Bouncing Sphere

Famous_Mortimer

Having some trouble with this one. Can't get the sphere to rebound. The model falls apart when the sphere contacts the plate. Without the contact, the ball falls thru the plate. Not sure where I am messing up. Any help would be appreciated.

Famous_Mortimer

Here is the file.

Famous_Mortimer

I hate to admit defeat.

Victor

I got it to bounce off the surface before it fails with these changes. Not sure if all of them are necessary:

• Removed back surface of plate from contact.
  
• Changed plate to quadratic elements.
  
• Refined plate.
  
• Turned on Automatic time stepping in Analysis
  settings.
  

Sergio

I did some comparations between Abaqus and CCX using exactly the same cards, and the results were different after the impact. CCX bounds more than Abaqus, didn't find a way to change this behavieur even using different damping options on CCX.

Victor

Does the CCX one rebound above its starting height, violating conservation of energy? I got a little bit of that with an orthotropic material but mostly it won't solve that far and I lost the special material that worked!

One source of a difference might be that Abaqus shells (I assume) don't have the normal stiffness that CCX shells have.

Sergio

Have checked and not, it doesn't go up higher than initial height. Plotting the position of one node is not accurate as the ball start to rotate after a few rebounds. Atacched is the model that I have run in CCX (using Scite to launch the solver) and a colleague in Abaqus as well (running a job directly from the .inp file) without any modification. The model is made all of second order hexa elements, no shell involved.

Is there any standard or benchmark case that we could use to validate the impact simulations results???

Victor

Oh, good point about the rotation. That itself is suspicious though because of symmetry and it might be taking energy away from the linear motion.

Normally, I'd say it's not reasonable to expect the same results from the same inputs with different programs but hex20 elements tend to be implemented the same way everywhere. There might still be other implementation differences that would vanish with mesh refinement like maybe lumped vs consistent mass matrix or differences in the contact formulation.

Once Guido pointed out a bug in dynamics that seemed to be related to this kind of problem but he fixed it several versions ago.

Can you show the Abaqus solution? How different is it?

JohnM

We will bounce a ball to get the damping characteristics of the material. Connecting the peaks forms a curve that allows you to calculate damping from fitting a logarithmic decrement.

Sergio

In the past had used a pendulum test as a way to determine the rebound properties of rubber, others ask you for a drop test as you perform, but inside a plastic tube to avoid wind effect. But for more accurate damping measurements we use dynamic test on a servohidraulic dynamometer (MTS).

disla

Some fun,

disla

Even more fun with Neo Hookean bouncing ball

JohnM

Plot your Z displacement on the top (non-impact) side of the sphere, I think you catch a higher order flexing mode of the sphere

disla

Hi JhonM,

Thanks for comment.
I'm still working on this model . My goal during last weeks has been to get a model in which I can validate the Strain energy density for hyperleastic material models. I think this could be finally the right direction.
I suspected it was the first mode because of the shape, not the frequency. A fast check seems right. Modal gives 78 Hz and a very raw extraction of the oscillation at the opsite contact point you comment gives 77 Hz.

JohnM

If you are designing a rubber collar for a rotating seal, those are the kind of things to watch for. If you are designing a rubber ball, I think you're good

disla

. Thanks