convergence problems for natrual shaped cartilage

[yantaaa]

Hi,

After doing models with simple geometries in FEBio, I moved on to whole joint models. However, I got convergence problems when the model is complicated.

My model is a human hip joint with upper half of the femur bone, the pelvis bone, and the cartilage in between. The cartilage which is modelled as biphasic material and the bone were tied together. As the femur and pelvis models are irregular in geometry, I meshed them in I-DEAS using tet elements. However, I always got the message: Negative jacobian was detected at element.. at gauss point..
If I remove the pelvis or the femur bone, this error message still appears.
To pinpoint the problem, I simplified the model to just the pelvis cartialge and the femur cartilage in contact with the femoral head underneath. The outer surface of the pelvis cartilage was fixed in all degrees of freedom.

If I mesh the spherical femoral head with tet elements, negative jacobian message still appears.

If I mesh the spherical femoral head with hex elements, negative jacobian message disappears but the model still cannot get converged.

If I move out the femoral head (only cartilage to cartialge), the model can get converged.

I guess the negative jacobial error message is caused by the bad shaped tet element but am not sure.

As for the model with hex element, I tried to increase or decrease the penalty value in contact control but the model still cannot get converged. I do not know why it is not easy for the model to get converged if there is tied contact in the model. Is there any special technique which can solve this problem?

I think in the end, I will need to model the whole joint with bones meshed using tex element. How to eliminate the negative jacobian message?

Any help will be greatly appreciated. Thanks in advance.

As the size of the file exceeds the limitation of this forum, I have sent it to Steve's email (three .feb files. one is cartilage to cartialge model; one is cartilage to cartilage with femoral head of tet element; one is cartilage to cartialge with femoral head of hex element). Please have a look, thanks a lot.

Regards,
Jerry

[yantaaa]

At first, I thought the convergence problem and the negative jacobian message were caused by the irregular geometry of the model causing mesh elements of low qualities.

However, these same problems still exist if the hip joint is simplified as a ball and cup contact model. I have sent it to Steve's email. Please have a look. Thanks.

Just like the problems with the model with natural geometry, for the simplified ball-cup model, if the two parts in the tied interface are composed of hex and tet element respectively, negative jacobian message appears. If the two parts in the tied interface are only meshed with hex elements, negative jacobian message disappears but still, the model cannot get converged.

I tried to increase or decrease value of the penelty and pressure penelty, but it is not helpful.

Any advice will be appreciated. Thanks.

Regards,
Jerry

[maas]

Hi Jerry,

The convergence problems are caused by the fact that you did not define a value for the convergence tolerance in your tied interface. Adding the following line should solve this.

<tolerance>0.01</tolerance>

As far as the negative jacobians are concerned, try reducing your penalty parameter for your tied interface and see if that helps. Let me know how it goes.

Cheers,

Steve.

[yantaaa]

Hi, Steve
Thanks for your reply.
As for the hex element model, convergence problems no longer exist after adding the tolerance value.

However, for the tet element model, reducing panalty parameter can increase the no. of increment but still, the model cannot get converged because of negative jacobian.

I tried to reduced the penalty value of tied interface from 1000 to 10, 1, 0.1, 0.01, 0.001 and the model still cannot get converged.

Although it did help when I reduced the penalty to 0, the tied surfaces were not actually tied together.

I met this problem even for the model only with tied contact.

Any help will be appreciated.

Thanks.

Regards,
Jerry

[maas]

Hi Jerry,

Sorry I haven't had a chance to look at this. Were you able to make some progress? Let us know if you still need help with this.

Cheers,

Steve.

[yantaaa]

Thank you Steve,

This problem has not been solved yet. The tied interface between different kinds of elements still causes convergence trouble.

Please could you look at the model I sent to your email address if it is convenient for you.

Thanks.

Cheers,

Jerry