unrealistic velocity field
 

Here is the situation, I am running an external unsteady incompressible simulation, flow over a bluff body (building).

However, by clustering the 3D mesh to the surface in resolving the wall layer, the mesh seems to be inevitably bad with high aspect ratio when they are extended to the far field.

So you can see as attached, by using dyn LES model, the velocity field wriggles greatly at the leading top regions of the building.


Any ideas how to remove this? Any suggestions?

Attachment 10773
Attachment 10774
Attachment 10775

This is not my area of expertise, but...

Your problem may be with time accuracy, so by reducing the time step the problem may go away. Or, given a physical time step, converge it more for that time step. Don't know the ins and outs of how, or if, an "inner iterative loop" works with OpenFOAM. The "inner iterative loop" is, for me, a concept from the RANS world. Not sure if that carries over to LES.

Or, you can go with overset grids. But, I think you would have to modify OpenFOAM to handle the transfer of information at the fringe points, etc.

Or, you can redo your grid so it is not cartesian like, i.e. make it truly unstructured. But I guess there is a reason you don't want to do that. Again, I'm not versed in the execution aspects of LES.

There is a "nesting refinement" in GridPro which does what you want. Unfortunately there is nothing about this on the homepage. Just send them an email asking for it.

Indeed, the nest function in GridPro is the only way I see to keep small aspect ratios while still refining towards the surface. A first step could also be to use few internal surfaces to keep points close to the surface.

But the mesh of Daniel doesn't look like a GridPro mesh. You used blockMesh?

... and a possible mesh topology for Daniels case might look like the attachments, without refined cells in the farfield.

Thank you guys, I use blockMesh to building that 3D building mesh. Hmm, around 4 hours.

I am still not skilled in GridPro, I tried to use internal surfaces before, but I will look into this nested refinement function too.

Vincent, if it were you, how many hours are you going to take to make this mesh?

High aspect ratio cells is a disaster to LES. JBeilke's figures are all that what I need. Thanks

If you include some nice nest refinement well placed, I would say around 30min. Up to one hour maybe.

And you will get something you will never have with blockMesh even spending one week on the case ;)

JBeilke's figures show a starting point and from that you can do a lot of nice other things. :D

Hi,老魏。

I also found the problem that unrealistic velocity field appeared in front of the cylinder as similar in your case. Have you solved the problem now?

Thank you
Zhou Qiang