CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   CFX (https://www.cfd-online.com/Forums/cfx/)
-   -   2D SST Simulation Airfoil - Convergence Problem (https://www.cfd-online.com/Forums/cfx/21710-2d-sst-simulation-airfoil-convergence-problem.html)

 Kraemer October 18, 2005 06:51

2D SST Simulation Airfoil - Convergence Problem

Hi everyone,

i am trying to solve a 2D Airfoil with SST. When starting the solver everything looks fine. Good convergence behaviour for the first 200 iterations (Auto Timescale) then the residuals start to fluctuate periodically and no final solution is accomplished. (MAX Res 1e-05) Does anyone know this behaviour?

What about the 3rd dimension in a 2D situation working with Icem CFD? The chord of the airfoil is about 1.5 m. I have set the 3rd dimension where nothing should happen (symetric BC's) to 0.01 m. There are some aspect ratio above 5000 because of that, is that a problem? Determinants, angles are fine.

Kraemer

 Glenn Horrocks October 18, 2005 18:24

Re: 2D SST Simulation Airfoil - Convergence Proble

Hi,

Is the airfoil stalling or creating a wake? There might be transient structures in the wake causing convergence problems for steady state.

Glenn Horrocks

 Kraemer October 19, 2005 03:42

Re: 2D SST Simulation Airfoil - Convergence Proble

Hi Glenn,

it could be stalling, when it is so what we be the next step to get a good solution. I need drag and lift values.

Kraemer

 andrew October 19, 2005 05:02

Re: 2D SST Simulation Airfoil - Convergence Proble

Don't you actually see wheather it stalls or not? I simulated 2D airfoil with SST option enabled and found no separation. Strange.

 Glenn Horrocks October 19, 2005 18:52

Re: 2D SST Simulation Airfoil - Convergence Proble

Hi,

When you stall the airfoil or have large separations off some bluff body you generate transient 3D flows. If you want accurate lift and drag numbers in this region you will need to do a 3D simulation, and possibly use a more advanced turbulence model, eg LES.

Glenn Horrocks

 iffy April 10, 2011 15:29

Hi everyone,
I have to determine boundary layer characteristics for 2D airfoil. For this I have to do grid refinement study. I created three C-mesh with 100000, 400000 and 1 million elements. I have to first obtain laminar steady state solution for these meshes and see if there is any variation in the solution. The first two meshes (with 100000 and 400000 elements) converge but the mesh with 1 million elements does not converge. I am using very low Reynolds number (30000). The residuals decrease to 1e-5 and then starts creeping up.
Anyone knows the possible reasons behind this?

Irfan

 ghorrocks April 10, 2011 22:20

http://www.cfd-online.com/Wiki/Ansys...gence_criteria

 iffy April 14, 2011 05:49

Hi Glenn,
As my steady state simulation was not converging for the fine mesh (mesh 3, 1 million elements, but it did converge for mesh 1 and mesh2 with 100,000 and 400,000 elements respectively) I tried to run the simulation transient (also recommended in the link which you mentioned). Although the simulation seems to converge as the residuals were decreased to about 1e-8 but I got totally different results. The transient simulations show no separation on the airfoil in mesh 3, while the stead state simulations for the mesh 1 and mesh 2 were showing separation at about 70% chord.
Now I don't understand which results should be trusted.
Any idea/clue why there is so much difference b/w steady and transient results? (the courant number for the transient simulation was less then 1)

With kind regards,

Irfan

 ghorrocks April 15, 2011 06:41

It sounds like you have only run one time step of a transient solution. You need to run a transient solution for enough time steps that you have captured the flow. This will probably be thousands of time steps.

 iffy April 15, 2011 08:53

Thanks again Glenn,
You guessed right ! I was running transient simulation for few hundred timesteps because the residuals became stable. Now I will run simulation for more timesteps.
One question regarding meshing: I am using unstructured hexa mesh for airfoil simulations, and do you think it is ok to use unstructured or should I use structured mesh? Is there any effect of mesh type on convergence and the accuracy of solution?

With kind regards,

Irfan

 ghorrocks April 16, 2011 07:22

The quality of the mesh is more important than the element type. But if you have a choice hexas are the preferred element (providing it is not by reducing mesh quality).

 All times are GMT -4. The time now is 07:03.