CCX Exiting without Results

protontim

I am running a non-linear simulation using CCX "contact" for a frictionless interface. I am using v10 beta. The simulation sets up and starts to run just fine, but after 1 step it exits without showing any data in the solution. The CCX window says nothing about CCX running to completion. Any idea what might be happening? It is a large model but I have 64GB of ram and it appears to have more than sufficient margin. My results are shown below.

Victor

That's one of the usual ways CCX exits when the model is too big. I don't think it can use all of 64 GB of RAM unless you compile it specially. Did you see how much it did use? It might be limited to 16 or 32 GB. 900,000 nodes could easily exceed one or both of those limits.

In case the mesher is generating unnecessarily fine mesh near corners and edges, the new Minimum element size setting may help to dramatically reduce the size.

protontim

Thank you Victor. I was wondering if that might be the case. How low would you suggest I try and get the node count?

protontim

Is there a way to get something similar to the CCX contact type of interface with the internal solver? I looked at the compression only constraint, but that only took one surface as an argument so I wasn't sure how to use that in a mated surface condition. Is there some documentation I could look at to see how that would be done?

Victor

I've found the limit is typically around 300,000 nodes with 16 GB of RAM. I expect you would be able to go higher though. You can delete half the mesh and solve to see if it succeeds.

The closest things to contact for the internal solver are:

• Compression only support. But one surface is ground so it can't connect two parts.
  
• Tension-only elements. You could use these for simple contacts with small compatible meshes by having an initial penetration so that the elements are in tension when the surfaces are pushed together.
  

protontim

Looks like 300K is the magic number, tried the high 300's with no luck. Dropped it below 300K and it appears to be iterating. Looks like CCX only uses about 5GB max for its computations. Curious as to why it is so limited in its model size. Is there a version that can handle more? A 64bit version perhaps?

Victor

I don't know how but I've heard of people solving much larger models. Perhaps check on the CalculiX Yahoo! group.