ABSTRACT
   Severe accident Steam Generator Tube Rupture (SGTR) sequences are identified as major contributors to risk of Pressurized Water Reactors (PWR). Their relevance lies in the potential radioactive release from reactor coolant system to the environment. The radioactive particles can be partially retained over the surfaces they find along the path from the reactor coolant system, in particular within the secondary side of the steam generator, even in the absence of water. Lack of knowledge on the source term attenuation capability of the steam generator has avoided its consideration in probabilistic safety studies and severe accident management guidelines. As a consequence, the steam generator filtering capability is not usually taken into account either in the probabilistic risk assessment of nuclear safety or in the severe accident management guidelines. 
   This thesis is a contribution to the technical understanding and quantification of the natural processes mitigating the consequences of SGTR accidents. It describes the main activities and results of a theoretical program focused on modeling the aerosol retention in the break stage of the secondary side of a dry steam generator.  The model developed, called ARI3SG, is a semi-empirical, Lagrangian model based on the filter-concept approach. It is built to compute retention efficiency according to dominant aerosol phenomena and gas fluid-dynamics underneath. Fluid dynamics have been analyzed through thorough 3D simulations with the FLUENT 6.2 code, which were validated against ad-hoc experimental data. 
   The model performance has been assessed through a verification process that has demonstrated its robust and sound behavior. Predictability was also assessed by comparing its estimates to open data and by analyzing the effect of associated uncertainties. Data-model comparison has been shown to be satisfactory and highlight the potential use of an ARI3SG-like formulation in system codes. Through a random sampling of the input variables of the model (i.e., inlet velocity and aerosol size and density), a theoretical correlation has been derived as a function of the Stokes and particle Reynolds non-dimensional numbers. As the average relative deviation with respect to ARI3SG is less than 7%, the correlation provides a useful way of implementing ARI3SG within severe accident system codes, like ASTEC and/or MELCOR.
   The work performed in this thesis is framed in the CIEMAT contribution to the ARTIST and ARTIST II projects (2003-2010) which were supported by the Spanish Nuclear Regulatory Commission (CSN).