A New Resilience Metric to Compare System of Systems Architecture

Bryan C. Watson,Ashray Chowdhry,Marc J. Weissburg,Bert Bras,Bryan Watson,Marc Weissburg

DOI: https://doi.org/10.1109/jsyst.2021.3062444

IF: 4.802

2022-01-01

IEEE Systems Journal

Abstract:By combining independent systems (e.g., financial systems, power infrastructure, transportation networks), systems of systems (SoSs) can provide additional services. The ability to withstand faults, minimize service disruption, and efficiently recover (resilience) is a key performance characteristic. Current resilience metrics, however, require improvement. Each SoSs is approached with different assumptions about fault magnitude and recovery strategy. Thus, engineers currently have difficulty answering the basic question: Between alternative designs A and B, which is more resilient? SoS resilience metrics should reflect that the system interfaces are the most potent leverage point, have a standardized fault and recovery strategy, incorporate SoS dynamics, and apply to both natural and artificial systems of systems. In response, we develop the system of systems resilience metric (SoSRM). SoSRM methodology uses the definition of SoS to identify appropriate fault locations and graph theory to define fault and recovery duration. SoSRM is applied to two case studies to demonstrate application. First, an intertidal oyster reef ecosystem's system dynamics model is presented to demonstrate SoSRM methodology. Next, a military SoS demonstrates the application of SoSRM to a stochastic multi-agent system. Validation simulations show that SoSRM reflects expected fault impact and that methodological choices for fault duration do not drive SoSRM results. This article presents a new design tool capable of comparing the resilience between two SoS architectures.