[david caseiro]

Hello everyone,

I would like to ask your guidance in a problem I still have after one month of continuous search for resolution.

I am testing transient flow around a circular cylinder for Reynolds number of 8.000. In literature it is found that the experimental values of Cd is +-1.2, for this Reynolds number. I am obtaining Cd of about 1.6 - 1.8.

I have made all my mesh in ICEM CFD and tested several in order obtain a result independent of the mesh.

ICEM-CFD:
- All meshs conduct to a maximum y+ of around 1, the grid is progressive (1.07 in progressive ratio);
- Number of elements of the 2D meshs varies from 59.340 to 637.972.
- According to fluent, all meshs have a quality of +-0.76 in minimum orthogonal quality and +-9.49 maximum aspect ratio;
- Diameter of cylinder is 0.01m; Domain size 37*D x 10*D.

Fluent:
I am not activating Energy Model, and in Viscous Model, I have already tested k-e with EWT, sK-W, sst-kW, transition-sst; low reynolds stress omega RSM, as advised by Razvan is response to Sham problem.
Between sst-kW, transition-sst and the RSM the difference is minor compared to error and I have opted to use sst-kW.

I have read:

http://www.cfd-online.com/Wiki/Ansys..._inaccurate.3F

And I performed an analysis of all the subtopics found in 3. Currently I am doing the same tests in 3D, checking the dependence in the new direction (since 2D is already independent of the mesh).

About timestep, I have calculated the frequency from Strouhal number and programmed a 40 times smaller time step. I checked the timestep size dependence and did also tests with maximum courant number of 1 - better results were not achieved.

Conditions of the test:
- Water: density: 997 kg/m3; viscosity: 0.0008899 kg/m.s;
- Velocity-inlet: 0.7141m/s normal to boundary;
- Pressure-outlet: 0 Pa;
- Upper and Bottom walls: Free-slip (Specific Shear = 0 in both components);
- Cylinder: No slip;
- Reference Values: corrected so the values of Cd are correct;
- Scheme: Piso with both corrections at 1, Gradient: Least Squares Cell Based and others 2nd order including transient formulation;
- I let URF's default (but also tried to change them without major difference);
- I changed the residual criteria for all to 1e-5 of absolute criteria;
- I start the solution with standard initialization computed from inlet (I have already tried starting with a previous laminar test, without results);
- I have tested various time steps including the one that makes the maximum courant number around 1 and no better results were achieved.

The Cd have the behavior found on the literature with oscilations but the mean value is way too high for this Reynolds, I have checked if the Reynolds number is calculated correctly and if Cd fluent result is correct - They are.

I have found some literature with my problem:

"Numerical Study of Flow Past a Circular Cylinder Using RANS_HybridRANS_LES and PANS Formulations", A. Elmiligui et all.
"Detached_Eddy Simulations Past a Circular Cylinder", A. Travin et all.

I know that there are a few theads on this forum trying to solve this problem. I read them but I have still not found a solution to my problem. Exemples of similar threads:

http://www.cfd-online.com/Forums/cfx...e-10000-a.html

http://www.cfd-online.com/Forums/flu...-cylinder.html

http://www.cfd-online.com/Forums/flu...-shedding.html

Thank you,
David Caseiro

P.S: I will atach 2 images of the coarser mesh and 1 of the cd.

[davibarreira]

Your domain seems to be too small (from the literature, people recomend around 10*Diameter for the top boundaries, 10*Diameter for inlet and 20*Diameter for the outlet), test different domains sizes and see the effect on the results.

Also, if you look in the literature, you will see that for Reynolds around your value (subcritical regime) the 2D numerical results using RANS may differ a lot from the experimental value. (See for example: "URANS Calculations for Smooth Circular Cylinder Flow in a Wide Range of Reynolds Numbers: Solution Verification and Validation"). So it might not be a problem with your modelling, but with the model that can't predict such flow.

[david caseiro]

Thank you davibarreira for your response.

I will try to change domain sizes as you recommend. I am currently using 3D domain with pi * D (z direction (new dimension)) and the cd got a bit lower but it is still far from experimental results.

I will update the post when i have some results.

[davibarreira]

Good luck, mate. But anyways, if you look at the paper that I referenced, they also did 3D, with kw-SST, and got ~1.5 Cd for Re 10000. So your results might not be wrong, maybe the model is just not that good for flows past cylinders... :/

[Jennifer Von]

Can you tell me how you solved the problem at last?