Reviews of safety criteria and qualification of models can be made on the basis of experimental data coming from three main types of tests, which are complementary and necessary :
- out of pile experiments on small scale samples, in order to identify and quantify the basis mechanisms,
- in test reactor separate effects experiments, on short fuel rods or sections,
- in test reactor integral tests, taking into account a maximum of mechanisms, and notably the bundle geometry.
This methodology can be more or less applied to the different topics included in this paper : power ramps and power cycling, Anticipated Transients Without Scram (ATWS), LOCA, RIA and severe accidents. Current status and pending questions related to each topic have been developed in the paper. Associated future experimental needs have been set out for current designs of reactors (Generation II). For Generation III reactors (extension of current reactors design), passive safety features have been introduced, so it is reasonable to think that safety needs will be covered by those of Generation II. Safety needs concerning future designs (Generation IV) are difficult to anticipate and are consequently not discussed in this paper.