CFD Online Discussion Forums

CFD Online Discussion Forums (
-   Main CFD Forum (
-   -   Reasons for failed validation at high frequency (

zonexo October 22, 2006 08:14

Reasons for failed validation at high frequency

i'm trying to simulate an oscillating/pitcing airfoil. I've 2 papers which does similar things at Re~1000. I've successfully validated my code against an oscillating cylinder before. Also at low frequencies for airfoil, the results are quite good. however at much higher frequencies, the results do not tally. i wonder what could be wrong.

Is there any things which need to be considered at high frequency simulation? usually, the CFL no. limits the time step. however, at high frequency, is the restriction more due to frequency? is there any rule to set the time step in this case?

Btw, I'm using a FVM fractional step code with structured grid.


vasanth October 22, 2006 15:42

Re: Reasons for failed validation at high frequenc
Does your boundary condition work well at higher frequency?

Dominic October 22, 2006 20:49

Re: Reasons for failed validation at high frequenc
zonexo : How high do you u mean ? What is your reduced Frequency value ?

Vasanth : If the domain is far enough and your mesh size is fine, what boundary condition do you think will work for a higher frequency ?

Cheers, dominic

zonexo October 22, 2006 22:33

Re: Reasons for failed validation at high frequenc
My BC are quite standard ie no slip at wall, symmetry at top/bottom of domain and convective at outlet. Do u mean there is a need to modify the BC at high frequencies?

My high and low are based on the minimum and maximum range given by the paper. Min is around reduced freq = 0.3 while high is around 3.2. Is that considered high?

The few factors i am considering are

1. farfield distance 2. time step 3. mesh size

In other words, one of them must be modified if frequency increases, although at low frequencies results are quite good.


ag October 23, 2006 07:41

Re: Reasons for failed validation at high frequenc
Have you checked to ensure that your timestep is sufficiently small to capture the physical variations in time? Your code may have no practical CFL limit, but that does not mean that you can resolve a kiloHertz frequency with a time step of 1 sec. Generally, to get good resolution of a transient response I try to get at least 50-100 timesteps per the highest frequency in the simulation.

Mani October 23, 2006 09:32

Re: Reasons for failed validation at high frequenc
I am assuming you already followed ag's advice. You did get good results with low frequencies and your time step (at any frequency) should be set to insure a minimum number of steps per fundamental oscillation period.

However, keep in mind that the temporal resolution (per period) will generally have to be larger as you increase the frequency. When 50 step per period were enough for a reduced frequency of 0.3, you may need more time steps per period for a reduced frequency of 3. Why? Because, depending on the flow, an oscillation frequency as high as 3 may result in a nonlinear response. Maybe due to shock motion, maybe due to separation, or other effects. The flow will respond at higher harmonics (multiples of the oscillation frequency) and you need to resolve those as well, not just the airfoil oscillation frequency! Increase your number of time steps and observe the solution.

If this is bad news to you, here's some more: You may actually have to increase your grid resolution as well. If you keep increasing your temporal resolution, and you get a converged result that is still far off, it may be because your grid resolution is not sufficient to capture the essential unsteady effects (like shock motion, separation...). Temporal and spatial resolution go hand in hand. Neglecting either of them will limit the accuracy of your results.

All times are GMT -4. The time now is 13:33.