Implementation of more automatic methods to validate criticality safety codes
Titre du congrès :ICNC 2007
Ville du congrès :Saint-Petersbourg
Date du congrès :28/05/2007
The CRISTAL criticality safety package has been developed by IRSN and CEA, in collaboration with COGEMA Company (AREVA group). This package consists of two calculation routes, using nuclear data based on the JEF2.2 nuclear data file:
- a standard route based on group-wise cross-sections (APOLLO2-MORET 4 or APOLLO2-Sn),
- a reference route using the TRIPOLI4 point-wise energy code.
APOLLO2 is a multi-group deterministic code allowing the calculation by the collision probability method (CPM) of macroscopic cross-sections, which are used in the MORET 4 Monte Carlo code. The validation of the APOLLO2-MORET 4 standard route, under the responsibility of IRSN, relies upon a database of more than 1800 benchmark-models (mainly taken from the ICSBEP Handbook). This validation database contains a broad variety of configurations in terms of fissile medium, neutron spectrum and geometry. It aims to cover a large range of industrial configurations in the nuclear fuel cycle.
The validation of the APOLLO2-MORET 4 route is an on-going process and leads to a growing database. For instance, in 2005, "only" 1300 benchmark-models were in the database. The management of data files and the analysis of all the results were done by "hand", but the growing number of cases made this work more and more difficult, and, as a consequence, less and less reliable. Thus, it became necessary to re-organize the database, to implement a procedure to guaranty the quality of the data files, and moreover to automate the calculation and the treatment of the entire database.
This re-organization is the first part of a more ambitious project that will lead to the development of more advanced tools dedicated to the administration of the validation database and its analysis.
The need for such tools is illustrated through the analysis's that are currently undertaken on the validation database.
A first example concerns the comparison between calculation platforms, since it has been decided to move the database on more powerful computers. The database has been run on the old and new platforms without any change in the inputs. In some cases, significant keff's discrepancies (up to 0.007) were observed and made possible to identify insufficient convergences for these cases.
A second example concerns the comparison of the results obtained with different versions of the codes. Note that an advanced tool could also allow the comparison between different codes and/or different libraries.
The last example focuses on the comparison between some calculations and experiments versus different parameters in order to highlight some trends.