A Time-domain Modeling and Simulation Framework for Comparative Analysis of Prognostics, Reliability and Robustness in System Design

Redundancy is an effective, high-level solution to the requirement for reliable safety-critical systems, but it comes at the cost of Size, Weight and Power (SWaP) and reduced capability. A modeling and simulation framework was developed to address the need for robust design alternatives to redundancy. Robustness, in our application, is treated as the insensitivity of the performance with reference to specification. The necessity to characterize both reliability and robustness in the same framework has resulted in a time-domain simulation approach to modeling behaviors associated with unreliability and a lack of robustness. The incorporation of these features offers a novel insight into potential applications of prognostic technology. Further development of this approach has the potential to allow designers to choose how risks associated with failures are mitigated, by redundancy, robustness, or prognosis. By modeling the life of parts, the factors that impact them and the resulting behaviors, the observability and predictability (even controllability in the case of optimized, fault-tolerant, closed-loop control) of faults and failures is identified. Designers can determine which parts of a system would benefit from prognostic health management (PHM) technologies, adaptive / tolerant features to yield robust design, or redundancy based approaches. The complex causality in the models requires a Monte Carlo approach analogous to the simulation of fleets of systems; this, combined with the ability to simulate systems made from new and old parts, can inform strategies for condition-based maintenance (CBM). We present the mathematical modeling concept and the simulation framework which permits comparative assessment of reliability, robustness and prognostics. The multi-hierarchical, systems integration aspects inherent to the concept make this technique highly applicable to real- world dynamic systems. The framework also supports statistical, standards based and physics-of-failure descriptions of stress, aging, fault and failure behaviors in a unified way. There are challenges to be overcome in realizing the benefits of this approach to model-based system design. Issues of model validation, data availability and computational burden are recognized and discussed. As we show, these challenges can be overcome to produce new design tools providing better products and transparent project quality.