Due to stochastic influences or more specifically due to aleatory uncertainties on
- the behavior of the technical system,
- the performance of the plant operators,
- the behavior of the physical-chemical process and
- on external factors
a variety of potential event sequences has to be considered, and the assessment of the consequences of an accident scenario can only be probabilistic.It is obvious, that the spectrum of event sequences may be tremendous, if aleatory uncertainties exist, for instance, on the physical-chemical process or on the timing of system function failures and of human actions.
Only if a PSA is able to account for all the sequences which may evolve and to rank the sequences according to their likelihood, it can provide a well-founded probabilistic assessment. But the conventional PSA approach may not be capable of doing that, because
- it is not an integral analysis where the interaction between the plant behavior, operator actions, and stochastic influences can be captured,
- it has no time axis and, therefore, cannot adequately account for the influence of the timing,
- it is decoupled from the physical-chemical process and therefore cannot adequately account for the influence of variations in the process, and
- it strongly depends on the experts who build the models (expert-to-expert variation).
The MCDET method allows for an integral probabilistic analysis of accident scenarios. The combination of Monte Carlo (MC) simulation and the Discrete Dynamic Event Tree (DDET) approach is capable of accounting for any aleatory uncertainty at any time. The implemented MCDET modules can in principal be coupled with any deterministic dynamics code simulating the behavior of a nuclear power plant under accident conditions (like, for instance, ATHLET, RELAP or MELCOR). Beside aleatory uncertainties, MCDET can also consider epistemic uncertainties which determine how precise probabilistic assessments can only be provided due to the knowledge uncertainties involved in the calculation.
The MCDET modules were supplemented by a crew module which enables calculating the dynamics of crew actions depending and acting on the plant dynamics as modeled in a deterministic code and on stochastic influences as considered in the MCDET modules. The crew module allows for running situation-dependent sequences of human actions as they are expected for a dominant cognitive behavior.
The capacity of the MCDET modules and the crew module has been demonstrated by several applications including a large-scale station black-out scenario and the analysis of an emergency operating procedure.