Development and Selection of Models of Infrastructure for Risk Analysis
Advisor: Paolo Gardoni
Abstract
Critical infrastructure are vital to the functioning and prosperity of modern societies. These include electric power infrastructure, potable water infrastructure, wastewater infrastructure, and transportation infrastructure. These critical infrastructure are vulnerable to damage from natural and anthropogenic hazards, including earthquakes, hurricanes, wildfires, tornadoes, landslides, and many others. Significant losses after past disrupting events have highlighted the importance of critical infrastructure, and moreover, the need to improve performance in the aftermath of disrupting events. Mitigation measures and recovery optimization can reduce the impact of disrupting events. Because these events are rare and uncertain, risk analysis uses models to support mitigation planning and recovery optimization. This dissertation focuses on creating and selecting models of infrastructure for risk analysis.
First, we focus on how models of infrastructure can be generated from available data. Commonly, the direct data to define the infrastructure are not available. To address this issue, this dissertation develops systematic procedures to generate network representations of infrastructure. Some of the proposed procedures may be applied to any type of infrastructure, while other procedures are specific to the type of infrastructure being generated. In this dissertation, we use wastewater infrastructure as an example. We implement the proposed procedures. We adapt and use algorithms from graph theory, and we use detailed geospatial data that are generally available. We provide reference to standard datasets with consistent coverage, so the proposed network generators can be directly used for many locations of interest. We first propose a community-scale procedure where we divide the total area into subareas based on mainlines in the street network. We determine the network topology that connects the subareas. We demonstrate the community-scale procedure for Irving, Texas, a mid-sized city with approximately 250,000 residents. We also propose a fine-scale procedure to generate the network in each subarea. We demonstrate and validate the proposed fine-scale procedure by generating a portion of the wastewater network in Seaside, Oregon, a small city for which the real network data are available. Together, the community-scale and fine-scale procedures can be used to generate a full network.
We also consider which model should be used for risk analysis in two example circumstances. First, we consider the impact of localized damage with a known location on wastewater infrastructure for which no direct information is available. In this case, the main source of uncertainty is in which components of the infrastructure are in the damaged area. To account for this, we adapt the community-scale procedure to generate multiple network realizations and produce a map of the probability of disconnection of each subarea. We also combine the community-scale and fine-scale procedures. We demonstrate the procedure for Irving, Texas. Finally, we consider the impact of an earthquake on a known network. In this case, we consider various combinations of network granularity and fidelity of the flow equations. For the network of Seaside, Oregon, we identify a granularity and fidelity at which the results are similar to the most detailed case. This model is an efficient representation for risk analysis. In general, the dissertation considers what model should be used to represent infrastructure for risk analysis and provides procedures to generate these models from available data.
