When analyzing a given weld, one is not required to remain faithful to a single model. Clearly, unless one wishes to analyze the interior of the weld pool, it is sufficient for all analyzes outside of the weld pool to specify the geometry of the weld pool as a trimmed surface patch preferably in the solid. Then for each point in spacetime on this surface patch specify the temperature, composition, phase fractions and tractions. All coupling in the welded structure with the welding process can be controlled by this interface.
We prefer the interface to be in the solid in order to leave all issues related to the mushy zone in the weld pool model. In most cases, the interface itself could be represented by four cubic triangular patches or FEM elements. This would need 25 nodes or points and usually would be more accurate than the data available to characterize the welding process.
The weld pool itself could be represented in most cases by four cubic tetrahedrons. This has the advantage that the surface is the boundary of this weld pool and all data is defined every where in the weld pool and not just on the surface of the weld pool. It could be advantageous to add a layer of cubic bricks in the solid to provide an overlap with the far field domain.
We recommend that whatever model is used to define the weld pool data, the Second Generation prescribed temperature model be used for all analyzes outside of the weld pool. This strategy simplifies the code and increases computational efficiency.