k-Wave

1. Adrien_eventacoustics
   

   Member
   Joined: Sep '15
   Posts: 2

   Dear all, Bradley Treeby and Ben cox,
   I’m coming to you with a question about this great toolbox. My name is Adrien Boisseau, and I am a French student doing my final internship in the Netherlands. The project that has been entrusted to me is to make a working MATLAB simulation tool/code to simulate, in 3D, wave propagation of a directional source through different media.
   The example we are interested in is festival tents like displayed on this link: http://imagesia.com/festival-tent_11gwl

   When we have a directional source loudspeaker array propagating inside a festival/circus tent how does it affect the directivity of the sound going outside the tent through the canvas? (Here a link to a simple drawing of the problem http://imagesia.com/drawing-of-the-tent-with-a-directional-source-_11gvg )

   To solve this problem I think my best option would be to use a finite difference time domain method (FDTD) in order to model how the sound propagates through the tent layer.
   My questions are: do you think that the k-wave toolbox would be able to help me out in 2D and 3D? Will it be possible to model how the sound propagates on a long distance outside the tent (up to a couple hundred of meters)? And is the 50Hz-250Hz a range of frequency that can be studied?
   Anyway, thank you a lot for taking from your time to read this, and I hope to hear from you soon!
   Best regards,
   Adrien.

   Posted 8 years ago #

2. Bradley Treeby
   

   Developer
   Joined: Mar '10
   Posts: 1,640

   Hi Adrien,

   Just regarding the length scales, this is achievable in k-Wave. At 250 Hz in air, the wavelength is around 1.4 m. Using 4 grid points per wavelength (for example), for a 200 m domain, this gives a grid size on the order of 512^3, which is very doable.

   The main thing to consider (whether using this or a FDTD code) is how you will represent the tent walls in the grid. The thickness is presumably much smaller than the acoustic wavelength. This might require some careful thinking about some type of equivalent medium, or perhaps extensions of the model, e.g., using the immersed interface method.

   Hope that helps,

   Brad.

   Posted 8 years ago #

3. Adrien_eventacoustics
   

   Member
   Joined: Sep '15
   Posts: 2

   Dear Bradley Treeby,

   First of all thank you for your quick answer, I didn’t expect this and I’m really grateful for that.
   I have then some more questions:

   - Will it be necessary to use the elastics-dynamic model (with the pstdElastic2D code) or will it also be possible with the normal FDTD (with the kspaceFirstOrder2D code) ?

   - In the case it is necessary to use elastics-dynamic, are there optimized (parallel processor or GPU) codes for this, or planned in the near future?

   - Regarding the thin layer of the canvas, I understand the problem; the nodes are too far spaced to define the thickness of the layer through the use of nodes. Therefore a layer with the thickness of a node-spacing with equivalent properties to the layer which is a fraction of the node spacing has to be used. This could be very tricky regarding material parameters such as Poisson’ number (although I’m not sure you can define that medium parameter with k-wave) etc. since I don’t know if these scale linear.

   So maybe it would be possible to have a locally (only at and around the canvas tent) much finer grid? I was thinking of using the subdomain FDTD method, could that be a solution?

   Thanks again for your help and time, looking forward to your answer,
   Best regards,
   Adrien.

   Posted 8 years ago #

4. Bradley Treeby
   

   Developer
   Joined: Mar '10
   Posts: 1,640

   Hi Adrien,

   My intuition is that you will only need the fluid model, but I say that with the caveat of never having studied this type of environmental acoustics problem directly.

   There are currently no C++ (OpenMP/MPI/CUDA) elastic codes. These are in the development pipeline, but it may be some time before they are released.

   Local grid refinement (e.g., using a subdomain method as you suggest) would be a good place to start if you want to model the layer directly. However, this is not implemented in k-Wave, and is non-trivial in the Fourier pseudospectral case. You could try looking through our list of other software. Clawpack in particular has some adaptive mesh refinement capabilities.

   Good luck with your simulations,

   Brad.

   Posted 8 years ago #

RSS feed for this topic

Reply

You must log in to post.