|
发表于 2002-5-1 09:21:44
|
显示全部楼层
请教各位大虾: 风机、流场、噪音
补充:
尽管模拟噪音很困难,但计算气动声学发展也很快的.有些软件宣称能模拟风机噪音的.
现从cfd-online上摘录几篇,供参考.咱也通过你的提问了解了一下行情,谢谢.
***1. We have recently coupled standard CFD codes with our computational aeroacoustics module. You will find information (papers, animations) in my web page "http://www.gre.ac.uk/~k.pericleous/".
***2. PAM-FLOW from ESI Group is a finite element based explicit solver that is used extensively for aeroacoustics simulations within the automotive industry. PAM-FLOW can handle both low frequency and high frequency phenomena.
Proven automotive applications include sunroof and side window buffeting, wind rush noise due to side mirrors, A-pillar, wipers, roof-racks etc., and internal components such as intake manifolds, engine cooling fans and HVAC systems.
We provide an integrated auto-mesher that facilitates boundary layer meshing, even for ultra-complex geometries. CPU requirements are addressed by highly scalable parallel solvers.
Please visit our website (www.esi-group.com) or contact me for additional details.
***3. Our software RADIOSS-CFD is mostly used for aeroacoustic numerical simulations. RADIOSS-CFD ios a FE code that solves the NS equations (LES turbulence) fully coupled with the structure in the time domain (explicit time integrator). We provide boundaries with prescribed acoustic impedance as well. So far, the code has been used successfully for problems like exhausts, intakes, centrifugal and axial fans, side mirrors and generally speaking components. Full car simulations require *lots* of CPU ressources. You can check our website for examples and published papers: www.mcube.fr <http://www.mcube.fr>
***4. STAR-CD offers a tool which claims to be able to do such analyses. But as far as I know this tool is "only" a post processing tool which generates pressure fluctuations based on the calculated turbulence field. May be this is sufficient to solve engineering problems like the reduction of buffeting.
***5. The Fluent6.0 version would be able to do aero-acoustic modeling. This can be done through use of udf built by FLuent. Lighthill-Curle analogy is used for noise computation in which it was not necessary to have grid upto the observers point.
***6. I'll give you three references. First, the classic Sir James Lighthill paper: M.J. Lighthill (1952). "On sound generated aerodynamically: General theory." Proc. Roy. Soc. A211, 1107. (For those who hadn't heard, Sir James drowned this year while making what was for him a routine swim around an island off the coast of England--a great loss.)
Second, a paper describing what is referred to as the "MGB method" after its developers, Ramani Mani, Phil Gliebe, and Tom Balsa: T.F. Balsa (1976), "The far field of high frequency convected singularities in sheared flows, with application to jet noise prediction," J. Fluid Mech., 330, 193.
And finally a more recent paper describing the implementation of this method using CFD results to prescribe the flow field: A.Khavaran, E.A.Krejsa, & C.M.Kim (1994), "Computation of supersonic jet mixing noise of an axisymmetric convergent-divergent nozzle," J. Aircraft, 31(3), 603.
***7. Use of CFD codes for acoustic wave propagation can be carried out succesfully. However, a very fine grid and timestep are required. Typically, with a second order code like Fluent you require at least 50 gridpoints per acoustic wavelength. This is due to the excessive dispersion and dissipation characteristics of the scheme used. Upwind differencing schemes are inherently dissipative, so very efficient at damping out acoustic waves. Not good if that's what you're trying to capture.
Also, in order to obtain a reasonable spectrum you will need to run the unsteady simulation for long enough to be able to capture several periods of the oscillation.
A MAJOR problem is that of non-reflecting boundary conditions. For an internal flow this isn't really a problem, but if any of the grid extends to freespace you will require a special non-reflecting condition. Otherwise acoustic waves will reflect from the boundary. You will not see these waves when plotting pressure contours, as acoustic waves are up to 5 orders of magnitude less than hydrodynamic pressuures. To try and get around this without specifically writing a new boundary condition, you can just put a stretched block at the outer boundary that goes to a very coarse resolution at the far edge. This will then use artificial viscosity to damp out any outgoing waves. The symptoms for this not working are that the simulation works for a transient period but then blows up.
|
|