Learning Forum

[meganhigley11]

I was wondering if there was a good guideline approximation for the amount of RAM needed to generate a mesh. We were thinking that it was approximately 2 Gb of RAM per 1 million elements. Can anyone confirm? Our model should be around 72 millions elements.

[Aniket Chavan]

mhigley It highly dependable on the number of factors such as type of elements, mesh methods.

But in general, if you are trying to create 72M elements on a 144+ GB machine as per your description, it is highly likely that it will be sufficient. For example, I have created 150M+ element tet mesh on 64 GB machine in the past. Again highly dependent on the methods and elements and others factors.

 

[meganhigley11]

Aniket, we edge size all the edges, and face mesh all the faces to create our mesh. Most of the bodies are mappable and sweepable. This is what part of our mesh looks like. We are trying for rectangular prism elements on the 3D model. What's taking a super long time and using a lot of RAM are those boxes with the circles in them. 

[Rob]

It's more likely some of the size functions or mesh connectivity that's slowing things up. Why do you have each tube in a separate fluid region? Pave+sweep should be more efficient. 

[meganhigley11]

Where do I find pave and sweep? I would love to try it.

[Rob]

Have a look in the documentation and in the tutorials. The best way to use Workbench (ANSYS) Meshing is to have relatively few volumes and let the mesher numerics do all the hard work. You do need some distinct volumes as you need to minimise cell count in the student version (you're capped at 512k cells) but the above decomposition is reaching levels we'd have used 25 years ago for a multiblock mesh!