CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   FLUENT (http://www.cfd-online.com/Forums/fluent/)

 p_kmb April 14, 2013 09:31

2D case with steady solver problems

Dear all,

I struggle with my case for a bit, and I need your help.
I am using Fluent steady solver for turbulent incompressible flow. I want to make comparison between flow over rectangular and rectangular with cylinders (within one domain), therefore I use triangle mesh.

The wind speed at velocity inlet is 10m/s, Iu=10% and the turbulence Length scale I calculated is L= 0.133. Other boundary conditions are symmetry and outflow. The walls of the domain are at least 10 times from the model.

I tried to used simple and coupled solution methods. First simple and default orders, then I switched to coupled second order. Second computations with coupled second from the beginning.

I cannot get convergence, my residuals oscillate. Is this because transient nature of flow? Should I use transient solver? If yes, how can I compare the results?

Paulina

 Far April 14, 2013 09:38

Please post some pics of your model and mesh. Why you are not using Hexa?

 p_kmb April 14, 2013 10:20

4 Attachment(s)
Thank you for quick replay.

There are two reasons I don't use hexa mesh. I try to simulate flow over rectangular and cylinders in one domain and I have four different angles of the wind attack. Therefore I created a circle with my model and I append it to the domain in Fluent, so the mesh is not changed, only the circle rotates.

My mesh have 457002 cells and Minimum Orthogonal Quality = 7.29122e-01
Maximum Aspect Ratio = 4.42918e+00.

I tried to use different mesh, hybrid, with boundary layer, different size. But I got some fluent sweep error using them. And I have to admit I almost like the results I got from computations on actual grid, if only it didn't behave like transient for steady solver.

If there is a mistake in my way of thinking from the beginning, please let me know.

I have problem with max file size that could be attached...

 Far April 14, 2013 11:06

I am pretty sure that your domain is too short. Make it 15 lengths upstream and 25-30 lengths downstream.

 p_kmb April 14, 2013 13:20

2 Attachment(s)
I changed the domain as you suggested.
I still run calculations, but after 2300i (1000 simple then coupled 2-nd order) I can say that results oscillate again.

Simulations are made with k-e RNG model.

Any idea what should I do? Change the mesh?

 Far April 14, 2013 16:55

Yes the flow is transient. To compare results you have to take into consideration time-average values of variables at particular location.

 p_kmb April 14, 2013 17:32

Thank you.

Unfortunately I know nothing about transient solver.
How can I get time averaged results? From post processing? How many time steps do I need too make a good average? I'll appreciate any info or tutorials.

One more question. When I change the angle of my model, should I also change turbulence length scale (especially with 90 degrees)? I have doubts if I can compare results with different turbulence length scale.

Can I even use the Fluent Manual equation L=0.4*d for external flows? Integral scale of turbulence for wind is ~100m. The result from the Fluent equation is 0.054m.

 All times are GMT -4. The time now is 17:16.