3D Separation Model using K-Omega SST Divergence

TempestCFD · February 12, 2013, 01:27

Hey guys,

I'm trying to model a 3-D wing in Fluent to study the effects of boundary restabilisation via Vortex Generators.

I started by using the k-e model, before realising that this was inappropriate for the high angles of attack I would be studying. I have now tried to switch to k-Omega SST to better capture boundary layer separation, and refined my mesh (50 layers of inflation, first off wall thickness of 1e-06 due to a fairly high Reynolds number of ~ 10^5). My domain is about 20 chord lengths radius (I have a cylinder) of my wing, and I'm using a pressure-far-field condition.

My solution doesnt last too long before it starts to diverge, and I have no clue why. I have been searching all over the forums for a good two days trying to solve this issue, but without any luck.
Any ideas?

-------------------------------------------------------------------------
Also, here are some of the parameters that I have set (just to reference):

1) Pressure Based, Steady
2) Energy - On, SST k-omega
3) Air - ideal gas (for pressure-far-field condition)
4) Velocity - 13.44 m/s
5) Courant # of 5

Attached is a picture of part of my mesh.

Thank you for your help

cfd seeker · February 12, 2013, 08:47

A common problem, not a big issue. Jut reduce your courant number to 1 at the start of solution and slowly increase it. By doing so, I hope your problem will be solved.

TempestCFD · February 12, 2013, 13:22

cfd seeker, thank you for your reply.

The initialisation now works, however the solution begins to diverge soon after, giving new errors (see screenshot attached).

I'm curious about the behavior of the continuity residuals at the beginning as well (staying at exactly 1 initially)?

I will try reducing my courant number further, but am open to other ideas as well.

Thanks in advance.

Frustrated but still determined,

TempestCFD

cfd seeker · February 13, 2013, 12:57

what is the mach no? in which sotware you made your mesh? what is the minimum qulity of your mesh? also trying with 1st order upwind for all equztions at the start of the solution.

TempestCFD · February 13, 2013, 13:24

The mach number is low (M = 0.0386).

I meshed the setup in Ansys Mesher.

My elements are fairly good quality and I have a fine mesh of about 6.5 million elements. Anything specific you want to know? Skewness ect?

I re-ran the simulation with a courant number of 0.1 and relaxed the energy from 1 -> 0.9 and the solution runs okay. The continuity residuals seem to flat line at 10e-02 though which is not satisfactory in my eyes. I'm going to look at improving the elements near the trailing edge of the airfoil as its sharp and not a great transition to the tet elements behind that section.

Any additional concepts or ideas would be greatly appreciated.

Thank you

cfd seeker · February 14, 2013, 08:35

Continuity convergence is very slow and secondly courant no and convergence are directly propotional, so let the solution run as it is and check the convergence of global parameters like CL,CD,CM etc.

TempestCFD · February 14, 2013, 08:39

I tried running the solution with a velocity inlet and pressure outlet instead, and the continuity went down to 10e-04. Much better.

My Cl and Cd seem exceptionally high however. I can't seem to find any issues with my reference values that would cause this.

Thank you for your guidence by the way, it is greatly appreciated.

cfd seeker · February 15, 2013, 14:21

Quote:

My Cl and Cd seem exceptionally high however

Have you resolved the boundary layer properly? what's the range of wall y+ ?
Have you scaled the mesh after importing it in Fluent?

TempestCFD · February 15, 2013, 14:29

My boundary layer (in my opinion) has been resolved well, even too well.

My y+ is ~ 0.15 - 0.89

What do you mean by "scaling the mesh after importing it into fluent" ?

cfd seeker · February 15, 2013, 15:10

Wall y+ 0.15 - 0.89 Wow it's great...what is the total mesh size? How much RAM does you machine have....
What is the units in which you created geometry and then meshed it? If it's other than "meters" then you have to scale it in Fluent, follow Mesh>Scale

TempestCFD · February 15, 2013, 15:25

Total mesh size is huge (about 5.5 million elements), 12 GB of ram, lol. It lags a bit

Everything was done in metres

Far · February 16, 2013, 08:44

Why energy is on? Are you using pressure based -coupled solver?

TempestCFD · February 16, 2013, 18:33

Sorry, I forgot to mention, I switched energy off, switching from pressure-far-field condition to a velocity-inlet/pressure-outlet setup.

Yes, pressure based, coupled.

February 12, 2013, 08:47		#2
cfd seeker Senior Member Join Date: Mar 2011 Location: Germany Posts: 552 Rep Power: 20	A common problem, not a big issue. Jut reduce your courant number to 1 at the start of solution and slowly increase it. By doing so, I hope your problem will be solved. mrenergy likes this.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Scaling omega in SST turbulence model	TedBrogan	Main CFD Forum	0	January 23, 2013 01:00
k - omega SST turbulence model with wall functions	kenan	Main CFD Forum	0	January 17, 2013 07:14
Water subcooled boiling	Attesz	CFX	7	January 5, 2013 04:32
Boundary Conditions for k omega SST	dancfd	OpenFOAM Pre-Processing	0	June 10, 2011 00:25
SST model	swe704	Main CFD Forum	0	February 5, 2010 09:36

February 13, 2013, 12:57		#4
cfd seeker Senior Member Join Date: Mar 2011 Location: Germany Posts: 552 Rep Power: 20	what is the mach no? in which sotware you made your mesh? what is the minimum qulity of your mesh? also trying with 1st order upwind for all equztions at the start of the solution.

February 13, 2013, 13:24		#5
TempestCFD New Member Join Date: Oct 2012 Posts: 7 Rep Power: 14	The mach number is low (M = 0.0386). I meshed the setup in Ansys Mesher. My elements are fairly good quality and I have a fine mesh of about 6.5 million elements. Anything specific you want to know? Skewness ect? I re-ran the simulation with a courant number of 0.1 and relaxed the energy from 1 -> 0.9 and the solution runs okay. The continuity residuals seem to flat line at 10e-02 though which is not satisfactory in my eyes. I'm going to look at improving the elements near the trailing edge of the airfoil as its sharp and not a great transition to the tet elements behind that section. Any additional concepts or ideas would be greatly appreciated. Thank you

February 14, 2013, 08:35		#6
cfd seeker Senior Member Join Date: Mar 2011 Location: Germany Posts: 552 Rep Power: 20	Continuity convergence is very slow and secondly courant no and convergence are directly propotional, so let the solution run as it is and check the convergence of global parameters like CL,CD,CM etc.

February 14, 2013, 08:39		#7
TempestCFD New Member Join Date: Oct 2012 Posts: 7 Rep Power: 14	I tried running the solution with a velocity inlet and pressure outlet instead, and the continuity went down to 10e-04. Much better. My Cl and Cd seem exceptionally high however. I can't seem to find any issues with my reference values that would cause this. Thank you for your guidence by the way, it is greatly appreciated.

February 15, 2013, 14:29		#9
TempestCFD New Member Join Date: Oct 2012 Posts: 7 Rep Power: 14	My boundary layer (in my opinion) has been resolved well, even too well. My y+ is ~ 0.15 - 0.89 What do you mean by "scaling the mesh after importing it into fluent" ?

February 15, 2013, 15:10		#10
cfd seeker Senior Member Join Date: Mar 2011 Location: Germany Posts: 552 Rep Power: 20	Wall y+ 0.15 - 0.89 Wow it's great...what is the total mesh size? How much RAM does you machine have.... What is the units in which you created geometry and then meshed it? If it's other than "meters" then you have to scale it in Fluent, follow Mesh>Scale

February 15, 2013, 15:25		#11
TempestCFD New Member Join Date: Oct 2012 Posts: 7 Rep Power: 14	Total mesh size is huge (about 5.5 million elements), 12 GB of ram, lol. It lags a bit Everything was done in metres

February 16, 2013, 08:44		#12
Far Senior Member Sijal Join Date: Mar 2009 Location: Islamabad Posts: 4,558 Blog Entries: 6 Rep Power: 54	Why energy is on? Are you using pressure based -coupled solver?

February 16, 2013, 18:33		#13
TempestCFD New Member Join Date: Oct 2012 Posts: 7 Rep Power: 14	Sorry, I forgot to mention, I switched energy off, switching from pressure-far-field condition to a velocity-inlet/pressure-outlet setup. Yes, pressure based, coupled.