[Panboy]

Hey Guys,
I have recently discovered this forum and so much good information. I am studying vortex shedding, and I believe I have read ever post on it, but I still have a few quick questions.

I have simulated a 2d cylinder at low reynolds number with results of vortex shedding as a lesson for myself.
I have now moved on to a 2d profiled body with a blunt trailing edge. Pretty much a flat plate with a smooth leading edge.

I can not for the life of me get vortex shedding on this body.

I am simulating Re 500. The body is .0254m thick and .3175m long.
I have refined my mesh many times and I believe I have a very fine mesh with 110,000 cells, but that might not mean much since I am new to CFD/fluent. its orthogonal quality is .85, and its all hex mesh. with biasing towards the body walls and wake area.

I have tried both starting it straight from transient, and also starting as steady, and using those results for initial values for transient solution.

I have tried to run it both with and without turbulent models
I have also attempted simulations in which i patched the y-velocity in order to help start the shedding.

I have used SIMPLE, and PISO models, and am using second order for both spatial and time discritization. I have sometimes attempted to use the PRESTO model a few times as well.

my time step size is .0025s and I am iterating 30 iterations per time step.


My main questions are:

1. The main problem is that I can't get my C_L to oscillate. It either goes straight to 0, or it oscillates a few times before it damps out to 0. Is this a common problem for a silly mistake that I am doing?

2. For the reference values, I am setting my surface area to my chord length, and the length value to the chord length also. Is this correct?


Thanks so much to anyone who can throw me some ideas.

[Panboy]

I wish I could help but I don't know

[LuckyTran]

Re = 500 is that based on chord length? So it is a laminar flow? Wall boundary condition is no-slip correct?

Are you having any convergence problems? 30 iterations per time-step, was your solution converged at each time step?

What's the velocity? Is your time-step small enough?

Are you sure your mesh resolution is good enough? How many cells across a vortex?


Reference values should be arbitrary (up to you). Depends on your definition for the coefficients, use the appropriate reference value. That said, wouldn't the plate thickness be more appropriate as a reference area? Anyway it is only a matter of reference.

[Panboy]

Re500 is based on chord length. I do believe its laminar, but according to an article i read, the point at which it turns turbulent is around 570.

I didnt specifically tell that the wall is no-slip. Its just a "wall" B.C.
I'll double check that, thanks.

I am not having convergence issues. It is easily reaching 1e-6

I am not sure how many cells across a vortex. I dont know how big the vortex is. I am currently refining my mesh even more right now, as I was comparing it with Cornell's CFD tutorial unsteady cylinder mesh and there's seemed much more refined near the sphere. I am hoping this is the problem.

If I did plate thickness as the reference value. Wouldn't that make my C_L pretty much a C_T value since it would be against the incoming air?


Thanks for the good questions,
Panboy

[LuckyTran]

Originally Posted by Panboy

Re500 is based on chord length. I do believe its laminar, but according to an article i read, the point at which it turns turbulent is around 570.
Panboy

Ok just double checking to make sure what reference length you were using. In that case just use laminar and do not even bother with turbulence models (you should still get the vortex street).

as an aside, the transition Reynolds number for a flat plate is 500,000 ish, and I imagine it would be similar for a plate with a rounded nose. Are you sure on this 570? Anyway, it should be immaterial to your simulation.

Originally Posted by Panboy

I didnt specifically tell that the wall is no-slip. Its just a "wall" B.C.

that is no-slip, again i was just making sure

Originally Posted by Panboy

I am not sure how many cells across a vortex. I dont know how big the vortex is. I am currently refining my mesh even more right now, as I was comparing it with Cornell's CFD tutorial unsteady cylinder mesh and there's seemed much more refined near the sphere. I am hoping this is the problem.

Obviously if the grid resolution is very poor then you would not be able to resolve any flow features. The mesh at the leading edge can definitely influence the solution since a separation bubble must develop just after the leading edge which starts the vortex street. Check your solution at each timestep to verify that this separation bubble is present, otherwise you will not get your vortex shedding.

Originally Posted by Panboy

If I did plate thickness as the reference value. Wouldn't that make my C_L pretty much a C_T value since it would be against the incoming air?

Nevermind the plate thickness. I just double checked and the most commonly used definition of drag and lift coefficient is to use the planform area which is the chord.

[Panboy]

Originally Posted by LuckyTran

as an aside, the transition Reynolds number for a flat plate is 500,000 ish, and I imagine it would be similar for a plate with a rounded nose. Are you sure on this 570? Anyway, it should be immaterial to your simulation.

Hmm, you are right
From a research article from Petrusma and Gai, 1996, they have a plot that compares Strouhal# vs Re# for cylinder and profiled body with blunt trailing edge. It states that at 570, there is transition to turbulence.
Thats confusing, I will look into this.

http://i854.photobucket.com/albums/a...rms/StvsRe.jpg

Originally Posted by LuckyTran

Obviously if the grid resolution is very poor then you would not be able to resolve any flow features. The mesh at the leading edge can definitely influence the solution since a separation bubble must develop just after the leading edge which starts the vortex street. Check your solution at each timestep to verify that this separation bubble is present, otherwise you will not get your vortex shedding.

Thanks, that is very true. Would I check for this separation through velocity and pressure plots? Most likely a large drop in velocity and large drop in pressure?

[LuckyTran]

Originally Posted by Panboy

Thanks, that is very true. Would I check for this separation through velocity and pressure plots? Most likely a large drop in velocity and large drop in pressure?

It will be obvious in simple velocity vector plots (or even better streamlines). Shouldn't be hard to find a separation bubble, look for the reverse flow under the bubble.

[Panboy]

Originally Posted by LuckyTran

It will be obvious in simple velocity vector plots (or even better streamlines). Shouldn't be hard to find a separation bubble, look for the reverse flow under the bubble.

Ok , I have refined my mesh and I definitely get separation at the leading edge.
However, I am still getting no lift oscillations nor any vortex shedding formations. Risiduals look good and drop below 1e-06.
I tried both starting from a steady state solution, and by starting transient but perturbing the top portion of the wake area in order to influence the vortex formation but it still drops to a steady flow at the wake very quickly....

I am not sure what could be wrong. I do the same solver setup for a circular cylinder and i get vortex shedding, but not with this profiled body w/ blunt trailing edge.