I am looking at flow past a cylinder (in 2D) using Fluent 4.4.8, and I am finding that I am unable to get the correct shedding frequency (which I am measuring, by considering the changes in velocity observed at a monitoring point situated in the wake). While I am getting shedding, which has the correct spatial wavelength (ie. the distance between the successive waves on a streamline plot looks to be about 5 cylinder diameters) and the flow field looks correct, the period of vortex shedding reported via the monitoring point is much to large (by a factor of between 4 and 5). At first I thought it was a scaling problem, but using the same grid (built using geomesh) with FLUENT UNS, gives the correct period and hence Strouhal number. I have discussed this with another user of Fluent, and when he compared some of his results for a completely different problem with those of other researchers, he found that his results had to be evolved for 5 times as many time units in order to obtain a similar result. So essentially I am asking if anybody has either experienced this problem, or has any ideas as to what is wrong.

Could you tell me the Re-number in the simulation?

(1). change the numerical scheme, check the input, run another case, (2). change the inlet condition slightly, say 1%, check the input, run another case, (3). change the mesh density only slightly, check the input, run another case, (4). if you are running a turbulent flow case, change the turbulence model, check the input, run another case. If all the solutions are about the same, then, the code is consistent. ( you can trust it) If you think FLUENT UNS is better, then use it. If you still have problem, then ask the support engineer to supply you a complete set of the input for the flow over a 2-D cylinder to see whether you can "duplicate" it. ALLWAYS CHECK OUT THE BLACK BOX USING A TEST CASE TO SEE WHETHER YOU CAN "DEPLICATE" THE RESULT AGAINST THE DEVELOPER'S RESULTS, BEFORE YOU EVEN START RUNNING YOUR OWN PROBLEMS. (A BLACK BOX IS A BLACK BOX! NEXT TIME WHEN YOU RUN THE CODE MAKE SURE TO DIVIDE THE PERIOD BY A NUMBER BETWEEN 4 TO 5. THIS WILL SOLVE YOUR PROBLEM. ASSUMING THAT THE RESULTS ARE REPEATABLE AT ANY TIME.)

The distance between successive wakes divided by your "freestream" velocity will give the approximate time period. If the value you calculate by other means differs greatly then you can be sure where the error lies! Almost any prediction will produce a reasonable result for the shedding frequency so long as vortices are shed. The frequency is fairly insensitive to numerical/modelling errors but the magnitude is another matter. The only other suggestion I can think of is to look for a missing pi (i.e. what are the units for your frequency?).

Hi Paul, There is some literature about the formation of secondary frequencies in the wake of a circular cylinder with vortex shedding. The main point is that when boundary conditions are imposed on the variables (rather than on the characteristics of the flow) then the boundary becomes (partially) reflective and this (resonnance) phenomenon can induce additional frequencies in the flow. The main reference to this is: Abarbanel, Don, Gottlieb, Rudy and Towsend, J. Fluid MEch., vol.225, p.557. So this might be another direction to look for the difference in frequencies. Cheers. Patrick.

For those that asked the Reynolds number being considered is 400 (Based on diameter).

Thanks for the suggestions, I have tried changing the scheme, but get the same results. I have tried using different grids but again the same thing happens. I am looking at flow for a Reynolds number of 400, so I am not using any turbulence model. I would use FLUENT UNS, but I want to look at free surface flows, and FLUENT UNS can't model them. I will try modifying the inlet boundary condition and see what happens.

Thank,

Paul

There is one thing you can check, that is, run a steady-state solution on both codes at a lower Reynolds number ( at a Reynolds number where you are sure that the steady -state solutions exist) and check the results to see whether you are getting the same steady-state solutions. (you are doing fine. It's like detective work. It's fun.)

Thanks for your comments, when I divide the the distance between the the vortices by the free stream velocity I do get the correct period. That is what I was suggesting when I mentioned that the distance between the successive waves on a streamline plot looks to be about 5 cylinder diameters, and since the free stream velocity is 1.0, the diameter is 1.0 and the Reynolds number is 400, the Strouhal number should be roughly 0.2, so the period should be 5 . It tends to suggest that perhaps there is either something wrong with my use of the monitoring point, or something wrong with the time reported by fluent in the time history file.

Thanks, I will check it out.

[rishitosh]

hiii..

i m also new in dis CFD area...
i m working on flow over two side by side cylinders... @ Re= 100 and 200..
i got good results..

but i want to find out Strouhal number for all cases...

i came to know by finding shedding frequency with help of FFT, using cofficient of lift data, in fluent 13.0 we can calculate Strouhal number...

but i m getting helpess..

how to use FFT to find shedding frequency....?????
-Rishi

[cfdnewbie]

do a temporal FFT of the vertical / orthogonal to inflow and cylinder axis force. that will give you the Strouhal number

[rishitosh]

hiii cfdnewbie....
thnx for response..

but i dont know hw to do temporal FFT of the vertical / orthogonal to inflow and cylinder axis force.. ???

can u guide me plz... ???

-rishi

[cfdnewbie]

record the force component over time, sample a sufficiently high number of cycles and the do the fft to find the dominant frequency! what is still unclear?

[rishitosh]

hii cfdnewbie...
thnx again 4 showing interest...

i hv recorded Cl (coefficient of lift) over time for high no of cycles (60 cycles) in xy file format..
den.... i m following dis step... record>plot>FFT and here i m loading dis file...
den checking FFT...

is it correct way..??? i m not getting desired result..

-Rishi

[cfdnewbie]

Originally Posted by rishitosh

hii cfdnewbie...
thnx again 4 showing interest...

i hv recorded Cl (coefficient of lift) over time for high no of cycles (60 cycles) in xy file format..
den.... i m following dis step... record>plot>FFT and here i m loading dis file...
den checking FFT...

is it correct way..??? i m not getting desired result..

-Rishi

Just load the file in Matlab and do the FFT there (or on octave). Or just plot it over time and determine the frequency by eye. Should be very easy to do, and exact enough to check your result.

[rishitosh]

hiii.
thnx again..

yes, i hv done just by seeing the plot..
Strouhal number(St) comes approx 0.2 for all cases..
but i hv to give precise value of St for all d cases to validate my results...