Automated Factor-Based Sensitivity Analysis

Large defense systems composed of numerous integrated subsystems and algorithms have complex interdependencies that require rigorous analysis to fully comprehend. Uncertainties in the inputs of component 1 can affect its output in such a way that component 2, dependent on component 1’s output, is also affected. This cascading ripple effect can induce significant changes in overall system behavior, possibly forcing the system out compliance with its top-level performance requirements. A sensitivity analysis explores the relationships between information flowing into and out of a process and studies how output uncertainties can be allocated to different sources of input uncertainty. By isolating individual algorithms and controlling input data flow, one can force the system down processing paths that otherwise would not have been exercised. The outcomes of these off-nominal cases can then be aggregated and analyzed to establish behavioral trends that aid in the overall understanding of sensitivities inherent in an integrated system. Sensitivity analysis design and execution, though a structured and systematic process, can be incredibly complex and time-consuming. Automating experiment design and execution through a combination of artificial intelligence and model-based engineering speeds up the process and reduces complexity, thereby alleviating the burden on the user and promoting more robust study of the system under test. In this proposal, we offer IDT’s design of an automated, interactive software tool that applies factor-based sensitivity analysis to mission-critical Aegis Ballistic Missile Defense (BMD) combat system algorithms. Approved for Public Release | 20-MDA-10643 (3 Dec 20)