When designing a new reactor, one of the major objectives is to reduce potential radioactive releases from severe accidents. For Gen IV reactors, in the event of a severe accident, there should be no need for population evacuation and sheltering shall be limited in time and space. Reactor design is based on the defense in depth (DiD) safety principle. The application of this principle leads to implement design provisions, in order to reduce severe accident frequencies of occurrence. This is the object of the first three DiD levels. Whatever the provisions taken in the design, the application of the DiD principle requires to postulate at the fourth DiD level the occurrence of severe accident, and to provide sufficiently reliable means to mitigate their consequences. The reactor design aims to put in place mitigation provisions with regard to all possible situations of severe accident. Nevertheless, there may be Severe Accident situations for which no reasonable mitigation means could be implemented and which could lead to large and early radiological releases, with insufficient time to organize population protection measures , or to massive radiological releases leading to population displacement over a significant period of time or in an extended area. These situations need to be identified at an early stage in the design process in order to make them extremely unlikely with a high level of confidence by appropriate design and organizational provisions. Practical elimination is an approach which, from the outset of the reactor design studies, identifies these severe accident situations that cannot be mitigated under economically reasonable conditions. This approach should be applied preferably for reactors in conceptual phase.

This paper issued by the GCFS (the French group EDF/CEA/AREVA NP on safety of generation IV reactors) proposes the methodology to apply for Gen IV reactors. It first consists in establishing a list of situations to be practically eliminated. This list should be confirmed by the safety authority, at the first stage of the design activities. Then guidelines are given to ensure demonstrations for all these situations. These demonstrations are mainly based on a deterministic approach, and aim firstly at making these situations physically impossible.

Then, as an example, the application of this methodology to the sodium fast reactor project ASTRID used during its conceptual design phase is given. The list of practically eliminated situations considered on ASTRID is explained, with some examples of design improvements associated.

In conclusion it appears that, beside the DiD approach, the application of this methodology at an early stage of design studies is an important element to improve safety of generation IV reactors.