CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   Mapping Pressure In Workbench (http://www.cfd-online.com/Forums/cfx/24269-mapping-pressure-workbench.html)

Erich July 16, 2007 10:42

Mapping Pressure In Workbench
 
Hello,

We have noticed really long run times for the mapping of pressure from CFX result files in Workbench. Are there any methods of decreasing this map time (short of reduced element count)? Is it normal for mapping to take over 10 times longer than solution time?

Thanks,

Erich

Wooster July 16, 2007 12:14

Re: Mapping Pressure In Workbench
 
I'd say not. My pressure map time is about as quick as any other map time and I generally see 2 million nodes in my models. What about when you just launch CFX-post from the launcher? Do you still have long wait times?

Erich July 16, 2007 12:50

Re: Mapping Pressure In Workbench
 
CFX-Post when launched is very quick to load (~ 10 seconds), and the transient time steps load in around 5 seconds.

The interesting part is that during the mapping process, we see only a 25% load on the CPU's and low memory usage as well. The model is 1.8 million nodes and is running on a dual core 3.0 with 8 gb of RAM.

All of the executables are local, and the only network interaction is for the license.

Thanks for your input.


Robin July 16, 2007 14:24

Re: Mapping Pressure In Workbench
 
Hi Erich,

The delay occurs when Post searches for the nearest nodes in the mapping. Post uses a different mapping when you load an ANSYS CDB file (which is done in the background when Simulation maps CFX results). The search is fairly quick when the regions line up (i.e. same CAD surfaces), but can take a long time if there are non-overlap regions.

To speed up the mapping, specify a CFX boundary condition on the same surfaces that match up with your Simulation model, don't just map the default boundary region.

Regards, Robin

Erich July 16, 2007 17:55

Re: Mapping Pressure In Workbench
 
Hello Robin,

I hoped you might answer. The mapped faces are identical in both CFX-Mesh and Structure models. I created a surface group of the corresponding areas of interest (< 20 surfaces) and made it a bc. We cut the number of surfaces dramatically from the default wall region. The plus side was that the final mapping did show 100% mapped surfaces.

The downside of this issue is that we have a transient solution with a fair number of mappings. Any other possibilities? If we cut out too many of the surfaces the loading on the unit will no longer be accurate. Is there an advanced parameter that would close the search distance for the nearest nodes? Thanks for any thoughts.

Regards, Erich


Robin July 17, 2007 12:53

Re: Mapping Pressure In Workbench
 
Hi Erich,

The number of surfaces shoudn't matter. What matters is that the source and target regions for the mapping are the same physical surface regions. The mapping algorithm takes significantly longer when there are non-overlap regions.

Regards, Robin

Erich July 31, 2007 15:27

Update
 
Ansys support recommended switching from solid elements to solid shells (type 190) on mapped, sweepable surfaces. This resulted in our pressure mapping taking 1/27th of the time! Thanks Ansys support.

Erich


All times are GMT -4. The time now is 11:59.