CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   OpenFOAM Running, Solving & CFD (http://www.cfd-online.com/Forums/openfoam-solving/)
-   -   Convective schemes for vector fields (http://www.cfd-online.com/Forums/openfoam-solving/57951-convective-schemes-vector-fields.html)

 edoardo February 9, 2009 14:11

Hi, I'm testing convective sch

Hi, I'm testing convective schemes in OF in LES calculations. In particular I'm comparing results from different convective scheme configurations in a case of 2D LES of flow past a square cilynder. I read in User Guide v 1.5 there are some variants of classic convective schemes for transport of vectors. I got very different results (FFT of pressure signal reported by a probe downstream of the cilynder) from some scheme and its V version counterpart (applied to div(phi,U) term): in particular SFCD, limitedLinear and vanLeer.
Strangely results from V schemes are worse: pressure amplitude spectrum doesn't present a clear peak at vortex shedding frequency.
Did anybody face this particularity? Does anybody have some suggestion for a best LES convective scheme?

 kev4573 February 9, 2009 16:03

You could try looking the pres

You could try looking the pressure lift coefficient on the cylinder instead of probing for pressure - might give more consistent results to use for comparison between the schemes. That's the way I've done this kind of thing in past with a RANS solver and gotten very good results when comparing the experimental and computational Strouhol numbers.

 maruthamuthu_venkatraman February 10, 2009 05:56

Hello Kevin, Pre

Hello Kevin,
Presently i am trying to predict the vortexsheddding behind cylinder for Re =10^5. I couldnot get a steady oscillation for both Drag and Lift even after 20 periods. The oscillations are very irregular.

Lift frequency is better than Drag. Actually drag period should be twice the Lift period as obtained from Str Number, but i couldnt see a regular pattern.

COuld you post the schemes that you have used and solver settings. It take lost of time to test individually all schemes. If you have good experience with this problem and schemes, then i can use ur settings.

Note: Yplus is almost 0.1 in my case , using Launder and Sharma low REKE model.

Thanks

 kev4573 February 10, 2009 11:42

Hello, I just used upwind s

Hello,
I just used upwind schemes on the div terms and the defaults for the solvers. That Re number seems a little high, think you need something on the order of 10^2.

 maruthamuthu_venkatraman February 11, 2009 03:23

But i need to predict the shedding frequency in that order . I do realize in laminar flow its quite easy to capture the shedding frequency than turbulent flows.

May be some other Foamers can recommend me for the schemes to capture vortexshedding behind cylinders at Re 10^5. my intentions are once i predict them in simple 2D model, then i will change it to LES in 3D .

First i have to capture it in 2D model. i have changed into vanleerV sheme for Div(phi v)and for time cranknicholson 0.5. Noe its running. Let me see the behaviour for Drag and Lift ...

 edoardo February 16, 2009 09:47

Hi Maruthamuthu. I'm running

Hi Maruthamuthu.
I'm running LES of flow over a square cylinder. I'm running a test case analogous to the one reported by J.S. Ochoa and N. Fueyo in their article "Large Eddy Simulation of the flow past a square cylinder" (Smagorinsky for eddy viscosity model and wall function for wallpatches) . As you could read from the article Re is just 21400.
My intentions are predicting vortex shedding frequency on a 2D case through a systematic test of convective schemes present in OF, then with the best one runnig a 3D case.
I used a pure Crank-Nicholson (cranknicholson 1.0); I tried the following convective schemes: SFCD, SFCDV, CDS, LinearUpwind, VanLeer, VanLeerV, Gamma and GammaV (with coefficient 0.2, 0,5 and 1), LimitedLinear and LimitedLinearV (1.0). I extracted Cdrag and Clift from results, actually just GammaV02 and GammaV05 (corresponding to 0.1 and 0.25 parameters Betam in "Error Analysis in Finite Volume CFD" in Franjo Juretic's thesis) gave me good results for Strouhal number (0.131 and 0.117 against 0.133 experimental Strouhal). All of the tested schemes overpredict time averaged drag coefficient (2.3:2.9 against 2.0); hope in better agreement with 3D simulation.
I reported a different result from yours about drag and lift frequencies: drag period is about half lift period.

 maruthamuthu_venkatraman February 16, 2009 12:05

Hello Edoardo, Bigger pardon

Hello Edoardo,
Bigger pardon in saying drag period as twice the lift period. You are right , Drag should be half the lift period. It was written Mistakenly...Its good you corrected in this Forum!

Iam not convinced with the results yet. Particularly for frequency predictions. I have maintained my reynolds number but played with velocity and viscosity , now the coeffcients seems to be close to Okey. But the frequencies are not in order for the above schemes.

So i have changed in to CN 1.0 for complete 2nd order accurate for time scheme.Hopefully i will get some good results tomorrow.

Also i havent adventured in to resolving the turb boundary layers by using High RE models. Still iam working with Low Re ke model and resolving the boundary layers..

If its doesnt work then i will follow the suggestions that you have given me for convective schemes...

Will update you soon..

 All times are GMT -4. The time now is 10:37.