Face of muscle inconsistent

[Lukuas31415]

In the PR #39 we add Neumann BC to two identical muscles. The direction of those BC point in opposite directions, which also means that the BC are anchored on opposite sides/faces of the muscles. To set up those BC we have the following code:

elasticity_neumann_bc_1 = [{"element": (variables.el_z-1)*variables.el_x*variables.el_y + i*variables.el_y + j, "constantVector": traction_vector, "face": "2+", "isInReferenceConfiguration": True} for i in range(variables.el_x) for j in range(variables.el_y)]

elasticity_neumann_bc_2 = [{"element": j*variables.el_x + i, "constantVector": traction_vector, "face": "2+", "isInReferenceConfiguration": True} for i in range(variables.el_x) for j in range(variables.el_y)]

In particular, we use "face": "2+" in both BC, even though one should be "face": "2+" and the other one "face": "2-", because they act on opposite sides. But if we do that, the muscles behave unexpectedly and move very inconsistently. With both BC using "face": "2+", the muscles behave exactly how we expect them to behave.

Why is that? Are our expectations wrong? Are there two errors in the setup that cancel each other out?

[carme-hp]

Thanks for opening the PR! Could you explain in more detail what happens when you use 2- ? Maybe a plot of the change in length to compare would help.

[Lukuas31415]

If muscle 2 is using "face":"2-", we get the following plot for the muscle lengths:

[Lukuas31415]

A theory about this weird behavior was the time step width, which was quite large with 1. Another try with time step width 0.1 resulted in this plot:

This means that the time step width was not the issue.

[carme-hp]

After checking in person, we agreed that the timestep is not the cause of the oscillations. The oscillations go away if a smaller k is used, e.g., k=10.

[Lukuas31415]

For k=10 the results look very similar for both cases 2+ with 2+ and 2+ with 2-.