[Getdp] Convergence problem with inhomogeneous Helmholtz equation

Colignon David David.Colignon at ulg.ac.be
Fri Jan 4 20:34:08 CET 2008


Hi Lars,

I have not run your file with getdp, but you should put back the Krilov_Size to 80 or 100 and
increase the Nb_Fill parameter to 40 , 60 , 80 , ... untill you converge to a solution in a
reasonable (between 30 and 100) number of iterations.

Cheers

Dave



> Here you go. Thanks for the help.
>
> Lars
>
>
> -----Original Message-----
> From: Colignon David [mailto:David.Colignon at ulg.ac.be]
> Sent: Fri 1/4/2008 12:07 PM
> To: Lars Rindorf
> Cc: getdp at geuz.org
> Subject: Re: [Getdp] Convergence problem with inhomogeneous Helmholtz      equation
>
>
> Hi,
>
> can you also send us the  solver.par  file you use ?
>
> Cheers,
>
> Dave
>
>
>> Dear all
>>
>> I am having some problems with the convergence of the solution of an inhomogeneous Helmholtz
>> equation. The problems occurs for frequencies of the driving term which are about more than five
>> times the fundamental eigenfrequency of the system. For the solver parameters I use the default
>> with the maximum number of iterations increased to 3000. I have tried to change the solver and
>> the dimension of the Krylov space. Changing the dimension of the Krylov seems to do something,
>> although it does not solve the problem. The driving term of the inhomogeneous equations is
>> applied
>> in a point. I tried a line, and a subdomain and for those the problem was the same.
>>
>> I've made the model in Comsol, and here there was no problem. I also tried to change my nodal
>> field to a 1P form, but this made no difference.
>>
>> Can anyone help?
>>
>> I have attached the files. The model offers two resolutions: 'static' and 'eig'. Choose
>> 'static'.
>> Also choose 'static' for the postoperation.
>>
>> Kind regards
>> Lars Rindorf
>> The Danish Institute of Technology
>>
>> _______________________________________________
>> getdp mailing list
>> getdp at geuz.org
>> http://www.geuz.org/mailman/listinfo/getdp
>>
>
>
>
>
>