Acoustics of automotive muffler

Altair Forum User
Altair Forum User
Altair Employee
edited October 2020 in Community Q&A

Hi,

 

I'm working on my master's thesis and the object of my studies is an aeroacoustic analysis of two automotive mufflers. For validation purposes of acoustics I have the experimental results so it is a benchmark for me. Basically I divided my case into two parts. The first part is to obtain aerodynamics characteristics of mean flow 10m/s inside the muffler , which I'e already done, and the second part is to compute the transmission loss of the muffler using transient CFD simulation and acuFwh post-processor. Before I manage to solve the real case, I decided to calibrate my acusolve environment on a simple expansion chamber muffler and find out if my results will cover with the experimental results. In order to do that I'm working with case from Fig 5, from this research:

https://www.acoustics.asn.au/conference_proceedings/AAS2004/ACOUSTIC/PDF/AUTHOR/AC040066.PDF

 

And now I'm really confused because of the boundary conditions and bad resuts I'm getting. As far as I know there are two ways to obtain such results:

1. Single period sinusoid pulse of velocity with ZERO mean flow

2. Single period sinusoid pulse of velocity with NON ZERO mean flow

 

In the paper above, they use the first case, where there is a single period 3200Hz sinusoid pulse with the amplitude of 0.05m/s.

But the question is how can I implement it to acusolve?

If, let's say, frequencies of interest are of 0-3200Hz as it is above and the sampling frequency is 200kHz than it gives me:

T_signal = 1/3200=0,0003125s

T_sampl = 1/200000=0.000005s

T_signal/T_sampl = 62.5 -> so this is an information for me, that the wave of 3200Hz will be sampled with 62 points using sampling frequency 200kHz, yes?

 

Altair says in their manual that it should be at least 6 points per wavelength in order to properly 'create' the wave, so 62>6, and that's OK

 

Now what should I do exactl using acusolve and acuFwh? Should I use Multiplier function at inlet surface of the muffler and create a single period sinusiod of f=3200Hz, but using 5us sampling period as the case above say? Ok, but it is only one period that lasts 0.0003125s so what TOTAL TIME of the simpulation should I use?  The thing I know is that I'm 'putting' a single pulse without mean flow into the muffler and now I'm 'waiting' for response of the system, but the question is how much should I 'wait' ?

I am really  confused so hope to see some help from you.

 

The other question is, if I'm using zero mean flow pulse, then should I turn off any turbulence modeling, because the flow is laminar?

 

PS: I know that CAA is difficult but I have access to a multi CPUs cluster so there is not an issue

 

Regards,

Adrian

Tagged:

Answers

  • Rahul_P1
    Rahul_P1
    Altair Employee
    edited October 2016

    Use acuFwh with external flows, not internal like the muffler in this example.

    This example can be problematic since the acoustic waves can bounce off

    the surfaces with open boundary conditions. The control volume would have

    to be bigger than just the muffler chamber of interest. Also, using compressible

    flow (for Mach < 0.8) will most likely not work well resolving acoustics waves,

    but potentially the sources can be computed.

     

    Assuming that you can solve the flow properly, there might be a chance to

    import the sources (from CAA_OUTPUT) into a separate acoustic solver,

    like the NVH module in OptiStruct. I have not see any muffler examples of this

    though, so I'm not sure it it works.

     

    The total time needed for the analysis depends on the lowest frequency,

    say a factor of five (or more) times the corresponding time.

     

    If the flow is laminar, the turbulence model should be turned off.

     

Categories