Unexpected behaviour of cuboid muscle in optimize-prestretch-length branch

[Lukuas31415]

I have tested the cuboid muscle example in the optimize-prestretch-length branch, where we stretch the muscle until it reaches a certain length. There are a few unexpected things:

1. The stretching seems a bit wonky.

This might be because it has been stretched to almost double its length, but that never happened with the prestretch-force example from my bachelor thesis, and it also has been stretched by this much before. Stretching it by 10cm seems to be the smallest length increase where this happens. For smaller length increases this looks completely normal.

2. The contraction doesn't seem to work properly.

The green solid part is the muscle after prestretch and before contraction, the white wireframe is the outline of the muscle after contraction. It seems like the muscle has deformed, because it has become squished together at the end, but it has stayed in place. And it has actually contracted closer to the middle as we can see by the displacement of the wireframe on the right side of the image. The file that tracks the contraction length of the muscle says that after 40ms there is a contraction of about 0.74, which should be more noticeable in my opinion.

[carme-hp]

Can you check if this is also the case for cuboid_muscle in branch fix-muscle-length ?

[Lukuas31415]

The prestretch part in fix-muscle-length works fine, even with stretching it to double its length. In the contraction part, it also stays fixed at the ends, but that was the purpose of the isometric contraction.

[carme-hp]

I've run the case and saw that with prestretch=10, the mesh after prestretch already looks bad. But actually the problem appears for much smaller prestretch. Although, it might visually look nice, every time we get a " J=det F is negative", means something wrong. This already happens with prestretch = 2, but not with prestretch = 1.

Error: J=det F is negative: -0.0983867. Result will be unphysical.
For dynamic problems, reduce time step width, for static problems, add smaller "loadFactors" or reduce load.

@Lukuas31415 try looking at the loadFactors I used in biceps and test them here.

[carme-hp]

I have looked at the settings for the biceps, and found that the loadFactors were also not set. What was different there, was

"loadFactorGiveUpThreshold": 1e-5, # iinstead of 0.1

This works for a prestretch of 5cm but not for 10cm. There we get the error message:

Solution failed after 0 iterations, residual norm 410.384, retry with load factor 7.62939e-06. Nonlinear solver reached load factor 7.62939e-06 (no. 0), which is below give-up threshold of 1e-05. Now abort, use best found solution with residual norm 5.81986

I even tried with a factor of 1e-8 but this doesn't work either. I think the current approach doesn't allow such a big prestretch.

[Lukuas31415]

I tried it with loadFactorGiveUpThreshold 1e-5, but I still get the usual error: J=det F is negative. The maximal prestretch for which I don't get this error is 1.83. For 1.84 I already get this error again. I also tried it for different starting lengths of the muscle and it seems like it's not an absolute constant, but a relative one. The muscle can be stretched by 15.3% without problems, but anything above that seems to be too much and I get an error.

An idea of another approach to this problem: Since we don't get an error for the prestretch force case (even with large prestretches) and there seems to be a bijection between prestretch forces and prestretch lengths, we could use a bisection method to find the force that gives us the prestretch length we want and use the prestretch force case just as we know it.