i am new to this forum but already appreciate a lot it exists.
I am wondering if it is possible to calculate models with acoustical free field conditions (non-reflecting boundary) via the helmholtz-solver. I would like to investigate radiation from different shaped sources in Elmer.
I know that "CrocoDuck" did great acoustical calculations. One of them is radiation of a point/spherical source which is documented here. It is (also for me) a reference calculation I already did with actran. The full description of the reference calculation of a point source of CrodoDuck is written on his website:
https://computational-acoustics.gitlab. ... ng-sphere/
If this results - which matches the analytical ones really good - would work with more complex models, I would be really happy. But my understanding is the following:
The same impedance of the soundsources radiation at the location of the boundary is used for the boundary layer to simulate "free field condition".
In another thread in this forum it is written that this methods only works for normal incident.
This means if I want t use a different geometry where sources are complex (and sound radiation too) it is not possible to measure something in free field conditions with Elmer at the moment, isn't it?CWeng wrote: ↑05 Oct 2019, 11:55 Hello Peter,
Another useful feature is the non-reflection condition (maybe it already has been implemented but I haven't read it yet; I am still learning Elmer). If I am correct, in the current version of Elmer the non-reflection boundary condition is realised by setting the boundary impedance to the characteristic impedance (density*speed of sound). This may work well for normal incident but not for oblique incident, and the latter is encountered in many applications. For example, in duct acoustics if the frequency is high enough or if the duct wall impedance are not symmetric, etc, non-plane waves are able to propagate. Then oblique incident may occur at the boundaries and the wave is reflected. Another example is sound radiation in free space. If the user want to compute the radiated acoustic power he has to make sure there is no reflection from the outer boundaries which is generally hard to achieve by using a non reflection boundary condition (Actran uses the infinite element and the perfectly matched layer to realise the radiation condition, see https://www.fft.be/product/actran-acoustics)
To create more general non-reflecting condition, I think the buffer-zone approach might be adopted. This may be realised using the perfectly matched layer (PML). I find the following introduction to PML very helpful
https://math.mit.edu/~stevenj/18.369/pml.pdf
and figure 1 in the above tutorial explains the idea behind the buffer zone clearly
pml_fromStevenJ.PNG
Since the Helmholtz model is a frequency solver so the implementation of PML might be straightforward.
Another buffer-zone approach might be the implementation the artificial viscosity. In the Helmholtz solver the artificial viscosity may be given as the imaginary part of the speed of sound. I also saw that in the WaveSolver the viscosity has already been implemented and can be given as parameters:
artificialViscosity.PNG
In the region of interest (ROI) the viscosity can be set to zero. In the buffer zone the viscosity can be set to an nonphysically large value to damp the wave quickly within the zone. To avoid reflection due to sudden change in the viscosity at the interface between the ROI and buffer zone, the viscosity may be turned on gradually from zero to the large value at the outer boundary (linearly or quadratically, etc..). I think this function would be easier to implement if Elmer simply allows the users to define the viscosity as a function of space, i.e., eta = eta(x,y,z); then, the users have the freedom to create the buffer zone and tune parameters such as the width of the zone and the turn-on function of the viscosity.