Workbench Mechanical - contact not working 2

[Jamdat]

Hello all,
I am working on simulating a fixture that will hold a ring for machining. Using Ansys Workbench, I've simplified my model down to 2D geometry made in design modeler, but I'm struggling to get contact working. The solution (picture attached) fails to converge and gives messages about underconstrained rigid body motion, contact status experiencing an abrupt change, and pivot warnings. The two objects that are supposed to be contacting pass right through like they weren't even there. I've read so many threads and tried a bunch of options, but nothing works.

Contacts are [ul]
[li]Auto generated[/li]
[li]Frictionless[/li]
[li]Symmetric[/li]
[li]Augmented Lagrange[/li]
[li]Stiffness Factor = 1 (update each iteration)[/li]
[li]Adjust to touch[/li]
[/ul]The initial information under the 'Contact Tool' indicates all are closed.

Any clues as to what I'm doing wrong? This seems like such a simple thing to do.

Many thanks in advance.

 

[lamboram]

Hello,

it is a boundary condition + mesh issue. Suppress the three smaller circles inside, and use the point of contacts to define your BCs. So the vertices on the top (two) will become simply supported or no friction support. The vertice on the contact between the circle with the force will have equivalent force load. Simulate it, it works. make sure your mesh is quite good. I suggest Hex, your mesh on such surface simulations will tend to have Hour-glass effect (google it). This effect will give unrealistic values. I simulated your case with an approximate geomentry. It works.

Cheers,
Ram

 

[lamboram]

If you still would like to simulate with those circles, use bonded contact. it will work the same way. but the previous method without cicles you might get the same results a bit faster, because its with lesser elements/computational efforts. Signs of an efficient Computational engineer

Cheers,
Ram

 

[Jamdat]

Sorry about the delay in replying, but thank you for your advice! I was able to get the simulation working well enough to be useful. I ended up using bonded contact with the circles. I'm going to experiment with contact over the next little while to try and become more proficient.
Two questions. You mention 'simply supported' supports. In workbench I can only get 'Frictionless Support' and it only works on lines (in 2D). Is this what you're talking about? Didn't work for me so I used the bonded contact.
Second, I'm trying to get 'Hex' mesh, but can only specify this when using a 3D model. When selecting 2D in workbench static structural, only triangles method is available. Did you model as 3D with thin thickness?

Thank you again!

 

[assafwei]

Hi,

Regarding your first question - what is the force applied? How many sub steps did you give for the analysis?
The setup you showed in the image should work if you keep in mind that:

If the force is too big, the small circle will fly away, because in the first step before contact is established it already flew away. In order to prevent it you have two options:

1. divide the problem to very small substeps, this way the force will be very small and the chance of the contact to engage is greater.
2. A better solution is to run the analysis in two steps, in the first step apply a very small force with small substeps, this will engage the contact. Then apply the rest of the force and now you can probably use
bigger time steps because the contact should already be engaged. You can also try and use displacement in the first step to engage contact.

Regarding Hex - in 2D hex is quads, in the image you attached it seems as if you have quads. If you want to have a better chance for convergence and for good contact, you should keep the mesh size in the contacting bodies the same size and play around with the mesh size until you get good contact convergence.

Hope this helped.

 

[lamboram]

Hi,

I used a 2D model with thickness.

answers to your questions;

1. I am not sure why you are not able to get 'simply supported BC', but you should be able to find it in Workbench Mechanical. some other similar BCs are Fixed support and frictionless, but frictionless cannot be used on point boundries.
2. Yes, as assafwei mentioned, Hex in 2D becomes quad mesh. triangles are actually much better these days compared to what they used to be, thanks to the improved algorithm. But, I would still use quad/hex, it saves element count and it is reliable.

Cheers,
Ram