The modeling method actually depends on the analysis you're conducting. If you study the behaviour of a region around a certain weld, you would probably model it with tet elements (HAZ study for example).
The same would probably apply also in case you conduct a durability analysis. However, if you're talking about a fatigue analysis with some fatigue solver for example, there are certain modeling techniques dictated by the solver.
Tets would be a way down the list of preferred strategies for modeling welds. You'll need a lot of them through the thickess of the zone of interest. Better to use bricks if possible.
In a lot of cases, you'd be better off starting with shells (at least you'll get the bending stiffness right)
Hi Everyone,
Thanks for the timely response. I am trying to perform a standard linear static analysis. The component I am trying to analyze consists of a t joint, which I modeled using shell elements. I considered modeling the fillet welds with 6-node wedge elements, however I am unsure about its accuracy.
There are a number of methods of extracting stresses for fillet welds in shell elements, a quick literature search will reveal their details (Maddox is a good one).
Generally you extract the stresses at a number of locations along a path leading up to that weld and plot them in excel (or wherever) for extrapolation)
Depending on the method (the locations differ) you select the stress at 0.5 x thickness from toe and 1.5 x thickness from the weld toe and stick a curve through them. Where this line intercepts the weld toe location you can extrapolate the stress - this removes the abrupt increase in stress leading to infinity (or meaningless results).
Other methods are Battelle's Verity (nodal force transformations), this is expensive but places less importance on mesh refinement.
