For a cantilever TWB, U cross section, 10 Hz excitation, I tried using general contact in Abaqus/Explicit (edge to edge contact, normal interaction: hard contact; tangential: rough), modeling with shell elements for 3 seconds of simulation time. It works, but here is the point: Is it a realistic simulation? The crack, located at a flange, is regarded as a very sharp V-notch: with mouth of 0.001 mm (length of the beam: 2000 mm; flange: 100 mm; crack depth: 50 mm).
I would like some opinions. Thank you in advance.
Is the crack at begin already there? if yes, is it modeled? if yes, then i think your simulation shall be realistic. If the crack is dynamic developed, are you try with damage model?
The crack is already there at the begin of the simulation. It is considered that the crack does not propagate, so its shape do not change through time. I had my doubts since it seems to be a strong idealization of a very complicated phenomenon. Thank you very much for your answer.
If you want to see whether and how the crack propagates, then you may try the extended finite element method (XFEM) in ABAQUS. But then two problems are to be solved: First, use 3-dimensional elements instead of shell elements. This might be computationally more expensive. Second, try to transfer the inertia forces of your dynamic simulation into an equivalent static load. If you can do this, then XFEM shows the crack path, if the crack propagates at all.
if you want to have more realistic solution, try 3D elements and model the crack tip shape.
For the moment, my research focuses on the global dynamics of the structure, that is, the alterations caused by breathing effect, without considering propagation. The next step that I take is to study the crack propagation. I've seen some simulations using XFEM and find them very interesting, so I will take into account your advice. Thank you very much.
I replaced the sharp V-notch by a seam (patition, with duplicate nodes). Time history response is similar, but not the same. I think this should be a more realistic model. General edge-to-edge contact works very well! The calculation takes two days and a half for 3 seconds of simulation time (750 S4R elements, Dynamic Explicit Procedure, Intel Core 2 duo 2.35 GHz CPU, 2Mb RAM).
I'm glad because my one-dimensional model has a very good agreement with these ABAQUS' results! Thank you all for your comments.
- Badges
- Science topic