Acoustics - Internal Wall/Hard Boundary

Jobin

Hello,

I'm a relatively new user to Code Aster, interested in muffler acoustics. I've been able to set up a very simple expansion chamber muffler and successfully solve the pressure fields, and also do the necessary post processing to extract the transmission loss of the muffler.

The inlet has a normal velocity boundary condition, outlet has an impedance boundary condition condition, and of course, the shell of the muffler by default is a hard wall boundary condition.

Now, I'm trying to work with mufflers with more complex geometries; specifically, an extended inlet/outlet expansion chamber muffler.

I've been able to successfully mesh the part. The image below shows the mesh, with the internal extension groups highlighted.

However, now I'm not sure how to proceed. The highlighted surfaces are "Interior Sound Hard Boundaries" (COMSOL terminology), which basically means that the normal component of velocity is 0.

I figure I need to apply this using something like VNOR=(0.0+0j). However, which matrix/vector takes it into account? The excitation vector or the damping vector?

Any thoughts on how I can proceed is appreciated.

Regards,

Jobin

CLF

Hi Jobin

Good to see someone doing acoustics with CA. :)

A limitation, unless you decide to tweak matrices yourself, is that CA acoustic elements do not compensate for flow effects so, it is limited to low Mach numbers.

For the inserts, I believe that all you have to do is not connect nodes and these will be separated by an infinitely stiff wall, i.e. just connect acoustically at the pipe ends.

FYI - CA does not have any fancy stuff like acoustic transfer matrices should you wish to model perforated surfaces. That said, the mass, damping and stiffness matrices can be tweaked as CA executes inside Python using Python commands.

Sound radiation to the exterior can be modeled. CA has a rubberband mesh (3D_FLUI_ABSO) that can be applied to add absorption to the boundaries. If this is not of importance, you can add an impedance boundary condition (IMPE) and, I believe, make it frequency dependent using a Table.

Pressure release (i.e. open end conditions) are set using the variable PHI = 0.0 This took me a while to figure out.

A very neat feature in CA is that it provides plane wave surfaces as an excitation.

As you may know - an alternative, often truer, muffler performance measure is to compute the Insertion Loss (IL). IL is the difference at the endpoint (where and how you wish to define it) between the output from a straight pipe and from the pipe with muffler, i.e. the difference in results from two separate runs.

Hope this helps

Claes

Jobin

Hi Claes,

Thank you for your response. It sometimes feels like I'm the only one trying to do acoustics work with CA - we need more discussion on these forums, haha.

1) Thank you for the heads up regarding low Mach numbers. Thankfully, my work primarily deals with low Mach numbers :)

2) Is there a way to adjust the mesh connectivity in Code Aster or is it something I need to do during mesh generation?

3) My plan was to introduce perforate surface using a perforate impedace model; i.e., applying a complex surface impedance on an internal surface. Does this seem like a viable approach?

4) You mentioned "tweaking matrices"... are there any guides for doing this kind of work?

5) 3D_FLUI_ABSO is quite interesting. Does it basically work as a perfectly matched layer?

6) Transmission loss is my current goal as much of the literature predicts it. So, for the initial phase of testing, I thought this was better. However, my goal is to eventually expand to insertion loss prediction - it's much more practically useful as you said. To do IL simulations, does a setup like this make sense?

- Muffler Inlet: Plane wave surface excitation

- Muffler Outlet: Pressure release

- Muffler surrounded with a fluid mesh with a 3D_FLUI_ABSO boundary

- Sound pressure probe somewhere in the fluid mesh

CLF

Hi Jobin

1) Yes, there could be more of us as CA is a good code for this kind of analysis.

2) You can duplicate meshes and nodes inside CA. To the best of my knowledge, you cannot modify the actual mesh. The major trick in keeping track of things is to apply group names for nodes and elements that makes sense.

3) Normal Impedance can be used, transfer impedance matrices cannot - unless you add them manually to the system matrices.

4) Surely. However, I do not know where and I have not done it. Ask/search the forum. Before asking, try using numpy during execution - just add a few lines of python in the CA analysis instructions script.

5) No, nothing perfect about it. It is referred to as a rubber band mesh. There are better boundary meshes, so called infinteFE around, though CA does handle not only acoustic P-waves but also other wave types, i.e. the rubber band mesh is usable also geotechnical and seismic analysis.

6) You almost nailed it - a) add the step that you do exactly the same analysis with a straight pipe of the same length as your reference in the IL computation and b) do not use pressure release on the outlet as you then reflect 100% of the incoming waves - you should just connect the internal and external meshes at the outlet. That said, most muffler SW computes IL at the muffler outlet using the impedance for a radiating piston and views the muffler end as a monopole and does the rest analytically to get the response at the receiver position(s). Again, a few lines of Python can do wonders...

Have fun.

susuklem

Hi everyone.

I´m trying to solve one case very similar to this one. I have been using ANSYS to calculate transmission loss in different kind of mufflers but now I want to do the same in code_aster. I have succesfully solved a simple expansion chamber, in which air is the only acoustic fluid. However, my results are a little strange around one determined frequency, and I don´t know if this is normal or something is wrong in my simulation in code_aster.

In order to easily evaluate the results, I have simulated a simple case from the article of "Duct Acoustics" in Wikipedia.

This is basically a simple reactive muffler with a ratio of 1/3 between inlet and expansion chamber, and a length of 0.5 meters. In ANSYS I have obtained values of transmission loss very similar to the results of wikipedia. In the case of code_aster, values of sound pressure at the inlet and outlet of the muffler are very simillar to ANSYS except for frequencies around 1000 HZ, where they are quite different.

I am using the following boundary conditions:

-normal velocity at the inlet: 0.0036 m/s

-anechoic outlet, with impedance rho·c

-walls

I attach the graphics with the results and my code in one .rar file, in case something is wrong.

Does anyone know which could be the problem?

Thanks in advance!!

Greetings!

Attached file:
muffler_case.rar, 113063kb

CLF

Your set up looks OK to me & still data is wrong.

For what it is worth. Theory is without any damping. Perhaps we should try to bark up this tree?

You input complex data with zero imaginary part. This should produce zero damping. However, in practice, there will always be some damping in there due to rounding. Also, there is the case where you phase shift excitation 90 degrees.

You solution scheme states an acoustic damping matrix C1. My suggestion - try to eliminate C1 and solve the case again.

It should not matter, but your parameter value PUIS_PULS=1 should not be needed as you state velocity as BC. Also, the 90 degree phase shift on excitation combined with complex data that should be zero may cause unfortunate numerical problems. It may be worth a try to toy a little with these to see if anything happens on the TL.

A comment - TL is one thing. Insertion Loss is the better descriptor of actual performance.

Hope this helps.

Sincerely

Claes

susuklem

Hi CLF,

thanks for the response. Finally I could solve the problem. I was applying the anechoic condition (impedance=rho·c) only in the outlet, and it must be applied in the inlet as well. With this, the results are very similar to ANSYS and the theory.

Greetings!

CLF

Glad it worked out in the end.

A request - for the benefit of other, new users, who also are likely to start with a case similar to yours. Can you please add a 100% working model, analysis files and output data to the forum? I believe this is good practice, i.e. not only to show problem scenarios but also ones that do work as expected.

Last, feel free to drop me a private mail. I may have something for you.

/Claes