MBSA Strategies for Handling Dormant and Multiple Dormant Failures

Maintenance allows keeping systems safe, as failed equipment are usually repaired quickly after their detection. However, an important threat to system safety is the presence of dormant failures. These failures, undetected during system operation or not annunciated when they occur, cannot be addressed immediately and are only verified during scheduled maintenance, thus resulting in a longer exposure time compared to standard failures.
Currently, modellers use construction patterns from standards of their work domain within Fault Tree Analysis (FTA) tools to address these failures, and some FTA tools offer automated generation of such patterns. Model-Based Safety Analysis (MBSA) should provide a unified approach to handle them.
This paper proposes several modelling and simulation approaches to represent dormant failures in all RAMS domains in which they are applicable.

• The first approach uses a dynamic modelling pattern that expressively captures the behaviour of a dormant failure. The component is modelled in a form where it undergoes regular maintenance. Simulation is performed using Monte Carlo methods.
• The second approach adapts ARP4761A-proposed modelling patterns for fault trees, enabling dynamic MBSA calculations that consider event sequencing.
• The third approach models components traditionally, with a failure event, and applies probabilistic laws to compute cut-set probabilities while accounting for dormancy. Some tools already implement such laws; we analyse their strengths and limitations and propose a general formulation, including average and maximum exposure times.
  Especially, we tailor these approaches to be able to not only tackle single dormant failures, but also double dormant failures, as the latter are more difficult to compute and require a more complex representation.
  We apply these methods to toy examples using Cecilia Workshop or SimfiaNeo, and analyse the pros and cons of each approach.

Keywords: MBSA, FTA, Dormant failures, Latent failures, Safety, RAMS.