Multi-physics and multi-objective optimization methodology for sodium-cooled fast reactor conception - 14176

In the French ASTRID project framework, CEA, AREVA and EDF are currently working together to design an innovative Sodium Fast Reactor. The objective of this project is to build a robust safety demonstration for those reactors while proving their economical efficiency. Therefore the design process must conciliate multiple goals which are often opposed. Since 2010 the reference concept for ASTRID is the CFV, a low sodium void worth core which presents radial and axial heterogeneities. Some design optimizations are still led on this concept to enhance its performances. The optimization process is commonly realized on the basis of expert advices and local parametric studies. Indeed, a strong knowledge of physical phenomena avoids a large number of expensive simulations. However there is no guarantee to obtain the most reliable design since the design study space is not fully explored. To solve this issue EDF started to develop in 2009 a method for global multi-objective optimization: the SHADOC-based Design Development System (SDDS). This method allows to evaluate quickly the performances of a wide set of different cores, i.e. a design study space. To do so, surrogate models are built using data from global parametric studies. SDDS evaluates the performances of each core, defined by a set of parameters (pellet diameter, fissile height...), in the study space. Then the user can select the most compliant core with a given set of requirement (maximum fissile radius, sodium void worth and so on). In addition to the core optimization aspect, SDDS is also a useful tool for physical analysis since it performs parametric study on a wide design study space. This paper presents the evolutions brought to SDDS since 2012, especially the modification of the neutronic scheme from a 2D RZ simulation to a 3D one which allows the chaining between neutronic, transient and thermal-mechanical simulations. The construction of surrogate model has also been reworked and modified to reduce uncertainties on core performances evaluation. This paper finally presents a practical case where SDDS is used to perform the optimization of a CFV like core. (authors)