Zahra Mohagegh1, Mohammed Modarres1, and Aris Christou1
1University of Maryland, Department of Mechanical Engineering, College Park, MD, 20740, USA
Abstract:
The modeling of dependent failures, specifically Common
Cause Failures (CCFs), is one of the most important topics in
Probabilistic Risk Analysis (PRA). Currently, CCFs are treated
using parametric methods, which are based on historical failure
events. Instead of utilizing these existing data-driven
approaches, this paper proposes using physics-based CCF
modeling which refers to the incorporation of underlying
physical failure mechanisms into risk models so that the root
causes of dependencies can be “explicitly” included. This
requires building a theoretical foundation for the integration of
Probabilistic Physics-Of-Failure (PPOF) models into PRA in a
way that the interactions of failure mechanisms and, ultimately,
the dependencies between the multiple component failures are
depicted. To achieve this goal, this paper highlights the
following methodological steps (1) modeling the individual
failure mechanisms (e.g. fatigue and wear) of two dependent
components, (2) applying a mechanistic approach to
deterministically model the interactions of their failure
mechanisms, (3) utilizing probabilistic sciences (e.g.
uncertainty modeling, Bayesian analysis) in order to make the
model of interactions probabilistic, and (4) developing
appropriate modeling techniques to link the physics-based CCF
models to the system-level PRA. The proposed approach is
beneficial for (a) reducing CCF occurrence in currently
operating