[fluidman]

I'm trying to simulate vortex shedding in a sphere at Re 600, but without success. I am using PISO pressure velocity coupling.

Also, i tried providing an initial perturbation to the flow by putting in a patch with a certain velocity in some part of the flow domain, but it seems to work only in the case of a 2D cylinder and not for a sphere.

I've been stuck with this problem for 2 weeks. Please help!

[DonQuijote]

Hi,

What error do you have? How is your domain?? Do you have a fine mesh? What are your boundary conditions?

[fluidman]

hey, thanks for replying.

It just converges after a few iterations and the value of Cd and Cl becomes constant. I have a cuboidal flow domain with a very fine mesh near the surface of the sphere and a coarse mesh farther away.

The boundary conditions i'm using are velocity_inlet, pressure_outlet and symmetry.

Hope that answers your questions.

[DonQuijote]

but if your problem converges, what is the problem you have? Otherwise, it is a good idea to make a fine mesh back to the sphere too, since fluent will solve better the sphere trail.

[chris85]

Ok havent got info on this till I look at one of my books but is vortex shedding not first observed at a higher Re number of 800. This is just off hand my thinking but havent got any literature on me to confirm.

[fluidman]

For a case where vortex shedding occurs, the solution should not converge and the value of Cd and Cl should oscillate. I tried this for a cylinder and that is what happened.

Also, most sources report that vortex shedding in a sphere begins at < Re 250.

[amin66]

hi
i worked in this problem before. you can use this article:
" Simulation of Flow Past a Sphere using the Fluent Code" copy this in google and download. hehe
i couldn't reach vortex shedding at that time and i leaved that. so if you can reach the good result please inform me.
best regards

[fluidman]

I have already gone through it and got no vortex shedding using the exact methodology that they have suggested.

Even the vortex shedding i obtained for a cylinder was accomplished using PISO instead of the SIMPLE that they've recommended.

[amin66]

yes PISO is more efficient for transient simulations. i will work on sphere next week again and if i reach good result then i will tell u

[scyllakeeper]

was anyone able to solve this problem?
i am facing a similar situation myself

[Far]

this should not be problem. If you setting up problem correctly.

1. Mesh

2. Time step

3. Boundary conditions and material properties.

[scyllakeeper]

I believe I have set it up all correct.
I'm trying to simulate vortex shedding at Re 300 for a 3D sphere, enclosed within a cylindrical domain.
Been stuck for over 2 weeks.

I am not getting the oscillations for the drag and lift coefficients that should be there. It all settles to a straight line as the solution proceeds.

Any suggestions?

[scyllakeeper]

Added info:

Dia for sphere: 0.1cm
Domain: D=15cm, Lu=10cm, Ld=20cm.
Mesh is as fine as it can be.
Making use of Pressure based, transient, viscous laminar analyisis.
Density=1000, Viscousity=0.001.
Boundary Conditions: Velocity inlet=300e-3; Pressure outlet=0 gauge, No-slip at sphere wall, and zero shear at domain wall.
PISO solver: Skewness Correction =1; Neighbour Correction = 1; No Skewness-Neighnour Coupling; Least Square Cel Based gradient, Second Order pressure, QUICK momentum.
(To add: I've pretty much tried all the variations in the solver settings here.)
Default solution controls.
Time step size = 0.02
Max Itr per Time step = 30

Any advice?

[Far]

how did you calculate that time step?

run simulation for long time.

[scyllakeeper]

I got my Time Step be making use of definition of Strouhal Number.
The Strouhal number for flow past Sphere would be between 0.16 to 0.18. (Found this in a research paper; please correct me if this is wrong.)

If my oscillations are dying as the solution proceeds, and I see no sawtooth curve, I don't think running the simulation for longer time would help.
Would still give it a try.

[Far]

Actually in transient CFD you have three regions:

1. Initial transients due to numerical model kick starting

2. Solution settling time aka steady state region

3. True transient region

Last region comes after a very long time and which needs patience.

Here I have included an image to explain these three regions. Not made by me, taken from internet.

[scyllakeeper]

Alright. Will give it a shot.
Thanks! Much appreciated.

[Far]

how many time steps did you use to divide one cycle time period?

[scyllakeeper]

Time step size = 0.02s
Max Itr per Time step = 30

Is this what you're asking?

[scyllakeeper]

My simulation has been running for the last 20 hours.
For now, the value for Cd is not oscillating.
Keeping the simulation running to see if any changes occur.