Probabilistic AI Reliability Analysis in MBSE and SysML

SUMMARY & CONCLUSIONS Systems engineers, using model-based systems engineering (MBSE) environments and languages such as Cameo and Systems Modeling Language (SysML), can effectively model and simulate structural and functional relationships within complex systems. However, the MBSE tools are insufficient in representing uncertainty in these relationships or uncertainty in the system performance under varying environmental conditions, which is critical in understanding performance and reliability of the created designs. In particular, reliability engineers using MBSE tools must manually map SysML designs into reliability block diagrams and analyze them in a new suite of tools. This process is error-prone and leads to a disconnect between the system design and reliability analysis processes. The solution demonstrated in this paper facilitates ProbSysML, a probabilistic AI plug-in to Cameo that supports evaluation of system performance and reliability under uncertainty using probabilistic inference within Cameo. The key elements of the plug-in include: 1) expressive representation of uncertainty in system designs and their impact on system performance and reliability, 2) automated conversion of this representation into probabilistic models, and 3) probabilistic inference for evaluating system performance and reliability based on probabilistic outcomes. The plug-in is demonstrated on an example of designing a solar panel for a satellite such that the reliability of the design is maximized while considering survivability trade-offs between power requirements, weight, and degradation due to solar storms.