model converge in febio2.5.0 but does not converge in febio2.8.5

[SChen]

Hi,

I have this simple tryout model that converge in febio2.5.0 but doesn't converge in febio2.8.5 (after changing fibers-3d-gkt-uncoupled to fibers-3d-gkt). Can I get some help to spot the issue?

I have been using febio2.4 for a while. Recently I decide to switch to a newer version since febio2.4 does not support tri6 elements. I am trying to decide which version to install on my work station, so I have been running simple tryout models using both febio2.5.0 and febio2.8.5.

Thank you!
FiberSetupTryout4_N1.feb

[ateshian]

Hi Sheng,

It is possible that febio2.5.0 had a bug where the nested fiber angles within a solid mixture were not being concatenated properly. (We would have to go back and review the release notes for 2.5.1 forward to double check, but I recall that this was fixed at some point.)

To run your model successfully with 2.8.5 I had to increase the number of integration points in the GKT scheme to <nph>11</nph> and <nth>31</nth>.

Best,

Gerard

p.s. Since you are using an uncoupled formulation I assume that you want a nearly incompressible response. In that case I recommend that you increase the value of the bulk modulus <k> to be at least 100 times the stiffest modulus in your mixture, i.e., k=48e2 based on the stiffness ksi of each of the the two fiber families.

[SChen]

Hi Gerard,

Thanks for the reply!

Does it mean febio2.5.0 was calculating it wrong? I did some tryouts to ensure the fiber information being considered properly. The y-stress distribution in this two fiber model is symmetrical. When I use one fiber, the y-stress distribution is not symmetrical. From these results I think the two fibers angles are considered properly in the results. Does “nested fiber angles within a solid mixture were not being concatenated properly” mean the fiber angle information were not considered correctly?

After I change the <nph>11</nph> to <nth>31</nph>, the model can converge in febio2.8.5. However I was not able to read the results from Postview. Prompt from below showed up. I wonder if you can reproduce this problem.
11.PNG

As to <k>, I wonder if each fiber family needs its own <k> values within the <solid> section or one <k> in the isotropic material (Mooney-Rivlin in this case) section is enough for the whole solid mixture.

If one <k> values within the Mooney-Rivlin <solid> section is enough for the whole uncoupled solid mixture, do you mean k=ksi * number of fiber family *100? Number of fiber families has to be multiplied, too?

Thank you!
Best,
Sheng

[ateshian]

Hi Sheng,

Actually I reviewed the release notes and didn't see a bug about the concatenation of material axes orientations, so perhaps I remembered it wrong. I also downloaded febio 2.5.0 and ran your model and the results look fine. In contrast, using 2.8.5, I get weird (asymmetric) results. So give me a little time to look into it in detail, I'll get back to you as soon as I figure it out completely.

Regarding <k>, the uncoupled solid mixture takes the sum of <k> values for each <solid> in the mixture. So it doesn't really matter which <k> you use, as long as you include at least one of them and pick its value to be ~100 times greater than the some of largest moduli in the various <solid> materials.

Best,

Gerard

p.s. Regarding the issue with PostView, I am not sure why that's happening. Perhaps you can try using the version of PostView that was released soon after 2.8.5 was released?

[ateshian]

Hi Sheng,

I've confirmed that there is a bug in "continuous fiber distribution uncoupled" in febio 2.8.5, which was incorrectly accounting for the fiber orientations. This bug has now been fixed in febio 2.9.0, which will be released in the near future. A possible workaround for now is to use "continuous fiber distribution".

Thanks for bringing this to our attention.

Best,

Gerard

[SChen]

Hi Gerard,

Thanks for the clarification!

Best,
Sheng