# Flow Past a Cylinder with Non-Zero Average Lift Coefficient

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

 May 21, 2020, 21:21 Flow Past a Cylinder with Non-Zero Average Lift Coefficient #1 New Member   Innocentio Loe Join Date: May 2020 Posts: 4 Rep Power: 3 Hello, I've been working on simple flow past a cylinder simulation with Re=100. Boundary condition is flat velocity profile on the inlet, 0 on the outlet, and slip top and bottom wall. The domain is 20d in width and 10d in height, with the cylinder centered at the origin, 5d to both left and bottom wall I'm trying to validate the result, compared to literature data. Expected lift and drag coefficient (CL and CD) around the cylinder wall are CL = +/-0.33 and CD = 1.34 +/-0.009 The simulation is working fine. However, what I notice is that even though my CL oscillates at the correct amplitude, it is not oscillating around 0 average but instead has small negative average value instead. Also my CD is oscillating around 1.42 instead, so I guess there is the same DC drift happening as well here. For the CL, it is easy to remove this additional drift, but for CD I cannot really do so, since there is no baseline of judgement. My question is whether this is commonly encountered? If so, how to validate such a case then? I'm using COMSOL Multiphysics at the moment. Might try to compare with OpenFOAM later on.

 May 22, 2020, 03:07 #2 Senior Member   Filippo Maria Denaro Join Date: Jul 2010 Posts: 5,736 Rep Power: 60 Have you checked that the grid is perfectly symmetric?

May 22, 2020, 06:41
#3
New Member

Innocentio Loe
Join Date: May 2020
Posts: 4
Rep Power: 3
Quote:
 Originally Posted by FMDenaro Have you checked that the grid is perfectly symmetric?
Hi, thank you for your input. Certainly I haven't thought about the mesh symmetry. Also, currently I am using an unstructured mesh which maybe will make it harder to create symmetrical mesh. I will try using a structured mesh.

 May 24, 2020, 05:20 #4 Senior Member     Gerry Kan Join Date: May 2016 Posts: 284 Rep Power: 8 Dear innoaloe: One thing you should also check is whether the CL oscillates about some value as the solver iterates. It could be that the solution is oscillating, as it happens physically. Look at your solution field to verify. Gerry PS I am sorry I did not read your entire message. I would second with FMDenaro and use a perfectly symmetric mesh. I don't know if you can do this natively with COMSOL but I would be surprised if you couldn't. Otherwise you need a very fine mesh to attenuate any asymmetric effects.

May 24, 2020, 07:19
#5
New Member

Innocentio Loe
Join Date: May 2020
Posts: 4
Rep Power: 3
Quote:
 Originally Posted by Gerry Kan Dear innoaloe: One thing you should also check is whether the CL oscillates about some value as the solver iterates. It could be that the solution is oscillating, as it happens physically. Look at your solution field to verify. Gerry PS I am sorry I did not read your entire message. I would second with FMDenaro and use a perfectly symmetric mesh. I don't know if you can do this natively with COMSOL but I would be surprised if you couldn't. Otherwise you need a very fine mesh to attenuate any asymmetric effects.
Dear Gerry,
thank you for your input as well. I did as FMDenaro's suggested to create symmetrical mesh, also ensuring the mesh around the cylinder to be more refined. And yes, it is possible to do so in COMSOL.

I managed to get an acceptable result by doing so. There is still a very tiny average value on the oscillation, but the value is negligible if compared with the CL oscillation itself, so in my opinion such small numerical error is to be expected.

Regards

 Tags comsol, drag coefficient, lift and drag, lift coefficient