CFD Online Logo CFD Online URL
www.cfd-online.com
[Sponsors]
Home > Forums > Software User Forums > ANSYS > FLUENT

Forcing an airfoil to separate-help

Register Blogs Members List Search Today's Posts Mark Forums Read

Reply
 
LinkBack Thread Tools Search this Thread Display Modes
Old   January 9, 2002, 13:03
Default Forcing an airfoil to separate-help
  #1
Andrew Parker
Guest
 
Posts: n/a
Dear all

I am looking for a little help in getting an airfoil to separate. I am running a 2D simulation of a NACA 0012 and do not seam to be able to get the flow to separate, despite a good grid, good y+ and so on. Has anybody had a similar problem and what was the remedy. Did it for example involve hiking up the inlet turbulence properties or one specific turbulence model gave separation when the rest did not. I am quite experienced at modelling this problem, so I believe the problem could be in a more advance function or feature, however, any suggestions would be very very welcome.

Cheers

Andy
  Reply With Quote

Old   January 9, 2002, 15:27
Default Re: Forcing an airfoil to separate-help
  #2
Jonas Larsson
Guest
 
Posts: n/a
What kind of turbulence model are you running? If you run a wall-function (std or non-equil) case it is quite common to get too late separation. You can't really expect a wall-function model to predict separation correctly. The two-layer model (with y+ below 1) is better at predicting separation. You could also try one of the low-Re models available from the text-interface - I'd recommend the Yang-Shih model. I have tested all these models on cases which involves separation on turbine blades and unfortunately none is very consistent - we are now mostly using the two-layer model since it is more robust.

Are you sure that it is a turbulent separation you are looking for btw? A quick check of what effect the turbulence model has is to run it laminar - if you don't predict a laminar separation then you will never predict separation with a turbulence model.

Btw, if you want to get an earlier separation you should *reduce* your inlet turbulence level. However, since you are running a wing I assume that you already have a fairly low level (below 1%).
  Reply With Quote

Old   January 9, 2002, 19:26
Default Re: Forcing an airfoil to separate-help
  #3
Rob
Guest
 
Posts: n/a
I had a similar problem while running a 2-D case of an airfoil. Using the S-A turbulence model, with proper y+ values, I was not able to find separation. However I am now using the standard k-e model with fixing a transition point from laminar to turbulent and I am getting pretty accurate separation results. I am comparing the Fluent results with MSES results, which we know to be quite accurate. Hopefully this helps you in some small way.
  Reply With Quote

Old   January 10, 2002, 06:00
Default Re: Forcing an airfoil to separate-help
  #4
Andrew Parker
Guest
 
Posts: n/a
Cheers

That's all very helpful, I may have to get back to you on some points later on, as I am unsure about some of your points. Mainly the inlet turbulent intensity =1%. I am using an old wind tunnel with a speed of 18 m/s and I would have thought the inlet turbulence would be higher. Are there any empirical formulas to calculate the inlet turbulence properties for my wind tunnel for this speed and the tunnel dimensions? I believe that my problem is that I am not specifying the boundary conditions properly, mainly because I don't know how to work them for for my wind tunnel.

After I posted last night I did manage to get my airfoil to separate. But I must admit it was strange. For the same y+ and all other parameters the flow does not separate on a tri mesh, but if I have a quad mesh on the surface, then it does separate, not sure about that??????

Cheers again,

Andy
  Reply With Quote

Old   January 10, 2002, 09:53
Default Re: Forcing an airfoil to separate-help
  #5
Jonas Larsson
Guest
 
Posts: n/a
Does your wind-tunnel have any turbulence grid? If not, 1% turbulence level is quite high, but could be okay for a fairly bad wind-tunnel. What turbulence level a wind-tunnel has varies a lot depending on how good it is. This is often a well known property of a tunnel though. Perhaps someone who has used the tunnel before knows it? It can be measued with a laser or hot-wire.

What turbulence model are you using and what kind of near-wall model (wall function, two-layer or low-Re)? What Y+?

It is quite common to get poor results on separation and skin-friction if you use a mesh of only tets/tris. Hex/quad meshes work much better in boundary layer flows since the cells will be aligned with the flow. Tets/tris are much more dissipative and it is essential that you use 2nd order schemes with them to get any accuracy. You should make some kind of grid-sensitivity study. Is it a 2d or 3d case? How many cells?
  Reply With Quote

Old   January 10, 2002, 10:13
Default Re: Forcing an airfoil to separate-help
  #6
Andrew Parker
Guest
 
Posts: n/a
Dear Jonas

Cheers again. I think my uni has a hot wire anemometer, but I have never used and I don't know what condition it is in. I am making comparisons between k-e standard with two-layer zonal, and S-A. Both meshes have a y+ on the upper surface of the airfoil of 1, and on the lower surface a maximum of 5. The cell count varies but is around 28,000 before adaptation on y+ and 41,000 roughly after adaptation. Thanks, I did not realise that an intensity of 1% was high and would be fine for even a bad wind tunnel, its not bad just old. I take it that the intensity in the S-A model and the k-e is the same and I can use one for both. At the moment I am specifying k and e at inlet on the k-e model and intensity and hydraulic diameter of the wind tunnel on the S-A model. I am going to do my best to find some data on the tunnel to see if the intensity is specified.

One last question Jonas, can I use Yang-Shih, that you recommended and the low-Reynolds equations. I have fluent version 5.5, but I did not think I could use one of these low Reynolds number models as my Reynolds number is about 0.18 million. What classes as low Reynolds numbers in this case. Lastly good to find out about the difference between tri, and quad meshes. I glad it is a known thing but I am pleased I managed to find this out for myself.

Cheers again.

Andy
  Reply With Quote

Old   January 10, 2002, 11:53
Default Re: Forcing an airfoil to separate-help
  #7
Jonas Larsson
Guest
 
Posts: n/a
An easy way to check that your specified inlet turbulence parameters are not completely off is to plot the eddy-viscosity ratio on the inlet - it should be a realistic value (not much below one and not huge, say above 1000). It should of course be the same for all models. For a "wing wind-tunnel" my very rough guess is that 50 is realistic, but I'm not that familiar with this type of wind-tunnels so please ask someone else (any better suggestions from someone else?). If I didn't know anything better I would guess a turbulence level of 1 percent and set the length-scale to say 5% of the channel-height.

If the inlet turbulence level is imporant the SA model will not work very well I think - it is not good at predicting turbulent decay and you will probably see that even if you have the same eddy-viscosity on the inlet it will be different compared to k-epsilon once the flow reaches the wing.

You can certainly use a low-Re model for this type of flows - the "low-Re" refers to the fact that it is possible to use these models down to the wall through the viscous sublayer where the turbulent Re number is low (the SA model and the two-layer model are also in some sense low-Re models). Think of "low-Re" as meaning that you can use it with resolved walls and should have y+ below 1.

To use a low-Re model in Fluent you have to activate it through the text interface by typing something like:

/define/models/viscous/low-re-ke

you select model with

/define/models/viscous/low-re-ke-index

The standard model is Launder-Sharma (index 2 I think). It is an old and well-established low-Re k-epsilon model which performs reasonably well in your type of flows I think. I'd recommend the Yang-Shih model though (index 3 I think - look in the online release notes for Fluent, they are specified there).

Did you try to run it laminar btw?
  Reply With Quote

Old   January 10, 2002, 12:34
Default Re: Forcing an airfoil to separate-help
  #8
Andrew Parker
Guest
 
Posts: n/a
Dear Jonas

I did run it on laminar flow, and the flow is badly separated, thanks for the suggestion. All the information is top-notch thanks very much. I think you are right about the sensitivity of the S-A model and to some extent the k-e for the turbulence boundary conditions. I will really need to get my definitions sorted for the turbulence parameter and get them defined correctly.

One question, when you say set the length scale to 5% of the tunnel-height, I take it that length scale is an option in S-A and k-e, I don't remember seeing it, and I take it you don't me hydraulic diameter. Cheers for the info on the index command, I was having some problems finding them. I also cannot find the details about these models in the help menu, are they there or on the web site?

Cheers again Jonas

Andy
  Reply With Quote

Old   January 10, 2002, 20:02
Default Re: Forcing an airfoil to separate-help
  #9
Steve
Guest
 
Posts: n/a
Have you tripped the boundary layer. If not then there could be a laminar region at the nose which would promote separation.

Also both the SA and k-e models will overpredict turbulent production at the stagnation point - this may supress separation. This effect will be dependent on the background turbulence levels. Perhaps try the kato-launder correction.

k-e will suppress APG driven separation considerably due to overprediction of the length scale in the defect layer .

Have you tried the v2f model (is the transition model in fluent)?

  Reply With Quote

Old   January 11, 2002, 11:15
Default Re: Forcing an airfoil to separate-help
  #10
Jonas Larsson
Guest
 
Posts: n/a
I think that you can only specify the length-scale with the k-eps models, not with SA.
  Reply With Quote

Old   January 11, 2002, 11:18
Default Re: Forcing an airfoil to separate-help
  #11
Jonas Larsson
Guest
 
Posts: n/a
The realizable k-eps model behaves better in stagnation regions - less over-production. Kato-Launder isn't available in Fluent I think. KL might be okay for a wing, but I don't like using it in cases where free-stream turbulence is important, since it essentially turns off the turbulence model outside the boundary layers and wakes.
  Reply With Quote

Old   January 13, 2002, 00:30
Default Re: Forcing an airfoil to separate-help
  #12
Daniel Bruno
Guest
 
Posts: n/a
Why the Full Low Reynolds is not described in the user's manual ? I've never heard about it. I thought the Two-Layer Zonal was the Fluent version of the low-Re model, since in the manual is said that it separates the near-wall region into two zones (based on a wall distance Reynolds number, Re_y) and solve down it to the wall.

And about the transion modelling ? Is Fluen capable of it or not ?!?!?

Thanks,

Daniel.
  Reply With Quote

Old   January 13, 2002, 06:44
Default Re: Forcing an airfoil to separate-help
  #13
Jonas Larsson
Guest
 
Posts: n/a
The two-layer zonal model is more robust and handles grids with a bit too high y+ values better. I think that Fluent decided that this model is better suited as a general low-Re model.

The real low-Re k-epsilon models available from the text-interface are provided as an extra feature for "experienced users", who are familiar with these models and know what they are doing. I tend to agree with Fluent's choice here - although I have long exprience from many low-Re models I mostly use the two-layer model instead. The low-Re models are not available with the any "realisability" fixes, which is a pity.

As far as I know there are no transition models in Fluent and I don't think you should try using any of the existing models to predict by-pass transition, although some of them could theoretically predict by-pass transition (transition caused by diffusion of turbulence from the free-stream into the boundary layers - not the same as natural transition caused by instabilities in the boundary layer). If you want to predict transition I'd recommend you to implement an ad-hoc transition model tuned for your application area using some kind of UDF. I haven't heard of anyone who has done so, but I have a vague memory of having seen this discussed here on this forum before - search the archives using the site-wide search engine in the upper right corner.
  Reply With Quote

Old   January 13, 2002, 10:20
Default Re: Forcing an airfoil to separate-help
  #14
Daniel Bruno
Guest
 
Posts: n/a
Thanks for the response, Jonas. I thought I was becoming crazy...

The lack of a transitioning model invalidates the use of Fluent to predict airfoil aerodynamic characteristics. How can you capture the effect of a longer laminar BL in a laminar airfoil, for instance?

Daniel.
  Reply With Quote

Old   January 14, 2002, 08:32
Default Re: Forcing an airfoil to separate-help
  #15
kim
Guest
 
Posts: n/a
Jonas, what do you mean by realisability?
  Reply With Quote

Old   January 14, 2002, 09:12
Default Re: Forcing an airfoil to separate-help
  #16
Jonas Larsson
Guest
 
Posts: n/a
I think that the term realizability in turbulence models originates from a paper by Schumann entitled "Realizability of Reynolds stress turbulence models" (Phys. Fluid, 1977, vol 20, pp 721-725). Shumann's ideas were later picked up by Lumley and others - the best early reference is probably his 1978 paper "Computational Modeling of Turbulent Flows" (Ads. Appl. Mech., vol 18, pp 123-176).

The basic idea behind realizability is that you can use Schawrz' inequality to mathematically derive a number of constratints which your turbulent properties much obey - an obvious example is that k, the turbulent energy, must never become negative. Using these constraints you can then derive additional constraints for your turbulence model. There are many models which use these realizability constraints - see papers by people like Shih, Durbin, Speziale, Lumley, ...

A standard k-eps does not satisfy realizability. By adding a realizability correction (Durbin has one popular "fix", Shih has another which I like) to a k-eps model you can remove some obvious problems with the standard k-eps, like over-production of turbulent energy in flows with large normal strain (strong decelleration, like in stagnation regions, or strong acceleration).

  Reply With Quote

Reply

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are On
Refbacks are On


Similar Threads
Thread Thread Starter Forum Replies Last Post
Low Speed Airfoil Mancusi FLUENT 7 April 3, 2014 07:11
[GAMBIT] Meshing airfoil using .dat file problem creggie ANSYS Meshing & Geometry 10 June 27, 2010 20:24
ICEM CFD Airfoil Tutorial Input Tom ANSYS Meshing & Geometry 1 September 25, 2009 07:50
Modeling Backflow for a 3D Airfoil (Wing of Finite Span) Josh CFX 9 August 18, 2009 12:31
Airfoil boundary condition Frank Main CFD Forum 1 April 21, 2008 19:36


All times are GMT -4. The time now is 21:40.