Soil-structure interaction (SSI) consists of an important step on the evaluation of seismic response of structures and equipment of nuclear power plants. In the current engineering practice, strong hypothesis such as vertical incidence plane waves with equivalent lateral boundaries and horizontal soil stratification is usually adopted so as to simplify the SSI studies. However, in reality, the incident wave field to be applied as an input excitation is more complex. Thus, it is necessary to properly model the complete wave propagation problem from the source to the site of interest which could be a computationally costly procedure. A good compromise between the accuracy and efficiency from a computational point of view is examined in this work.

For this purpose, a weak coupling approach based on the domain reduction (DRM) introduced by Bielak et al. [1] is used in this work. The definition of a 3D complex incident wave is provided through a spectral element method. It is used so as to study wave propagation from source to site in a regional scale model. Then, the 3D incident wave field is then injected in a finite element code (code_aster) so as to study the response of a reduced domain model in a more accurate way. The aim of this paper is to verify the weak coupling methodology implemented in code_aster for some cases related with i) the used soil behavior and ii) the mesh conformity between both numerical codes. The case-studies are the ones of the PRENOLIN benchmark and the canonical case CAN4.

For cases of strong ground motion, soil behavior plays a major role, thus, the impact of a nonlinear soil response is examined in the finite element framework so as to evaluate the efficiency of the DRM approach treating this kind of studies. To represent the structure, the well-known case of Kashiwazaki-Kariwa nuclear power plant is used. In addition, a modification of the sedimentary basins' geometry is proposed so as to study the consequences of surface waves propagation in the efficiency of the DRM methodology. It is showed that the implemented methodology in code_aster provides a good approximation of the whole system response when the DRM approach is used.