Mesh Adaption Across A shock

Cbas · August 21, 2010, 05:32

Hi all
I'm using a normalized, density gradient mesh adaptive scheme(coarsen and refine) in a shock wave simulation. The problem is, as the shock diffracts around a corner a sound wave is formed. Without going into too much detail, the density gradient across the shock near the wall is not very large, and thus it does not refine in this area- the area I need the most refinement. Any suggestions?
P.s. I may not understand the adaption method correctly, in which case my question is simply, why is it not refining adapting near the wall?

ganesh · August 21, 2010, 15:37

Dear Sebastian,

There are possible solutions for this issue, depending on whether you are looking specifically at this problem (and related class of problems) or mesh adaptation at large. To start with you could increase the percentage of refinement or adjust the threshold for refinement. If the refinement doesn't happen, it is simply because the sensor that you employ (in this case the density gradient) is not "large" enough (in terrms of the threshold) to signal refinement. Another possibility is to use other indicators (such as say the divergence of velocity) or a combination of more than one indicators. Or more simply, if you are focussed only on this problem (such as to understand flow physics) and adaptation is only means for a higher resolution, then you could simply geometrically refine the specific area for this case.

Hope this helps.

Regards,

Ganesh

Cbas · August 22, 2010, 10:43

That was exactly what i needed! I was thinking that i could increase the refinement threshold, but was worrying that it would just increase the refinement on the far field flow, and I would create a very 'expensive' model. I like your suggestion of more than one sensor- although i might run into the same problem. Changing the geometry - such as using inflation?

ganesh · August 22, 2010, 12:37

Dear Cbas,

Well, refinement but using threshold is always heuristic, you may have to tweak and play a bit there; there is no "universal" threshold, though people make claims of problem-independent thresholds that work well. Of course, multiple sensors may not be a complete solution to the problem; all I say is that it might be a partial solution. What I meant by geometric refinement, was to refine the area that you need by using a "geometric" sensor ie. mark that area for refinement always by using geometric constraints because you know that it needs refinement but your indicator won't flag it. That will sort the problem, but only this specific problem; it is neither intended to be a general solution nor is it one.

Regards,

Ganesh

Cbas · August 22, 2010, 15:16

Ganesh
Thank you! I've done that, and running the simulation now. Lets hold thumbs.

August 21, 2010, 05:32	Mesh Adaption Across A shock	#1
Cbas New Member Sebastian Bode Join Date: Aug 2010 Posts: 4 Rep Power: 16	Hi all I'm using a normalized, density gradient mesh adaptive scheme(coarsen and refine) in a shock wave simulation. The problem is, as the shock diffracts around a corner a sound wave is formed. Without going into too much detail, the density gradient across the shock near the wall is not very large, and thus it does not refine in this area- the area I need the most refinement. Any suggestions? P.s. I may not understand the adaption method correctly, in which case my question is simply, why is it not refining adapting near the wall?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
engrid -> save as .stl with boundarie codes	Zymon	enGrid	31	August 29, 2011 14:40
[mesh manipulation] TranformPoints gives skewed mesh Possible Bug	andersking	OpenFOAM Meshing & Mesh Conversion	3	March 25, 2008 22:33
Convergence moving mesh	lr103476	OpenFOAM Running, Solving & CFD	30	November 19, 2007 15:09
mesh adaption	John	CFX	2	December 1, 2005 01:46
Mesh for 3 dim Geometry	Phil	FLUENT	9	July 12, 2000 05:39

August 21, 2010, 15:37		#2
ganesh Member ganesh Join Date: Mar 2009 Posts: 40 Rep Power: 17	Dear Sebastian, There are possible solutions for this issue, depending on whether you are looking specifically at this problem (and related class of problems) or mesh adaptation at large. To start with you could increase the percentage of refinement or adjust the threshold for refinement. If the refinement doesn't happen, it is simply because the sensor that you employ (in this case the density gradient) is not "large" enough (in terrms of the threshold) to signal refinement. Another possibility is to use other indicators (such as say the divergence of velocity) or a combination of more than one indicators. Or more simply, if you are focussed only on this problem (such as to understand flow physics) and adaptation is only means for a higher resolution, then you could simply geometrically refine the specific area for this case. Hope this helps. Regards, Ganesh

August 22, 2010, 10:43		#3
Cbas New Member Sebastian Bode Join Date: Aug 2010 Posts: 4 Rep Power: 16	That was exactly what i needed! I was thinking that i could increase the refinement threshold, but was worrying that it would just increase the refinement on the far field flow, and I would create a very 'expensive' model. I like your suggestion of more than one sensor- although i might run into the same problem. Changing the geometry - such as using inflation?

August 22, 2010, 12:37		#4
ganesh Member ganesh Join Date: Mar 2009 Posts: 40 Rep Power: 17	Dear Cbas, Well, refinement but using threshold is always heuristic, you may have to tweak and play a bit there; there is no "universal" threshold, though people make claims of problem-independent thresholds that work well. Of course, multiple sensors may not be a complete solution to the problem; all I say is that it might be a partial solution. What I meant by geometric refinement, was to refine the area that you need by using a "geometric" sensor ie. mark that area for refinement always by using geometric constraints because you know that it needs refinement but your indicator won't flag it. That will sort the problem, but only this specific problem; it is neither intended to be a general solution nor is it one. Regards, Ganesh

August 22, 2010, 15:16		#5
Cbas New Member Sebastian Bode Join Date: Aug 2010 Posts: 4 Rep Power: 16	Ganesh Thank you! I've done that, and running the simulation now. Lets hold thumbs.