CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   OpenFOAM Running, Solving & CFD (http://www.cfd-online.com/Forums/openfoam-solving/)
-   -   2d aerofoil simpleFoam (http://www.cfd-online.com/Forums/openfoam-solving/94812-2d-aerofoil-simplefoam.html)

plm November 26, 2011 14:34

2d aerofoil simpleFoam
 
Hi all,
I'm using simpleFoam to model turbulent flow over an aerofoil. When the angle of attack increases to a point where there is separation from the aerofoil, my model is unable to predict this.

Is this due to the use of the spalmartAllmaras turbulence model or is this a limitation of simpleFoam?

I've also read online about the use of DES which is potentially a better solver but I'm unsure how to set it up and would welcome any comments on it's effectiveness

plm

plm November 28, 2011 14:21

Is it even possible to run DES/LES cases in 2D?
Really would welcome any comments! Thanks!

wiedangel November 29, 2011 06:39

Quote:

Originally Posted by plm (Post 333833)
Is it even possible to run DES/LES cases in 2D?
Really would welcome any comments! Thanks!

Hi plm,

it is possible to run 2D cases using DES (Detached Eddy Simulation). I did it for a thick airfoil, you just need a fine enough mesh.

plm November 29, 2011 07:20

Thanks wiedangel,
I have managed to run a DES case with a fine mesh but I'm still not predicting the stall condition - any ideas?

wiedangel November 29, 2011 11:01

Hi,

can you tell me which solver you are using? Are you simulating for each angle of attack individually or do you have some motion handling solver like pimpleDymFoam??

I am using pimpleFoam and no motion, I simulated pre- and post-stall situations and it seems to work.

Regards.

vkrastev November 29, 2011 11:45

1) before shifting to a new solver (e. g. transient pimpleFoam instead of steady-state simpleFoam) or a new turbulence model (e. g. k-omega SST instead of Spalart-Allmaras) you have to be sure that you have reached the maximum reasonable accuracy with your initial choices, and not simply change everything (solver+modeling) because of not satisfactory results...how about your mesh quality and solver tolerances? What is your wall treatment (and consequently the mesh near-wall spacing)? What about the numerical schemes?;

2) running a DES (or any kind of LES-like turbulence model) in 2D is not consistent with the vorticity dynamics, which is inherently three-dimensional (and unsteady): hence, if you want to keep a 2D modeling it will be better to try an unsteady RANS approach.

Best regards

V.

plm November 29, 2011 14:36

Thank you for the instructive comments vkrastev and wiedangel

I have tried using both the steady state simpleFoam and transient pimpleFoam solvers, both with Spalart-Allmaras turbulence modelling.

I have had a DES running but from what vrakstev said will look to improve my unsteady RANS approach instead....

I have been using various wall functions and numerical schemes - is it likely that these may have an effect on whether I find stall or not?

vkrastev November 29, 2011 14:58

Quote:

Originally Posted by plm (Post 334012)
I have been using various wall functions and numerical schemes - is it likely that these may have an effect on whether I find stall or not?

Yes, of course, but also solvers tolerances and, especially, boundary conditions and mesh quality and resolution (with respect to the particular near wall modeling) can have a large impact on the solution. To catch the stall phenomenon with a steady approach is quite hard, but what I'm saying is that simulating a case is not simply a mix of type of solver+type of turbulence model, and this is true whichever the solution approach. As I said before, for your particular case probably a well resolved unsteady-RANS simulation will be the best solution, but only with a proper combination of the factors cited above. Good luck for your work.

V.

wiedangel November 29, 2011 17:22

Hi, plm. A good idea is also to check the previous work done by other researchers, there are some papers about the best practices when dealing with stall and turbulence modeling. Like vkrastev mentioned, sometimes one has to go for the simplest model and not just mix models and spring from one to another just because one did not work. I will be happy to send you some references if you are interested.

good luck with stall ;)

plm November 29, 2011 17:28

I'd definitely be interested in your references wiedangel, cheers!

I was jumping around a bit because I was unsure if the simple methods I am using can actually predict stall - at the moment they are simply predicting an increase in the predicted pressure distribution, as if stall has never occured.

I'll keep at it and see what I can come up with.

Thanks :D

thinkagain November 30, 2011 08:36

I am also intersted in your reference!

wiedangel December 9, 2011 05:36

sorry for the delay in sending the references ... I could not find that post :o
here are some references about DES and the grid quality they have to satisfy:

http://books.google.de/books/about/D...sC&redir_esc=y

Spalart, P. R., Jou, W.-H., Stretlets, M., and Allmaras, S. R. (1997), "Comments on the Feasibility of LES for Wings and on the Hybrid RANS/LES Approach", Advances in DNS/LES, Proceedings of the First AFOSR International Conference on DNS/LES.

INVESTIGATION OF GRID RESOLUTION REQUIREMENTS FOR
DETACHED EDDY SIMULATION OF FLOW AROUND THICK AIRFOIL
SECTIONS
Jesper Madsen, Kaja Lenz, Pavitran Dynampally, Sudhakar P.
LM Glasfiber A/S, Denmark, LM Glasfiber R&D (India) Pvt. Ltd.

Profile Catalogue for Airfoil Sections
Based on 3D Computations
Franck Bertagnolio, Niels N. Sørensen and Jeppe Johansen

They helped me a lot, I hope it will help you.

plm December 9, 2011 05:46

Thank you :)

nabilhaneef December 18, 2011 05:58

Regarding results obtained for K-Ɛ, K-Omega and K-Omega SST
 
2 Attachment(s)
Hi,

I have been running simulations on a simple 2D Nozzle guide vane. I have used different turbulence models and found out the results for the K-Ɛ and K-Omega SST were very close to each other for various points on the suction side and pressure side of the blade. The plot is for Velocity profile versus different points on the suction side. I am attaching a graph that i have plotted for the various turbulence models. Can someone guide me why is this is happening. Why K-Omega values are not close to the other models. I have my submission of my project report very close by and would highly appreciate if someone could guide me with the answers to why this is happening.

lucaBonfiglio April 11, 2012 16:44

Hi plm,

I'm having the same problems simulating an hydrofoil naca0012 at Re 3e6, did you figure out something from the pimpleFoam solver??
I'm also trying the kklOmega turbolence model with the simpleFoam in order to reproduce laminar flow and the transition, but unfortunately with no success.


All times are GMT -4. The time now is 13:45.