake Event Simulation Tool and is also used to execute baseline Hazus analyses. Aggregate results show generally consistent damage estimates between the methods, in part because, the analysis relied on generalized structural parameter distributions. However, when evaluated at the network realization level, differences between the two approaches become more pronounced, with Hazus tending to predict larger network disruptions.
The analysis also incorporates correlations among bridge characteristics and damage responses, demonstrating that such correlations can influence predicted network disruptions. The extent of this influence varies with factors including intra-event ground motion variability, system classification, damage thresholds, and correlation weighting schemes. These findings highlight opportunities to model inter-bridge dependencies and refine regional earthquake damage assessments for transportation networks.
Postdoctoral Scholar University of Nebraska-Lincoln
Flood Fragility Portfolios for Network Resilience of Transportation Systems
Co-Authors: Milad Roohi (University of Nebraska-Lincoln) and Christine Wittich (University of Nebraska-Lincoln)
Abstract: Flooding represents a critical hazard to transportation infrastructure, routinely precipitating bridge failures, network disruptions, and cascading socio-economic losses. Historical data indicates that approximately 53% of U.S. bridge failures between 1989 and 2000 were attributed to flooding and scour. This vulnerability was highlighted by the 2019 Midwest floods, where Nebraska suffered damage to 27 bridges and over $190 million in highway losses, and more than $1 billion in total statewide damages. To mitigate these risks, there is an urgent need to transition from asset-level vulnerability assessments to system-level resilience strategies. However, the computational scale of such networks, comprising over 15,000 bridges in Nebraska alone, renders high-fidelity modeling of every individual asset infeasible. To address this challenge, this study presents a portfolio-based computational framework for quantifying transportation network resilience. We develop scalable fragility functions, including suitable intensity measures and engineering demand parameters, for the entire system by grouping bridges into suitable structural classes based on National Bridge Inventory (NBI) data. These fragility models are integrated into a network-level assessment that evaluates performance using multi-disciplinary resilience metrics, including mobility, critical service accessibility, economic disruption, and recovery time. Ultimately, this work advances network-based resilience planning by bridging the gap between component-level reliability and system-wide performance.
PhD Student Arizona State University
Advancing Urban Flood Management and Design through Coupled Infrastructure and Storm Modeling
Co-Authors: Mikhail Chester (Arizona State University), Mattheus Porto (Arizona State University), and Giuseppe Mascaro (Arizona State University)
Abstract: A systems-level management and design of stormwater infrastructure requires modeling tools that capture the impacts of spatiotemporally varying storms on the entire stormwater drainage network. This includes the representation of wind-induced power outages that can trigger pump failures, which is not currently comprehensively acknowledged in infrastructure management practices. We addressed this gap by developing (1) a coupled mechanistic model of stormwater and power networks, named PSI-Cascade, and (2) frameworks to generate high-resolution precipitation and wind fields. After describing the model structure, we showed its capabilities in Phoenix, AZ, for a set of historical storms with documented street flooding impacts. We devote particular attention to generating realistic wind gust fields that cause power outages, a challenging task due to the lack of dense observations and the isolated nature of strong windstorms. We then demonstrate how PSI-Cascade can be utilized to identify areas with a higher risk of flooding-driven transportation disruptions, and subsequently compare how various solutions mitigate the consequences of flooding.
Masters Student Oklahoma State University
Regional-Scale Vulnerability Assessment of Electric Transmission Infrastructure subjected to Flood Hazards in Oklahoma
Co-Authors: Maha Kenawy (Oklahoma State University) and Brian Giffin (Oklahoma State University)
Abstract: Overhead electric transmission infrastructure is highly vulnerable to physical damage during severe weather events, sometimes leading to substantial electric power disruptions and lengthy recovery. Extreme precipitation events, specifically, tend to trigger cascading threats such as flooding and debris flows that may be devastating to electric transmission assets. This study assesses the risks of structural damage to electric transmission networks in Oklahoma due to rainfall-induced flood and debris loads. We create a regional-scale framework which integrates flood inundation maps generated using the National Water Model with nonlinear structural simulations of the resulting hydrodynamic and debris loads on steel lattice tower structures. We quantify the uncertainties associated with extreme limit states of the tower structures due to variations in the loading conditions including flood depth, flow velocity, and debris sizes using the NHERI SimCenter Quantified Uncertainty with Optimization for the Finite Element Method (QuoFEM) tool. The simulation results are used to develop structural fragility curves for electric transmission towers subjected to extreme precipitation-induced loads, and regional-scale flood risk maps for electric transmission networks in eastern Oklahoma. These maps can support transmission line hardening and expansion efforts and improving the long-term resilience of the electric grid against extreme weather hazards.
Professor University of Kansas
Validating water quality predictions in a synthetically generated drinking water distribution system
Co-Author: Justin Hutchison (University of Kansas)
Abstract: Public infrastructure plays a significant role in community resiliency and response to disruptions. Drinking water distribution systems (WDS) are needed to maintain public health through the resilient delivery of safe, clean drinking water. Assessing the resiliency of these water systems requires high-spatial-resolution maps for use in drinking water distribution simulation tools such as EPANET. Unfortunately, not all utilities have the data available to construct these maps.
To address potential limitations in high-resolution data, this work presents a Python-based script package, ROAD2H2O, that generates an EPANET-compatible water distribution map. The underlying assumption was that water distribution pipes run parallel to roadways. Road network information was collected from OpenStreetMap and converted into links and nodes representing the distribution system. The script incorporates additional model input parameters, including elevation, population, and commercial demand, to appropriately size the WDS. The pipe sizing was determined by minimum state or regional fire flow requirements, followed by iterative pipe sizing calculations. The generated map was validated against a real community distribution system.
Further, water age was assessed across a synthetically generated system as a metric for potential variations in water quality. Water quality assessment validation was conducted with weekly sampling of monochloramine residuals at sampling locations throughout the real WDS as a surrogate for water age. Over a seven-day simulation period, the system had an average water age of 38.6 hours (33.0 hours when demand-weighted), close to the real system’s operating goal of 48 hours, demonstrating the simulated system’s potential model water quality parameters.