I’ve been able to conduct cylindrical interference study using Calculix/Pardiso solver on relativaly fine mesh but always seems like I’m reaching some limits that I’m struggling to get my head around. If the interference is kept below half of the element width (I’m considering C3D20R/C3D20R or C3D20R/C3D15 with the wedge aligned to the contact normal surface) the calculation goes well and converge towards expected values. However when the mesh is refined and some mid-edge nodes starts to be in interference too, the solver doesn’t even seem to try to resolve the interference and loop through increments until it gives up. The results show some reaction forces at the contact area that would vary depending of the surface contact values but no displacement..
I’ve tried with and w/o incrementation, Nlgeom on/off, swap slave/master without success. So if someone have encountered the same behaviour I’m curious what workaround could do the trick
There is a lot going on, I deal with half and full revolution cases. The half case have symmetry BC at the symmetry plane. However the largest corrected displacement is not necessarely on one of those nodes when the case is not converging with the fine mesh but rather in a zone with ‘refined mesh’ and elements almost fully in interference. Furthermore I set up cases often reaching more than 1million spring elements..
I will try to create two simple test cases that demonstrate the problem soon
here is the zoomed out picture (the blue part is fixed at the top (not visible) for the interference solving step):
That would be best. To see how it behaves on a small representative segment, maybe even with axial symmetry for simplicity. CalculiX needs good quad/hex mesh for accurate resolution of interference fit but proper node alignment may also be necessary.
ok sorry for the delay on this topic. I’ve created a test case to illustrate the problem.
It’s made only of C3D20R elements and the model is made of a holed block into which two pins are inserted 90° from one another to create a traction load case.
I’ve played with the mesh size of the pins. In the file Shear_setup-INTER01-deactivate1thenreactivate.inp there is one pin with a mesh size slightly coarser than the hole mesh while the other pin have a slightly finer mesh compared to their respective hole. If you deactivate (or remove using *MODEL CHANGE) the contact pair of the later, the simulation starts, until you reactivate the contact with the pin with the finer mesh. At that point the simulation spins and show no residual force and quasi-nul correction displacement (0.1000E-29).
In the other case, with both pin having a coarser mesh, the simulation seems to be running fine.
Using node to surface instead of surface to surface doesn’t seem to have an effect
Please find the files at the following link : Download
