XFEM SIF calculation for inclined crack

[Marina Grigorovitch]

Hi all

may be someone can help ...

I have a thin 3D plate of 60*60 cm (single element depth) and an inclined not propagating crack in the middle. I tested different crack length of 1-3 cm. Mesh size of 0.45 cm. element type C3D8R.
The load is uniaxial tension in y-axis.
I am using plain strain analysis, so the plate is fixed in z direction.

The problem:
The results are not consistent with analytic solutions. In addition, Kiii is pretty large. However, in my understanding, in this case SIF mode III must be zero.

I tried to refine the mesh, but it cause negative values of Kii.

It worked pretty good with horizontal not inclined crack

Thanks a lot in advance

 

[FEA way]

You are using C3D8R elements so it's not actual plane strain analysis (for that there are separate 2D element types) but rather its approximation in 3D.

If the crack is stationary then maybe you could use the conventional FEM approach instead of XFEM for this study.

 

[Marina Grigorovitch]

Hi FEA way
thank you for your reply.
My task is eventually to analyze crack propagation with my own model , and this one is analyzed for comparison. For a start I wanted to analyze simple case of stationary crack to validate the models. This is the reason and a precondition for choosing XFEM.
Nevertheless, XFEM should work for such a simple case. What kind of elements would you suggest to use ? A non reduced C3D8 also tested and it provides pretty similar results.

thanks again

 

[FEA way]

When XFEM is used for contour integral evaluation it's not necessary to prepare the conforming mesh around the crack tip. However, proper mesh refinement around the crack front is still needed for good results. Maybe that's the problem here - try further partitioning the model and refining the mesh around the crack.

Another thing is that values from the first contours are usually ignored because they are impacted by numerical errors. Values from the remaining contours can be averaged to get the desired result.