Hi Thanh,
I've cc'd the list for archival. Hope you don't mind.
Thanh Luan Nguyen <Thanh.Nguyen-w7t(a)ruhr-uni-bochum.de> writes:
I'm looking for a right tool to do simulation of
elastic wave
propagation in the domain of the subsurface within around underground
tunnel. There is commonly used free software such as SPECFEM3D, which
implements the spectral element method, for doing the job. However, as
my colleage from the geophysics department at the Ruhr-University told
me that it took nearly two days for one run of simulation on a computer
cluster of 100 cores. I don't have such a computer cluster or
supercomputer at my group.
You have shown very attractive results of the DG on GPUs. I have the
following questions that I'd like to ask you.
1. Does 'hedge', which implements the Discontinueous Galerkin, have the
capability of simulating elastic wave propagation for a 2D/ 3D
structural model taking into account absorbing boundaries, attenuation,
etc.? In short, can one get a seismogram as a result of simulation when
the model (dimensions, velocity profiles, source points, boundaries) has
been defined?
Full elastic waves aren't there yet, but they shouldn't be hard to do
with the examples provided below.
2. If not, could you give me some comments on the
possibility of program
extension? I think that 'hedge' could be a good tool for simulating
seismic wave propagation of a relatively small domain for one who does
not have the high computing facility at hand - which is my case.
Check out this simple example--this file provides constant and
variable-coefficient wave propagation:
https://github.com/inducer/hedge/blob/master/hedge/models/wave.py
The hardest part is getting the numerical flux right. Check out the
Maxima scripts in hedge/doc/maxima. Those can help with that. Also, they
can help set up absorbing BCs. Attenuation is pretty simple, I
think--just add a decay term to the PDE.
Hope this helps,
Andreas