[Getdp] plane wave scattering by a sphere

laurent bernard ldbernard at gmail.com
Thu Dec 21 13:53:15 CET 2006


Hi,

I am back with my scattering problem.
Setting the Dropping_Tolerance to some small values, I obtained convergence
for lower (but still high) values of Nb_fill.
The results are good using second order elements on a finer mesh (see
attached pdf), but spending almost 1GB of RAM and 1 hour.
I don't manage to obtain convergence for other values of the parameter Freq
of my .pro files (for example, Freq = 1e7 instead of 1e8).

So I am still interested in any advice which could make the resolution
easier.

Thanks,

Laurent Bernard

PS: the .geo .par and .pro files are attached, here are the corresponding
processing messages:

P r o c e s s i n g . . .
Operation : Generate[A]
Solver    : Loading parameter file
'/home/lbernard/getdp/sphere_homo/solver.par'
Info      : Setting System {A,b} to zero
Resources : cpu 14.5586 s
Operation : Solve[A]
Solver    : No scaling of system of equations
Solver    : RCMK algebraic renumbering
Solver    : N: 62588, NZ: 2629288, BW max/avg: 112/42, SW max: 62315
Resources : cpu 16.8311 s
Solver    : ILUTP (Float, fill-in = 1500)
Solver    : N: 62588, NZ: 96125406, BW max/avg: 3000/1535, SW max: 10681
Resources : cpu 3432.68 s
Solver    : Generalized Minimum RESidual (GMRES)
    1  2.4992055e+02  1.0000000e+00
    2  ...

   34  1.4337349e-08  5.7367627e-11
Solver    : 34 Iterations / Residual: 1.43364e-08
Resources : cpu 3544.23 s
Operation : SaveSolution[A]
Resources : cpu 3544.41 s
E n d   P r o c e s s i n g




On 12/1/06, laurent bernard <ldbernard at gmail.com> wrote:
>
> Hello,
>
> I am trying to compute the field scattered by an homogeneous sphere
> exposed to an incident  electromagnetic plane wave.
> The unknown is the scattered field (electric field for example).
> The source is the incident field in the volume of the sphere. (See for
> example P.D.Ledger's thesis §5.3 for the variational formulation of the
> problem http://www.sam.math.ethz.ch/~ledger/publication.html<http://www.sam.math.ethz.ch/%7Eledger/publication.html>
> ).
> PML are used to troncate the computational domain.
> Only one-quarter of the sphere is represented (taking advantage of the
> symmetries).
>
> The mesh is generated using Gmsh (netgen algorithm with optimization)
> (18159 elements)
>
> Problems:
> Nb_Fill must be set to 3000 in order the obtain convergence. For lower
> Nb_Fill, the solver faces break-down. In consequence, the resolution is
> highly RAM and time consuming. So i don't manage to solve the problem on a
> finer mesh, and i can't use second order elements.
> The solution shows unexpected irregularities (see attached pdf file, which
> shows the components of the total electric field on a segment passing
> through the sphere).
>
> Questions:
> Is there any mistake in the .pro files? In the geommetry? (see attached
> files)
> If not, what should i do to improve the results? to solve the problem
> using less RAM?
>
> thank you,
> best regards
>
> Laurent Bernard
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.geuz.org/pipermail/getdp/attachments/20061221/dc9c0ea2/attachment.htm 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: electric_field2.pdf
Type: application/pdf
Size: 17463 bytes
Desc: not available
Url : http://www.geuz.org/pipermail/getdp/attachments/20061221/dc9c0ea2/attachment.pdf 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: geometrie_x.geo
Type: application/octet-stream
Size: 17743 bytes
Desc: not available
Url : http://www.geuz.org/pipermail/getdp/attachments/20061221/dc9c0ea2/attachment.geo 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: solver.par
Type: application/octet-stream
Size: 4788 bytes
Desc: not available
Url : http://www.geuz.org/pipermail/getdp/attachments/20061221/dc9c0ea2/attachment.par 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: IntJac.pro
Type: application/octet-stream
Size: 2003 bytes
Desc: not available
Url : http://www.geuz.org/pipermail/getdp/attachments/20061221/dc9c0ea2/attachment.pro 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: MWE0.pro
Type: application/octet-stream
Size: 2900 bytes
Desc: not available
Url : http://www.geuz.org/pipermail/getdp/attachments/20061221/dc9c0ea2/attachment-0001.pro 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: sphere_homo.pro
Type: application/octet-stream
Size: 8466 bytes
Desc: not available
Url : http://www.geuz.org/pipermail/getdp/attachments/20061221/dc9c0ea2/attachment-0002.pro 


More information about the getdp mailing list