Fillet weld static strength results accuracy

[rileino2001]

I am studying static strength of fillet weld according to eurocode, (1993-1-8) which suggests studying stresses at the throat thickness mid-line (a-dimension). I have a few options that came to my mind but neither one seems quite accurate. I have bilinear material, 3D geometry and nonlinear analysis. stresses are far beyond yield but below UTS. Nodal forces = ON. Tet mesh with 6 elements through-thickness overall. 

I came up with these:

1: splitting geometry in half and using bonded contact to capture reaction forces

2: splitting geometry in half and using shared topology to connect pieces and capture force on the splitting surface

3: a single fillet, using construction surface to capture stress components

4: a single fillet, using construction surface to caputre force components with force probe.

 

Here's the construction surface for options 3 and 4. My issue is that with option 4 the extraction options "nodes on surface", "nodes on positive side" and "nodes on negative side" give remarkably different results and it seems quite unaccurate.

My concern with option 1 is that nonlinearity is not captured accurately enough due to material plasticization and bonded contact gives misleading results. share topology seems somewhat reasonable but I am still not sure about accuracy. Option 3 also seems quite robust and gives a clear stress mapping instead of vectors as in option 4, but still not sure about the accuracy.

 

So, I would appreciate some help and shared experience. My goal is to get as accurate results as possible. Thank you in advance :)