pimpleDyMFoam issue for pitching foil
I have a problem using pimpleDyMFoam when simulating a pitching hydrofoil (pitching from 0° to 15°): the solution converges but the lift coefficient is lower than expected, up to 50% ! The expected values come from CFX AND experiments so I think they are reliable.
I use OpenFOAM 2.1.1 with the k-omega SST turbulence model. I have tried two different ways to obtain the pitching but both give bad results :
- deforming mesh (dynamicMotionSolverFvMesh)
- sliding mesh (AMI)
I have read in other threads that pimpleDyMFoam may be unstable for CFL higher than 2 so I chose an adaptive time step with CFL = 1 (I have also tested fixed time steps dt=10^-5 and dt=10^-4).
I have tested 3 different meshes (22k, 41k and 76k cells).
I have tested 2 different angular velocities (6°/s and 63°/s).
Results of all these different tests differ a little but none of them gives good results...
It is strange because calculations at fixed angles of attack with simpleFoam and pimpleFoam give very good results for both lift and drag coefficients.
Do you have an idea regarding this issue ? Does it come from the solver pimpleDyMFoam ?
Thanks in advance,
I would try with fixed smaller time step.
How many iteration / time do you need to converge with SimpleFoam / PimpleFoam ? =>
I guess your pitching is too fast to catch the physics.
So try to lower time step, or look for a more coupled solver.
Thank you for your reply. I have tried with a fixed time step of 10^-6 but it still gives bad results (approximately the same as with dt=10^-5 and with the adaptive time step CFL=1). I don't think the problem comes from the time step since the 6°/s case is a quasi-static case (reduced frequency = 0.18) and should not require such small time steps. I know I should not compare the two codes, but CFX only requires dt = 10^-3 for this case...
For the pimpleFoam calculation at an angle of attack of 10°, which showed good results, I used the following parameters:
The time step was 5x10^-4 and the lift coefficient was converged after 2000 time steps.
I now try to use OF 1.6-ext to check if it gives better results (I suppose it has more experience with this kind of simulation).
When you say "look for a more coupled solver", to what solver do you think ?
|All times are GMT -4. The time now is 20:50.|