Unstable curved hyperelastic NLFE beam

Simon Križnik

Hi,

 

simulating a curved hyperelastic NLFE cantilever beam returned unexpected results: the beam is unstable instead of holding the initial arc shape in the absence of gravity load. The issue does not happen with a linear material. What is the cause and how can it be resolved?

 

 

<?xml version="1.0" encoding="UTF-8"?>

Unable to find an attachment - read this blog

Chris Coker_21312

A couple of suggestions:

1. It appears you haven't made any adjustments to integrator settings. For such a specific problem like this, including an advanced element like NLFE (which will have a lot of high frequency content) the default settings will likely not be appropriate.

 

2. The mass and inertias are very small, which makes the solution more numerically difficult.  This usually requires (as a minimum) reducing error tolerances.  Often, a better approach is to use model units appropriate to the size/scale of your problem.  In this case, changing your mass units from kg to grams would be appropriate.  More importantly, due to the transient nature of the solution, I change the time units from seconds to msec, and this gives the desired result.

 

Note that your print interval (the values that are being output) is much smaller than your maximum step size (default of .01).  MotionSolve will forces the max step size to be equal to the print interval, but this really isn't a robust way to model in multi-body dynamics.   Usually it's a good idea to have the integrator take smaller step sizes.

 

I would highly encourage you to sign up for one of our training classes, if you have not already attended one.  A lot of these kinds of topics are covered in detail.

 

Lesson here:  The default model units and integrator settings are robust for a large class of problems.  But, if you start to look at more advanced problems, and/or problems with very small (or very large scale), you also have to consider whether or not your model units are reasonable for the problem you are attempting to solve.

 

Simon Križnik

Thanks for your suggestions.

However, changing the units to milliseconds does not solve the issue.  The displacements even after 1 microsecond are not zero and running longer duration reveals the same instability:

<?xml version="1.0" encoding="UTF-8"?>

Chris Coker_21312

The displacements are essentially zero. .0004 mm.

With performing a thorough convergence study on integrator parameters (as recommended previously) I was able to get this model to run with minimal transients for .1 seconds.   I agree that the trends are a little strange, so I've reached out to our developers to see if they have any suggestions.  I'll follow up here when I learn something new.  But it's very likely that what we are seeing here are numerical artifacts of such a small part, with small mass, inertia, stiffness, etc.

 

Note, this is a rather strange usage of NLFE.  For typically applications (belts, cables, coil springs, torsion bars), using NLFE within the context of a complete model produces resonable, reliable results.

Chris Coker_21312

Update from our developers:

The Hyperelastic material model used by MotionSolve for NLFE assumes a straight relaxed configuration.

 

We'll work on improving the software to remove this assumption.