[Getdp] An example: non linear solution to the GetDP distribution 2D demos/test example

David Colignon David.Colignon at univ.u-3mrs.fr
Tue Dec 3 18:09:52 CET 2002


Hi,

here are some comments to improve the convergence and the results in the
table in "Results.pdf"

a) you can accelerate a little bit the calculations by lowering the value
of Nb_Fill to 14 in solver.par


b) for the calculations with the Picard scheme, you should choose a value
of NL_Relax between 0.5 and 1.0.

For example, in MagSta_phi_pic, for HcMag = 500000, you can obtain a mean
error of 1e-6 after 17 iterations if you put NL_Relax equal to 0.6 
You then get H center core    = 2.287e+2
             B center core    = 4.92e-01
             (H core near mag  = 2.86e+02) ( where exactly do you pick the
values for B and H core near mag ?)
             (B core near mag  = 6.09e-01)  

In MagSta_a_pic, for HcMag = 920000, with NL_Relax = 0.51, the system
converges and after 100 iterations you get a mean error of 5.5e-6

Dave





A 20:02 02/12/2002 -0200, vous avez écrit :
>Hello
>
>It took some time for me and for people that helped me in the list to
>get a complete magnetostatic 2D non linear calculation working. I'm
>submitting this example to the list hopping to help newbies and to save
>developers time to further improve this excellent software.
>
>Comments, remarks and corrections are welcome.  
>-- 
>Janvrot IVM
>janvrot at infolink.com.br
>janvrot at cetuc.puc-rio.br
>Attachment Converted: "C:\Documents and
Settings\colignon\Bureau\test-nl-2d.tgz"
>

Dave