Integrate Common Cause Failures on Demand into Dynamic Models

Investigate how to integrate on-demand common cause failures—failures occurring upon demand signals—into dynamic probabilistic risk assessment models, specifying their representation within shock-based CCF frameworks such as the Atwood model.

Background

Current dynamic PRA implementations in the paper focus on time-distributed failures and shocks affecting running components. The authors note that integrating CCF events that occur on demand remains uncertain, which is critical for scenarios where components are challenged at specific times.

Resolving this question would extend the dynamic CCF modeling framework to demand-driven contexts consistent with PRA practice, aligning shock occurrence and conditional failure behavior with demand events.

References

Nevertheless, some open questions concerning how to model CCF in time remain and further research should be performed in order to :

  • Validate the Atwood model for the case of CCF group of different sizes;
  • Validate the proposed staggered mechanism for longer mission times (which may be affected by the lack of operating data);
  • Investigate how to integrate CCF on demand;
  • Explore how to model CCF in case of repair;
  • Compare parametric model (Alpha) vs shock models (Atwood) for dynamic studies.
— Modeling Common Cause Failure in Dynamic PRA  (2406.08879 - Picoco et al., 2024) in Section 5. Conclusions and Perspectives

This may also reflect our independence assumption: correlated failure modes that simultaneously disable filters and trigger character fragility could amplify the CMF channel's importance, an extension we leave to future work.

— Rules or Character? Scaling Laws for AI Safety Design  (2608.13345 - Takahashi et al., 13 Aug 2026) in Section Discussion, subsection “The Quantitative Role of Common-Mode Failure”