Faire avancer la sûreté nucléaire

La Recherchev2


Improving quality of NPP PSA by international comparisons



Email :

Mot de passe :

F. Corenwinder, P. Dupuy, J.M. Lanore1, D. Gryffroy, P. De Gelder2, H.M. Kussman, S. Rokita3
1IRSN, 2 AVN Belgique, 3NNR Afrique du Sud,
PSAM - 8th International Conference on Probabilistic Safety Assessment and Management, Nouvelle Orléans, USA, 14-19 mai 2006,
Rapport DSR N° 125


  A PSA comparison of various PSAs performed by different teams for nuclear power plants with a similar design has been conducted for three Framatome 900 MWe PWRs. This exercise appears to have been very constructive in particular for improving PSA quality and credibility.
The plants under consideration are the French 900 MWe PWR-series, the Belgian Tihange-1 PWR and the South African Koeberg PWR. The PSA comparison was conducted among the regulators of these countries, in a first stage for the French and Belgian PSAs, subsequently the comparison was extended to the South African PSA.
Up to the present, the comparison has been limited to internal initiating events and power states. The methodology used for these comparisons is a “top-down” approach, starting from an overall comparison of the results and trying to explain the most striking differences descending to an appropriate level of detail.
The differences identified may be due to differences in design, assumptions, completeness, data, or a combination of several reasons.
Many insights may be derived from this PSA comparison:
·         Firstly, a general consistency of the models is observed.
·         Many differences in the PSA results can be attributed to plant specifics;
·         Design differences have in several cases had an unexpected effect, often being either a large effect of small differences or the contrary.
·         Similarly, differences in assumptions, data, or completeness do not always have the expected effect.
·         In many cases, a combination of differences is found, with an impact on the results that is difficult to foresee.
Such a detailed PSA comparison is, in fact, complementary to a very detailed review, yielding several additional insights attributable to the involvement of different PSA teams. The PSA comparison allows for the identification of potential PSA improvements as well as beneficial plant modifications, along with the effect on the risk profile to be expected. The comparison also indicates that improvements in PSA quality should not always be sought in more detailed or more realistic models, but may also be found in clarified presentations or improved justifications. In general, the insights allowed by the comparison are important aspects for effecting improvement in the PSA with respect to completeness, internal consistency, documentation and presentation.
The paper illustrates, by means of several specific examples, the value of this comparison exercise for improving PSA quality and usefulness. These examples particularly demonstrate the impact of:
·         initiating event definition and frequency,
·         design specifics,
·         sequence analysis, including modelling assumptions and success criteria,
·         data and common cause failures
The main benefit of this PSA comparison is an increased confidence in the studies and consequently a wider use of the PSA results.