Identifying Components Important for System Resilience

Abstract

The resilience of a system is often assessed using models that consist of numerous interdependent system components. To better understand the system’s resilience, the performance of each component needs to be evaluated. Such an evaluation requires significant component-level data collection and modeling efforts, which are, in some cases, restricted due to privacy or security concerns. However, the performance of certain system components might be irrelevant for system’s resilience. This paper presents a method that can identify such irrelevant components. More specifically, we focus on identifying those components whose vulnerability and recoverability are irrelevant for system’s disaster resilience, with minimal information on these two component parameters. The presented method employs a Sobol’ sensitivity analysis using uninformed input distributions to initially rank the system’s components. An iterative heuristic upper and lower bound search is performed next to identify the subset of highly ranked components whose vulnerability and recoverability significantly affect system’s resilience. The efforts aimed at better understanding the performance of system components that are not in the “important components” subset can be avoided without a major impact to system’s resilience assessment, simplifying the system model and preventing unnecessary component-level data collection and modeling efforts. The method is illustrated on a virtual community whose disaster resilience is assessed in terms of its ability to meet its community resilience goals.