PID controller potential bug

[hwiputra]

Hi All,

Ihave a simulation of a pressure expandsion of an aorta. I applied a load controller with PID and without PID in the two codes as lc="2". This load controller is not being used iin with PID script, literally just hanging out. However, the presence of it is enough to break the code. It did not show the aorta expanding in the .xplt file. It is likely a bug for a PID controller. I am planning to use PID controller for a robin-like boundary, but it seems to not be working now.

Thank you.

Regards,
Hadi

[maas]

Hi Hadi,

I tried it with the latest version of FEBio (version 4.7) and I don't seem to have any issues. It looks like you are using version 4.6, so perhaps try to update to latest version and then give it another try. Let me know if you still find issues.

Best,

Steve

[hwiputra]

Hi Steve,

I reran the cases with febio 4.7 and found that the PID case returns very small deformation, while no PID case returns much larger deformation. The PID case didnt attach the loadcurve to any parameters. I believe the two codes need to return the same values but did not.

Thank you.

Regards,
Hadi

[maas]

Hi Hadi,

According to the log files you attached, the two problems ran pretty much identically. There is a small difference in convergence norms towards the end, but that is not entirely surprising if you are running on a computer with multiple processors (See here for a note on repeatability). What makes you say that one case has a small deformation and the other has a big deformation?

Thanks,

Steve

[hwiputra]

Hi Steve,

That is weird, I saw that too, but the xptl files gives out different output. I attach the video I got and the displacement from the two cases.

Regards,
Hadi

[maas]

Hi Hadi,

Okay, I think I finally figured it out. It turns out that it was a problem with the plotfile. By default, FEBio will write the PID values to the plotfile, but there was a bug that corrupted the plot file. As a result, although FEBio was working fine, the data was not written correctly to the plotfile. And that's why you were not seeing the correct results in the file with the PID controller.

This should be fixed now, so please update to the latest development version (see here for instructions) and then give it another try. Let me know what you find.

Best,

Steve

[hwiputra]

Hi Steve,

I tried it, it works now! Thank you.

Regards,
Hadi

[hwiputra]

Hi Steve,

The .feb file I sent in the first post is a "test" code, and today I ran the full problem with the attached feb file. It showed the same error as described previously. I ran one without a PID controller before and it ran fine.

https://drive.google.com/file/d/1nbYmnFAz4Ukkod_gAivfjJZ8jMGe95v0/view?usp=sharing

This .feb file takes quite some time to run. It seems to still have the same zero-displacement bug. The earlier code I sent you not longer have that issue.

Is it because I skipped some load controller indexing or that some of the load controller were not in use?

Im sorry for the trouble, thank you.

Regards,
Hadi

[maas]

Hi Hadi,

It turns out that this was a different issue. In FEBio Studio, there was an assumption that there are a maximum of 256 data fields defined in the plotfile. Since all the PID controllers are also added to the plotfile, that limit was exceeded for your model, and strange issues ensued. I upped the limit to 65,535 (16 bit instead of 8 bit), which hopefully will be enough. Please do a dev update and give it another try. Since the problem was with FEBio Studio you should not have to rerun the file in FEBio. Let me know what you find.

Best,

Steve

[hwiputra]

Hi Steve,

I've tested the .feb file again and it works like a charm.

Thank you.

Regards,
Hadi