URANS vs LES for Acoustic Simulations
 

Hi Everyone,

In a case where pressure waves do not propogate from a turbulence source (no quadrupole), the only way is to reflect from solids. Thus, no need to solve turbulence as LES does.

In this case, is URANS solver able to capture the pressure waves at desired frequencies with a fine mesh (e.g. 20 elements per wavelength) and time step size (sufficient for the frequencies).

Is there any drawback using URANS for acoustic simulation cases like this ?

Looking forward your comments.

Thank you.

You can simulate propagation of acoustic waves using any unsteady solver, even unsteady laminar as long as you are using an ideal gas equation of state (or some kind of compressible medium).


Without knowing the details of the flow you are trying to avoid solving, we cannot say more about what are the drawbacks.

If there is no turbulence, you have to use the NS equations without any statistical averaging or filtering.

When you are using RANS for acoustic analysis, you can only infer the broadband acoustic quality from the TKE, for instance, with the Lighthill relation. You can probably estimate the general loudness of the turbulence generated, but it is often not very accurate and cannot be used to compare flow noise between different systems.

A more comprehensive way is to resolve the pressure fluctuation directly, either using LES or DNS and obtain the frequency spectrum, which is more representative, particularly for human perception.

Gerry.

Just to clarify, the issue is not only in using RANS, but also in the adopted formulation. Resolving the acoustic field means you solve the compressible form of the equations, otherwise, for the incompressible model, also using DNS/LES you have to couple the resulting velocity field with the Lighthill relation

Yes, you are right. The fact that the governing equations are (weakly) compressible should be implicit, if you want to solve for pressure fluctuations.

Although the Lighthill relation is a diagnostic relation so you could calculate this as a post processing step. Since it is only a function of the TKE you can apply this also to incompressible flows.

I think the problem with (U)RANS is that since it is time averaged, you lose the ability to capture the acoustic spectrum, as the time averaging serves as a ripple filter in the frequency domain. Thus you can at most perform broadband acoustic analysis, where contributions from all frquencies are assumed to be equal.

These pertain only to the acoustic spectrum generated by turbulence which the original question specifically excludes. I'm not sure why we keep bringing it up.

Actually the problem includes acoustic reflections between blade rows of multirow compressor blades due to the vibration motion of the rotor. The acoustic reflections from upstream and downstream stationary parts may trigger flutter instability on rotor. That is why I need the solve acoustics.

Dear Boris:

I assume that you are also simulating the flutter (using fluid structure interaction, for instance) or at least parameterizing the degree of fluttering using the pressure distribution on the turbine blade? This part I don't know very much and there are others in the forum who is in a better position to help you.

As Tran pointed out, I realized that it is not a turbulence flow noise problem, so you can throw the turbulent fluctuation stuff out of the window. This does not mean that the modelling of turbulence is not important. Particularly, if you expect separation on the blade, URANS might not give you an accurate separation point, and your results could be very much off.

It is probably best to run a steady case, for which you know where the separation point should be, so that you have a good idea whether the turbulence model you use could represent your case reasonably well.

Of course, I would go with LES if you could afford the computational power, which would give you a better estimate for this.

Gerry.

My question is: what do you know to say that there is no turbulence influencing the acoustic field?

I think what he meant is that the acoustic disturbance caused by fluttering presumably from flow expansion / compression dominates and the turbulent contribution can be ignored.

Interesting, how can you be sure the contribution from quadrupole term is negligible in such a kind of complex flow?