CFD Online Discussion Forums

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   Main CFD Forum (https://www.cfd-online.com/Forums/main/)
-   -   Node and element count (https://www.cfd-online.com/Forums/main/166302-node-element-count.html)

rere February 6, 2016 11:05

Node and element count
 
I have in the region of 16E6 nodes and it takes ages for Mesh Generation as well as the solution when analysing the domain around a truck.

How do i go about changing the node and element count? Would it also be ideal to chnage the element size from 0.1m to 0.5m?

Thanks

mprinkey February 6, 2016 11:58

Quote:

Originally Posted by rere (Post 583998)
I have in the region of 16E6 nodes and it takes ages for Mesh Generation as well as the solution when analysing the domain around a truck.

How do i go about changing the node and element count? Would it also be ideal to chnage the element size from 0.1m to 0.5m?

Thanks

That is a large problem to try to solve on a single system. I think you are not really understanding the details involved in meshing if you are considering a factor of five increase in cell size/node spacing. That would mean going from 16 million nodes to roughly 128,000...if your problem is 3D. Meshing is still more art than science--you need to find the happy medium between putting sufficient mesh density in areas of importance (boundary layers, recirculation zones, flame fronts) without getting a mesh so uniformly dense that you cannot get an answer in a reasonable time. There are meshing strategy guide out there if you search, but I can tell you that these are skills that are only learned (at present) by experience.

The alternative is to borrow or buy a cluster to run your job in parallel. Assuming that this is a turbulent, steady state simulation, you should be able to solve this in a reasonable time on 50-100 cores with Ansys Fluent, Star CCM+, or OpenFOAM. I have clients with clusters of this scale running problems up to 100M cells in steady state on a regular basis. If, however, you are doing LES or some other transient analysis, I think that you need to go back to the drawing board and reassess exactly what you are trying to accomplish.


All times are GMT -4. The time now is 01:01.