AeroAcoustic Simulations
 

Hi,

I am currently trying to find out all that I can about what it takes to carry out good noise/aeroacoustic simulations. In particular we are interesting in predicting noise from internal flows such as from fans.

I have done a lot of searching on the www but have only had limited success in finding good information on this subject. In particular, I am trying to answer some basic questions such as:

What are the differences between the Ffowcs Williams – Hawkings, Kirchhoff and Curle formulations? Are some of these methods better suited to different types of applications?

Are some of the methods not applicable to internal flows? (It appears that much of the research has been carried out to study noise produced by aeroplanes and helicopters).

Do you really need to use LES to predict the flow? Can useful information be obtained from transient RANS or even steady RANS simulations?

Do you really need a fine mesh with about 10 mesh points per wavelenght?

Any references to good www sites or books would also be appreciated.

Thankyou for any light you can shed on this matter.

Regards,

Kerrin Burnnand, Fisher & Paykel Appliances, New Zealand

Re: AeroAcoustic Simulations
 

Hi,

You shall have a look at our website www.mcube.fr You will find many published papers in the MCube publication section and examples in the M-Explicit applications section. (in particular there is a fan section)

Our code (RADIOSS-CFD) performs a direct compressible transient CFd simulation (LES) coupled with the structures (if any vibrating struct is involved). We then post process recorded time domain signals into the frequency domain. THe frequency range that we can tackle currently (with existing CPUs) ranges from 10 to couple 1000's Hz.

Re: AeroAcoustic Simulations
 

the lighthill, ffowcs-williams hawkings and kirchhoff (i am unfamiliar with the curle) formulations are noise propagation calculation techniques. they are useful for propagation of known noise through a known simple background flow, particularly a uniform flow. the noise generation process must be separately either modeled or captured.

one canonical problem for noise generation by internal fans is the much-studied BVI (blade-vortex interaction) problem. generally, the euler or linearized euler equations are solved with mesh resolution suitable for gusts of known frequency, such as the frequency of motion of rotor blade wakes relative to the stator blades.

the physics of the more complex generation of fluid self-noise is less well-understood, and is being actively researched. for example, self-noise of turbulent flows, where the noise arises from interior turbulent eddies rather than at the fluid-wall boundary. for some turbulence-noise problems, one approach is RANS, followed by semi-empirical models relating turbulent eddy viscosity and kinetic energy to generated acoustics, followed by propagation of acoustics to farfield using kirchhoff or similar technique. chris tam has a particularly successful approach in this regard to self-noise of round jets.

a more fundamental approach is the capturing of self-noise generation by LES. of course, it is correspondingly more expensive. bailly et al in france, lele and students in california, freund in california are some of the researchers with recent work in LES for aeroacoustics.

after looking at the range of approaches, you may decide for yourself the strengths and weaknesses of each. you may find it useful to look at some benchmark problems in CAA. the 4th CAA workshop on benchmark problems will take place in october of this year. you may search for websites or journal papers of work by members of the scientific committee of the workshop. also, look for proceedings of the three earlier workshops. there have also been specialized workshops on fan noise. the website for the workshop announcement and problem specification is

http://www.math.fsu.edu/caa4/index.html

(that's at florida state university, where chris tam and some other acoustics researchers work). the best introductory book on acoustics that i am aware of is Fundamentals of Physical Acoustics by David Blackstock.

Re: AeroAcoustic Simulations
 

Concerning books on aeroacoustics. Tam has written a new book on computational aeroacoustics - it should be a great reference for anyone in this field. You can pre-order it at amazon here. I ordered it recently and estimated delivery time was in early may.

Re: AeroAcoustic Simulations
 

Thanks Dimitri, Ananda and Jonas for your responses. I am especially interested by the computational aeroacoustics book by Tam.

However, there were still a few things I was trying to figure out:

(1) I am trying to get a feel for the computational expense of aeroacoustic noise simulations. Does anyone know if it *REALLY* necessary to carry out a LES simulation for accurate results? Are RANS/transient RANS simulations of any use?

(2) Do you need to use different methods for internal flow problems than for far field external flow problems (eg noise from a jet)?

Thank you again!

Kerrin Burnnand Fisher & Paykel Appliances (New Zealand)

Re: AeroAcoustic Simulations
 

kerrin,

i do not know the answers to your questions.

my best guesses, based on the little that i do know are as follows. (steady) RANS does not directly yield the acoustic information, which by its nature is unsteady. if the acoustics is self-noise associated with turbulence, then some kind of empirical correlation is necessary between the calculated turbulence statistics and the radiated noise. it seems to me that such correlations are nowhere near universal. individual companies or research institutions have built up such correlations, based on available experimental data. chris tam has in recent work demonstrated particularly successful modeling of this type for fine-scale turbulence noise from jets. i too, am glad to hear from jonas that tam has written a book on computational aeroacoustics.

if the noise is not dominated by turbulence-caused high-frequency sound, but turbulence is present, then an unsteady RANS approach may capture the sound, provided the mesh is fine enough. a VLES (very LES) approach may capture noise due to the large-scale coherent structures in turbulence, which will be adequate if the fine scales do not contribute much noise. but for any turbulence-related noise in general flows, or for broadband shock noise, my guess is that yes, you do require LES to capture the noise generation. then lighthill or ffowcs-williams hawkings or ... to propagate sound to the farfield. tone-noise such as some fan noise and screech of jets, can be done without meshes as fine as those needed for LES.

with internal flows, my guess is that you do not have to propagate the sound to the farfield, which saves a lot of computation. pressure readings within the channel may provide figures of merit for the acoustics. other than that, i expect internal flow noise computations to be similar to the external case.

Re: AeroAcoustic Simulations
 

Thank you Ananda. I think I will try and get a copy of Tam's book.

Kerrin

Re: AeroAcoustic Simulations
 

No, it is not compulsory to do LES but there are some advantages to do so: (1)If you use a steady RANS solution, you need to make assumptions to derive the noise field out of a frozen image of your flow. Using a transient method is better in any case. (2)If you try to catch frequencies higher than 1000Hz, you indeed will need small elements and small time steps. Our experience is that these will be practically small enough to be able to carry out a LES simulation.

Regarding the internal flow problems, the critical issues will be to have a good acoustic boundary at the inlet/outlet, with an user defined cutoff frequency and if the structure is vibrating and/or if you are looking forward to capture the noise outside the system, to have a fluid strucutre coupling (including the capability to give an acousitc impedance BC on the outer surfaces of the structure). The good things is that models are usually physically smaller hence you can go up to 5-10kHz. For the external flow, the main problem is the computatinal domain size. For propagation you need to have about 10 elements per smallest wavelength hence if you want to capture 5000Hz, the coarsest mesh you can use (propagation zones) is ~.7cm !!! Practically, this is usually not achievable so you have either to limit your domain and use pure propagation schemes between the domain and the measurement locations, either to use a 2 steps approach, mixing transient LES CAA simulation in the noise source zones and Steady RANS in the far field.

Regards,

Dimitri

Re: AeroAcoustic Simulations
 

Dear Dimitri,

Thank you for you reply. Would it be possible for you to direct me towards books and/or papers which describe how noise predictions can be made from steady state flow simulations as you mentioned? Although this is no doubt a far more approximate method than an unsteady (eg LES) approach, I would like to try it to get a feel for how good or bad the results obtained are.

Also, are you aware of any sort of best practice guidelines for predicting aeroacoustic noise with CFD?

Thank you,

Kerrin Burnnand Fisher & Paykel Appliances, New Zealand

hey can I get your help on predicting of fan noise using cfd please.

kindly regards
Fares