2D Transient flow around a cylinder
Hi, I am trying to simulate the vortex shedding process behind a cylinder but have been unable to do so. Just wondering if anyone know how to solve the problem?
I've set up the problem as a very thin plate with symmetry on the top and bottom side of the cylinder. I am solving it using the LES Smagorinsky model, and have also applied wall roughness to the cylinder. In the workbench mesh, I have also set up 5 boundary layers with first thickness of 5E-8m. I've run several simulations, some running over 24 hours but have been unable to produce vortex shedding. Anyone encountered the same problem or know how to fix this? Any help will be greatly appreciated. Cheers, Ben |
Re: 2D Transient flow around a cylinder
Hi,
You can't run LES in 2D - but it should shed vorticies regardless. What Re are you modelling? You will need second order differencing in time and space. How many flow time scales did your 24 hour go for? A small asymmetry should start when the flow has travelled about 1 diameter or so, and build up into a vortex as the flow progresses. Glenn Horrocks |
Re: 2D Transient flow around a cylinder
I'm looking at a cylinder in 3-D with the LES model.
The fluid domin is rectangular (length=60m, height=width=20m) and the cylinder diameter=1m and sapns the domain. The centre of the cylinder is 10m downstrean of the inlet of the rectangle. At Mach=0.3 and ISA sea-level gives a Reynolds no.=7million. I want to try the "define first layer by y+" option. I've set it to y+=1 so a 1st layer is 3.9x10-6 and I've put in 20 layers at 1.2 expansion factor. Not sure if this is a good y+ value. With more layers the memory was maxed out. I'll have to set up a few Controls at get more element detail in the wake and around the cylinder to merge neatly with the large elements on the domain boundaries. Is this good enough for the boundary layer? Thanks. |
Re: 2D Transient flow around a cylinder
Cheers for your post Glenn,
My Reynolds number is in the order of ~50000, and I have used physical timescales of 0.02seconds. I stopped the last simulation at 0.6seconds. (I have a flow velocity of 20m/s so by 0.6 seconds the flow would have travelled more than 1diameter down the cylinder) Yes, I have used second order backward euler for my transient scheme and a specified blend factor of 0.8 for my advection scheme. Glenn, is there a way of introducing flow perturbation in ANSYS-CFX? I've just read somewhere on another forum that this might have to be done to induce vortex shedding..... When you mentioned that LES can't be run in 2D, should I used k-epsilon model instead? Thanks for your help! Ben |
Re: 2D Transient flow around a cylinder
|
Re: 2D Transient flow around a cylinder
Hi,
Yes, for 2D you should use a standard RANS turbulence model. SST or k-eps are typical choices. When you go 3D you should consider the SAS model. It should be ideal for these types of flows. Generally you don't need to perturb the flow, enough numerical noise exists to get it to start. Regards, Glenn Horrocks |
Re: 2D Transient flow around a cylinder
Hi,
Meshing to y+=1 at such a high Reynolds number is difficult. Sounds like you should consider the SAS turbulence model. Glenn Horrocks |
Re: 2D Transient flow around a cylinder
Do u have these files of that simulation (http://br.youtube.com/watch?v=jMmPYuKqKCU)? (.blk, .tin, .ccl, and .cfx files)
very nice this movie, congratulations! |
Re: 2D Transient flow around a cylinder
Thanks Glenn I'll try that aswell. Just sort of trying different options in CFX.
How do you work out an appropriate y+ value? I've seen the equation but to me it seems like guess work/experience. |
Re: 2D Transient flow around a cylinder
Het, that's a great video! Cheers for that!
|
Re: 2D Transient flow around a cylinder
Yeap, finally got it in the end. Simply changing from LES to k-epsilon and cutting down the timestep solved the problem. Cheers for your help!
Ben |
Re: 2D Transient flow around a cylinder
Hi,
Only mesh to y+=1 if you need the sublayer or transition models. For most high Re flows normal wall functions are as good as anything and allow you to use y+ from about 15 to 100 or maybe higher. You have to do a sensitivity check to find out for your case what mesh you need here. Glenn Horrocks |
All times are GMT -4. The time now is 08:29. |