Enhancement of the simulation of scaled vessel failure experiments by a recursive coupling of the thermal and mechanical FEM-models

For the simulation of experiments investigating the behavior of the lower head of a nuclear power station in case of a core meltdown scenario like FOREVER (performed at the Royal Institute of Technology, Stockholm) it is necessary to model the melt pool convection and the temperature field within the vessel as well as creep and plasticity processes. Therefore a 2D Finite Element Model with 3 different physics environments is developed based on the code ANSYS® Multiphysics.
A thermal environment was build up including planar and contact elements for conductive heat transfer, additional surface elements to simulate convection and radiation from outer surface areas and a radiation matrix to account for internal radiative heat exchange. Normally a CFD-simulation would have been required for the natural convective heat transfer in the melt pool, but at very high internal Rayleigh numbers no turbulence model is capable for a correct simulation. Therefore an Effective Conductivity Convectivity Model (ECCM) was developed to simulate the heat transfer from the melt pool to its environment.
The resulting temperature field of the vessel wall is applied to the mechanical model. To describe the visco-plastic deformation a numerical creep data base (CDB) is developed where the creep strain rate is evaluated in dependence on the current total strain, temperature and equivalent stress. In this way the use of a single creep law, which employs constants derived from the data for a limited stress and temperature range, is avoided. For an evaluation of the failure times a damage model according to an approach of Lemaitre is applied.
The third physics environment is a kind of fictitious physics environment: it uses hyperelasticity and contact to move the melt pool along with the creeping vessel wall.
In this paper problems on the numerical side are explained and differences between the results of a simple coupled and a kinematically coupled FE-simulation are highlighted. The final comparison with the experiments shows that the kinematically coupled model is closer to reality than the single step model.