[ojha.mayank485]

Hello,

Am trying to simulate the displacement of a cylinder in a cross flow at Re=120. I first did a mesh and time convergence study on a stationary cylinder and got quite satisfying results but when I used a moving mesh in CFX 12.0, I am unable to get the desired Amplitude. Can anyone help me out as to what is going wrong ????

1. My timesteps are small enough to capture the vortex shedding phenomena.
2. I am using a higher order scheme.
3. No turbulence involved as flow is laminar.
4. Am using a CEL expression for cylinder displacements.

Has anyone done simulations for low Re flow over cylinder????


Thank you very much,

Mayank Ojha

[ghorrocks]

Assuming you have accurately modelled the stationary cylinder then it suggests your implementation of the cylinder motion is either incorrect or not accurate. Have you considered doing this with the various built-in rigid body solvers in CFX V13, such as the immersed solid approach and the rigid body solver?

[ojha.mayank485]

Originally Posted by ghorrocks

your implementation of the cylinder motion is either incorrect or not accurate.

Does that mean that the ODE [ mass*acc + damper*vel + spring stiffness*disp=force ]
is being modelled wrong ????
When I tried to give a pre-defined forcing value such as force=sin( t / 1[s]), It gives me the O/p what am expecting. I even used a higher order RK4 scheme for it. But it does not work.

Originally Posted by ghorrocks

Have you considered doing this with the various built-in rigid body solvers in CFX V13, such as the immersed solid approach and the rigid body solver?

How will using a different solver matter ??? I am having CFX 12.0 and not 13. It will be some time till the University server gets updated with CFX 13

Thank you for your reply.

P.S: I have been breaking my head over this for a long time now. I need to get done with this ASAP. Any kind of suggestions is highly appreciated.


- Mayank Ojha

[ghorrocks]

I said either incorrect or not accurate. If you are sure your implementation of the equation is correct then your approach is not accurate. Numerical accuracy is a very different thing.

I suggest V13 as it has several methods of doing exactly what you are doing built-in. As it appears the implementation you have done is not accurate, the different implementation by ANSYS may be accurate.

If your university has up to date TECS/leases then you are entitled to V13. You have paid for software you have not got around to installing - this sounds like waste to me. You could even install CFX V13 on a machine you have access to yourself just to test whether V13 fixes your problem.

[vmlxb6]

Originally Posted by ghorrocks

I suggest V13 as it has several methods of doing exactly what you are doing built-in. As it appears the implementation you have done is not accurate, the different implementation by ANSYS may be accurate.

Hey Glen, I have got the V13 installed on my m/c. Now I ran a bunch of case on it using the Rigid Body Solver. I tried from tstep=0.01,0.001 & 0.0001. My vortex shedding time period at re=100 is 0.14 [s]. Even if I take 1% of the tstep its 0.0014. But for all three tsteps I do not match the Displacements nor the freq.

Question: Should I try reducing my tsteps ? I think its already way too low. Going even lower doesn't make any sense.

When I ran a turbulence model (SST) just to check my y+, it was 0.06 which is super low. Is it possible to have problems because of very fine mesh.

Originally Posted by ghorrocks

such as the immersed solid approach

Documentation say that immersed body solvers should not be used when the simulation requires accurate boundary layer prediction.

Is there any thing that I can do about this ????? Please HELP!!!!!!!

[ghorrocks]

Should I try reducing my tsteps ? I think its already way too low.

Why is it too low? This is a common mistake. If the result is approaching the correct value as you decrease timestep then I would keep going smaller.

Is it possible to have problems because of very fine mesh.

Yes, especially when you have a very fine mesh which expands to a much larger mesh. This can be helped by using double precision solver.

If you do not like the not-as-accurate boundary layer approach used in immersed solids then use the moving mesh rigid body solver approach. It will be far slower and if the motion is large you will have to be careful to not fold the mesh, but you can retain a good boundary layer mesh.

[vmlxb6]

Hey Glen,

When the Rigid body is defined and we define the Spring constant and external forces should the gravity term be defined along with it ??? The solver theory guide: Ch-9 Rigid body theory (pg 342), It is written that the eq of motion is written as force equal to mass into acc. [mx"=Faero+mg-Kspring(x-xso) + Fext] here the mg force is considered automatically or should we define the gravity term ???

In my case should I be defining the gravity term ???

[ghorrocks]

If gravity is a significant force on the body then yes, include gravity.