A Decision Framework for Improving Resilience of Civil Infrastructure Systems Considering Effects of Natural Disasters

There can be massive ramifications from natural disasters for civil infrastructure systems. Damage to infrastructure components, such as roads, bridges, levees, dams, buildings and houses, causes economic loss and disrupts critical lifelines. It is necessary to build more resilient infrastructure systems capable of withstanding and recovering from damaging effects caused by natural hazards. This thesis addresses the need for a framework capable of determining the decisions that can improve system resilience and reduce system-wide risk to natural disasters. A brief background is given of the concept of interdependent infrastructure components and the importance of including interdependencies in integrated modeling of infrastructure systems. Different categories of past modeling efforts are reviewed, with categories defined by both model structure and abilities, especially the abilities to change system behavior, prescribe decisions, and incorporate uncertainty in analysis.One of the issues with current infrastructure modeling is a deficiency in defining meaningful and varied system performance functions. Different types of quantitative performance measures (or metrics) for infrastructure systems are considered. Metrics are developed for characterizing serviceability (the potential for an infrastructure to fulfill the lifeline needs), property damage, travel time, and cost of upgrades and retrofits. These metrics are intended to evaluate the collective performance of the components of the system, and to prioritize and determine the effect of decisions such as upgrades and retrofits.Models are developed that are realizations of different scenarios, each based on particular combinations of the infrastructure system and its traits of interest, especially the interdependencies and interrelationships. The range of models developed begins with systems that have a protective infrastructure, such as a levee, and flexibility allows the modeling framework to consider numerous other situations. The definitions of the decision variables and the model expressions allow the formulation of generic mathematical optimization models that could be solved using mathematical programming techniques dependent on input data. An illustrative example demonstrates the validity of the concepts.