Convergence issues with contact

[GerryJDail]

I am analyzing a planar symmetric model of a bearing housing with preload on the cap bolts and tie-down bolts.  The model is very similar to one that I solved successfully last summer for a client.  I have checked the model and as far as I am able to see, the models are set up identically for contact, offset contact to mimic the bolt preload and forces applied to the tie-down bolts to mimic preload of those components.  Normally, after establishing the preload on the cap bolts by adjusting the offset, I can then load the bearing and obtain stresses for the housing.  However, this model is not converging and seems to be dithering around the first load step and even subdividing the load step while still dithering around.  I am using FEMAP as the pre/post processor.  I am providing a link that will allow for anyone interested in assisting me to download the .ini file, the .nas file and the FEMAP file for review.  Assistance is most appreciated.

Cordially,

Gerry J. Dail, PE

https://spaces.hightail.com/space/kbyjma1cut

[jdalidd]

Hi Gerry,

I took a look at your model and tried running the .NAS file. During the NL solve the displacements became very large indicating the model is either not constrained properly, or more likely, the contact isn't setup sufficiently. Doing a quick look through the contact I already see one issue (see image below). 

I recommend fixing that contact segment and then doing a detailed audit of all the contact segments to make sure everything is setup correctly and no contact is missing. If the model still diverges, change all contact to welds and ensure the model runs well. If it does, slowly convert the welds back to separation contact until you find the unstable area. You can also look at the GRID number with max displacement in the .LOG file during a run and that would indicate one of the parts that is probably unstable (if the displacements are large). 

[GerryJDail]

Thanks!  I will check things out again and make corrections.  However, I remeshed the entire model and fixed the contact surfaces such as you noted.  I then re-ran the model and found that if I changed the offset contact surfaces that are designed to create a preload on the bolts holding the cap of the housing to the base from symmetric contact to unsymmetric contact, I was able to obtain results that appear to be valid.  However, if I then changed the input deck only by creating a subcase to solve for the bolt preload, the model would not converge.  This is very confusing.  I am continuing to work to debug this, but something appears to have changed with the solvers from what I used last summer on very similar problems.

Gerry