NACA0012 Grid Convergence - nor converging

aerospain · April 21, 2010, 09:27

Hi everyone,

I am caught in a loop and have tried everything I can think of to solve my problem.

I am trying to run a grid convergence study with a NACA0012 with blunt trainling edge: Re=2.88E06, M=0.16, alpha=3deg, chord=0.76. I am aiming for a cl value around 0.324.

The grid is a C-H shape type, with high definition at the leading and trailing edges, and grid lines departing orthogonally from the surface. Outer boundary is around 30 chord-lengths away from the airfoil in any direction. I have created a coarse grid to be refined in FLUENT twice in order to get the medium and fine ones. My y+ values are 1, 0.5 and 0.25 in the coarse, medium and fine grids respectively. Using k-w SST with no low-Re correction. Air is setup in density as ideal gas and the energy URF value is changed to 0.5.

I am not getting converged results from medium to fine grid. Either a straight line or a V-shape line when plotting lift coefficient for the three grids.

Any step-by-step guideline, suggestions, comments are very appreciated.

It's a bit urgent.

Thanx

Chris D · April 21, 2010, 09:54

Are you using Roache's method? Also, are you sure that your solutions are actually converged?

http://www.grc.nasa.gov/WWW/wind/val.../spatconv.html

aerospain · April 21, 2010, 10:30

Thanks for the quick reply Chris,

I am almost certain each calculation is converged, i.e. flat plot for several hundred iterations. I have run my calculation up to 2000, 4000, 6000, ...

Thanks for the link. I will tell you what I'm doing and we'll see how far off I am from Roche's Method. With the coarse grid opened in FLUENT I refine the domain by Adapt --> Region, and add one point in between points in both directions, i and j. And repeat a second time for the fine grid.

I am doubtful about my wake, it might be too refined.

cheers!

aerospain · April 21, 2010, 10:32

BTW, what about a reasonable value of shear wall stress for my test case in order to calculate y+?

I'm even thinking it could be my boundary layer refinement which is picking up "noise" and feeding the solution with it ...

Chris D · April 21, 2010, 10:49

Aren't you using FLUENT to calculate y+? Also, what spatial scheme are you using? Finally, going back to your OP, you said that you're not getting convergence from the medium to the fine grid. Maybe that means you need even more grid? I'm not really sure what you mean about the "noise" wrt the boundary layer.

aerospain · April 21, 2010, 11:03

I'm using the formula y+=rho*y*u_star/viscosity=2 for my coarse grid, assuming I will get 1 and 0.5 when refining it with FLUENT. When I plot y+ in FLUENT after a calculation is done I'm getting half the value I was expecting (which keeps me in the safe side), that's why I was asking about wall shear stress to use a good value of u_star.

I know FLUENT is capable of Adapting a grid in terms of y+ but I am not sure I would do it correctly. What would you recommend me to do in order to have three grids with the values of y+ I am aiming for?

I am not too certain about adding more points to the grids in order to get convergence between the medium and fine grids, I am already using grid dimensions that worked in a structured 2D code a few years ago.

With "noise" I mean numerical effects that keep my physical solution from converging nicely.

jack1980 · April 21, 2010, 11:08

Hi, also you might improve grid convergence by refining in Gambit. This will give a nicer grid that refining in Fluent.

aerospain · April 21, 2010, 11:10

My spatial discretization is Standard for Pressure and First Order Upwind for the other variables during the first couple of hundred iterations and changing to PRESTO! and QUICK respectively after that. I am not coupling pressure and velocity because I haven't been able to get a grid that doesn't blow up.

aerospain · April 21, 2010, 11:13

Thanks Jack.

I am using Gridgen to generate my grids. Because I am just adding points so I don't change the mesh topology and parameters, I can use FLUENT to add points with no worries.

BTW, I am in 12.0, so no GAMBIT anymore. I've never liked it, though. Personal preference.

jack1980 · April 21, 2010, 13:20

Hi, there are several things a grid generator does better. One is this:

Another is that it will decrease the growth factor. Also you can choose to have a constant first cell width in all grids (affecting your wally+).

Good luck!

aerospain · April 21, 2010, 13:33

Thanx Jack, I'll look into that.

At this moment something very wrong is happening in my grids. From coarse to fine mesh, my cl values curve describes a V-shape most of the times, some other times it is just a straight ascending or descending line.

I am just puzzled and about to miss a very big deadline.

Thank you guys for spending some time with me. I really appreciate it.

cheers!

April 21, 2010, 09:27	NACA0012 Grid Convergence - nor converging	#1
aerospain Senior Member aerospain Join Date: Sep 2009 Location: Madrid, Spain Posts: 149 Rep Power: 17	Hi everyone, I am caught in a loop and have tried everything I can think of to solve my problem. I am trying to run a grid convergence study with a NACA0012 with blunt trainling edge: Re=2.88E06, M=0.16, alpha=3deg, chord=0.76. I am aiming for a cl value around 0.324. The grid is a C-H shape type, with high definition at the leading and trailing edges, and grid lines departing orthogonally from the surface. Outer boundary is around 30 chord-lengths away from the airfoil in any direction. I have created a coarse grid to be refined in FLUENT twice in order to get the medium and fine ones. My y+ values are 1, 0.5 and 0.25 in the coarse, medium and fine grids respectively. Using k-w SST with no low-Re correction. Air is setup in density as ideal gas and the energy URF value is changed to 0.5. I am not getting converged results from medium to fine grid. Either a straight line or a V-shape line when plotting lift coefficient for the three grids. Any step-by-step guideline, suggestions, comments are very appreciated. It's a bit urgent. Thanx

April 21, 2010, 13:20		#10
jack1980 Senior Member Jouke de Baar Join Date: Oct 2009 Posts: 127 Rep Power: 17	Hi, there are several things a grid generator does better. One is this: Another is that it will decrease the growth factor. Also you can choose to have a constant first cell width in all grids (affecting your wally+). Good luck! Last edited by jack1980; April 21, 2010 at 13:21. Reason: image missing

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Grid size, convergence issues	franzdrs	Main CFD Forum	3	June 18, 2009 08:57
convergence for higher grid points	sandip	Phoenics	3	April 15, 2009 14:25
Convergence on non-uniform grid	dragon	Main CFD Forum	2	January 31, 2005 03:23
Help, gas combustion, grid quality, convergence	boji	CFX	1	September 16, 2002 09:46
Combustion Convergence problems	Art Stretton	Phoenics	5	April 2, 2002 06:59

April 21, 2010, 09:54		#2
Chris D Senior Member Chris Join Date: Jul 2009 Location: Ohio, USA Posts: 169 Rep Power: 17	Are you using Roache's method? Also, are you sure that your solutions are actually converged? http://www.grc.nasa.gov/WWW/wind/val.../spatconv.html

April 21, 2010, 10:30		#3
aerospain Senior Member aerospain Join Date: Sep 2009 Location: Madrid, Spain Posts: 149 Rep Power: 17	Thanks for the quick reply Chris, I am almost certain each calculation is converged, i.e. flat plot for several hundred iterations. I have run my calculation up to 2000, 4000, 6000, ... Thanks for the link. I will tell you what I'm doing and we'll see how far off I am from Roche's Method. With the coarse grid opened in FLUENT I refine the domain by Adapt --> Region, and add one point in between points in both directions, i and j. And repeat a second time for the fine grid. I am doubtful about my wake, it might be too refined. cheers!

April 21, 2010, 10:32		#4
aerospain Senior Member aerospain Join Date: Sep 2009 Location: Madrid, Spain Posts: 149 Rep Power: 17	BTW, what about a reasonable value of shear wall stress for my test case in order to calculate y+? I'm even thinking it could be my boundary layer refinement which is picking up "noise" and feeding the solution with it ...

April 21, 2010, 10:49		#5
Chris D Senior Member Chris Join Date: Jul 2009 Location: Ohio, USA Posts: 169 Rep Power: 17	Aren't you using FLUENT to calculate y+? Also, what spatial scheme are you using? Finally, going back to your OP, you said that you're not getting convergence from the medium to the fine grid. Maybe that means you need even more grid? I'm not really sure what you mean about the "noise" wrt the boundary layer.

April 21, 2010, 11:03		#6
aerospain Senior Member aerospain Join Date: Sep 2009 Location: Madrid, Spain Posts: 149 Rep Power: 17	I'm using the formula y+=rhoyu_star/viscosity=2 for my coarse grid, assuming I will get 1 and 0.5 when refining it with FLUENT. When I plot y+ in FLUENT after a calculation is done I'm getting half the value I was expecting (which keeps me in the safe side), that's why I was asking about wall shear stress to use a good value of u_star. I know FLUENT is capable of Adapting a grid in terms of y+ but I am not sure I would do it correctly. What would you recommend me to do in order to have three grids with the values of y+ I am aiming for? I am not too certain about adding more points to the grids in order to get convergence between the medium and fine grids, I am already using grid dimensions that worked in a structured 2D code a few years ago. With "noise" I mean numerical effects that keep my physical solution from converging nicely.

April 21, 2010, 11:08		#7
jack1980 Senior Member Jouke de Baar Join Date: Oct 2009 Posts: 127 Rep Power: 17	Hi, also you might improve grid convergence by refining in Gambit. This will give a nicer grid that refining in Fluent.

April 21, 2010, 11:10		#8
aerospain Senior Member aerospain Join Date: Sep 2009 Location: Madrid, Spain Posts: 149 Rep Power: 17	My spatial discretization is Standard for Pressure and First Order Upwind for the other variables during the first couple of hundred iterations and changing to PRESTO! and QUICK respectively after that. I am not coupling pressure and velocity because I haven't been able to get a grid that doesn't blow up.

April 21, 2010, 11:13		#9
aerospain Senior Member aerospain Join Date: Sep 2009 Location: Madrid, Spain Posts: 149 Rep Power: 17	Thanks Jack. I am using Gridgen to generate my grids. Because I am just adding points so I don't change the mesh topology and parameters, I can use FLUENT to add points with no worries. BTW, I am in 12.0, so no GAMBIT anymore. I've never liked it, though. Personal preference.

April 21, 2010, 13:33		#11
aerospain Senior Member aerospain Join Date: Sep 2009 Location: Madrid, Spain Posts: 149 Rep Power: 17	Thanx Jack, I'll look into that. At this moment something very wrong is happening in my grids. From coarse to fine mesh, my cl values curve describes a V-shape most of the times, some other times it is just a straight ascending or descending line. I am just puzzled and about to miss a very big deadline. Thank you guys for spending some time with me. I really appreciate it. cheers!