CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   How important is clean geometry for CFD? (http://www.cfd-online.com/Forums/main/87036-how-important-clean-geometry-cfd.html)

Kevin De Smet April 9, 2011 19:03

How important is clean geometry for CFD?
 
I am well aware of the importance of a clean topological model when it comes to performing FEA analyses, because the mesher really has a tough time with tiny faces and short edges and other such guck.

Since CFD has more of a grid approach to discretization, does it matter as much if your geometry is clean or not?

Does it also 'mesh' the model similar to finite elements or are finite difference/volume methods working entirely different? What about a solution that takes into account heat conduction in solids as well, what about then?

cfd_newbie April 10, 2011 02:48

Quote:

Originally Posted by Kevin De Smet (Post 302913)
I am well aware of the importance of a clean topological model when it comes to performing FEA analyses, because the mesher really has a tough time with tiny faces and short edges and other such guck.

Since CFD has more of a grid approach to discretization, does it matter as much if your geometry is clean or not?

Does it also 'mesh' the model similar to finite elements or are finite difference/volume methods working entirely different? What about a solution that takes into account heat conduction in solids as well, what about then?

Yes it does matter a lot in CFD weather your geometry is clean or not. It wil have big problems since you create a very detailed volume mesh and if your geometry has some bad feature(small extrusions holes, non water tight geometries) you will not only have a tough time meshing, but also solving the simulations.

minger April 10, 2011 10:48

In my experience having a clean geometry is crucial. Without a clean geometry it is very difficult to obtain a quality mesh without some of the bad things that were mentioned in the post above.

Having those "bad things" can lead to divergence in your simulations, lack of proper convergence or poor results in those areas.

DBFIU April 11, 2011 20:05

Bad geometry is a recipe for disaster in every way.

If you have bad geometry, chances are something got messed up somewhere and no one wants to take the time to fix it. It is frustrating, but it has to be done or else you'll shoot yourself trying to debug your mesh and solutions.

The question is, why WOULD you begin any type of simulation with corrupt geometry? If it's a time thing, then just be patient and fix it... you'll be better off in the end.

Kevin De Smet April 12, 2011 14:28

Glad to hear it, I'm a proponent of clean models myself. Not only for the repercussions a dirty model can have - just seeing nasty topology makes me just go crazy and I want to fix it! Whether it would make any deal or not.

pritish April 13, 2011 11:48

Quote:

Originally Posted by Kevin De Smet (Post 303317)
Glad to hear it, I'm a proponent of clean models myself. Not only for the repercussions a dirty model can have - just seeing nasty topology makes me just go crazy and I want to fix it! Whether it would make any deal or not.

i can tell you...it takes double the time or more to debug a volume mesh....so have it cleaned before that ! ! !

Pauli April 13, 2011 11:52

Quote:

Originally Posted by DBFIU (Post 303182)
... why WOULD you begin any type of simulation with corrupt geometry? ...

Because it is the only thing available! Which leaves you the analyst in the position where you can clean it, wrap it, or recreate it. What works best is wholly problem dependent. Operator experience can pay serious dividends.

The other option is to throw your hands up & complain the CAD data is unusable. Since your customer likely won't want to recreate the CAD data, you run the risk of losing business to someone who will work through bad CAD data.

DBFIU April 13, 2011 21:44

Quote:

Originally Posted by Pauli (Post 303451)
Because it is the only thing available! Which leaves you the analyst in the position where you can clean it, wrap it, or recreate it. What works best is wholly problem dependent. Operator experience can pay serious dividends.

The other option is to throw your hands up & complain the CAD data is unusable. Since your customer likely won't want to recreate the CAD data, you run the risk of losing business to someone who will work through bad CAD data.

Every analyst should have medium to high experience with CAD for this very reason. No reason to throw your hands up if the geometry is corrupt, just correct it and move on. :cool:


All times are GMT -4. The time now is 12:32.