System-of-systems architecture analysis and design using Bayesian networks

A System-of-Systems (SoS) is characterized both by independence and by interdependency. This interdependency, while allowing an SoS to achieve its objectives, also means that failures can cascade throughout the SoS, creating development delays or operational system failures. New approaches are needed to evaluate the risks associated with cascading system failures in the SoS, both during development and during operation. A method based on Bayesian networks (BNs) is adopted to analyze interdependencies between constituent systems. The analysis in the operational domain provides the means to analyze the performance of the new SoS and its evolution with time. On the other hand, the analysis in the developmental domain deals with the risk in project management, associated with schedule and costs. The results from the two analyses can support the SoS system engineer in the design of a desirable SoS which meets the needs of the stakeholders and has a reasonably small risk of project schedule and cost overruns. Evaluating SoS alternatives is a necessary step in the SoS architecture analysis and design. Various aspects of the overall SoS performance can be used to evaluate SoS alternatives. In this study, we use the concept of resilience as one of the overall SoS performance aspects. Resilience has sparked interest in the last few years because of its ability to demonstrate SoS performance such as flexibility, robustness, and repair-ability. Resilience is generally understood to indicate the capacity of a system to survive, adapt and grow in the face of change and uncertainty. In this dissertation, SoS resilience is defined as the capability of a system to survive in the face of constituent systems failures. In other words, SoS resilience defined here is only considered as the subset of the general resilience. A conditional resilience metric and several resilience patterns are used to quantify SoS resilience. A conditional resilience metric measures each constituent system\u27s contribution to overall SoS resilience, and a resilience pattern shows how the SoS performance degrades as systems fail. In addition to the resilience metric and patterns, trade space analysis is proposed to assess potential alternatives to optimize the balance between two or more aspects, such as operational capability and financial efficiency. A Pareto frontier is formed and used to visualize such trades. The application of the proposed approach is demonstrated using two SoS-oriented case studies: a Naval Warfare setting with air, sea, and undersea assets and a Next Generation Air Transportation setting. Study results provide evidence that the proposed resilience metric and patterns, and the Bayesian interdependency analysis, can assist SoS system engineers in identifying architectures and component systems that increase expected SoS resilience