Proceedings of the Eighth International Conference on Probabilistic Safety Assessment & Management (PSAM)
1 Development and Structure of the German Common Cause Failure Data Pool (PSAM-0020)
Download citation file:
- Ris (Zotero)
- Reference Manager
In probabilistic safety analysis (PSA) of highly redundant systems common cause failures (CCF) of components in general contribute to a large extent to the systems unavailability. As CCF events are rare events, operating experience of “classes” of components of the same type (like centrifugal pumps or motor operated gate valves) combined from “similar” plants and systems is used to generate CCF data with sufficient statistical quality.
In the German approach, the operating experience of all German nuclear power plants since the beginning of their operation is combined although these plants belong to different types and generations and although maintenance and operational practices have advanced. In order to take into account plant and system specific differences in technical and operational details, specific boundary conditions were introduced to be able to adjust the generic CCF data to component specific conditions in the analysed plant and system.
Most of these boundary conditions consider modern maintenance strategies and surveillance methods which were introduced in the plants to allow an early detection of potential common cause failure mechanisms before these mechanisms develop to multiple failures. Other boundary conditions take into account system specific technical aspects like quality of the process medium which may have influence on the probability of occurrence of some types of CCF mechanisms.
In a common effort of all German utilities and GRS the existing German CCF event data base recently was re-evaluated in order to introduce these aspects in the German CCF data pool. GRS discussed each of the originally approximately 250 CCF events in the pool with PSA, system and maintenance experts and operational staff from each German utility. Plant specific maintenance policies and specific system characteristics could be identified which would reduce or increase the probability of occurrence or the impact of a CCF mechanism.
For each discriminating aspect a respective boundary condition was formulated. Then all involved experts firstly assessed the event as it occurred and secondly assessed a hypothetical event showing the same CCF mechanism under the assumption that the new boundary condition is effective. Thus, two quantitative event assessments based on the component impairment vector approach of the international CCF data exchange project ICDE were introduced in the CCF event data pool for one observed CCF event.
The result of these discussions was that for about half of the events in the original CCF event data pool additional boundary conditions were defined. For some of these events up to four different boundary conditions were specified. In total there are now 321 event assessments in the German CCF event data pool.
First experience with the new data pool could be gained by calculating CCF probabilities for different types of components with the coupling model. Compared to coarser former results, the results achieved now are more realistic as they differentiate more for specific characteristics of the analysed components.
The paper describes the concept of boundary conditions and presents the common activity of the German nuclear power plant utilities and the GRS to set up the new CCF event data pool. Furthermore, some comparative results from the calculation of CCF probabilities for different boundary conditions are presented and an outlook on ongoing activities of the German utilities and GRS to update the CCF event data pool is given.