Questions - Valparaíso - Chile

Christophe Geuzaine Christophe.Geuzaine at ulg.ac.be
Wed Sep 13 10:33:08 CEST 2000


Alejandro Angulo wrote:
> 
> Christophe:
> 
> In these moment I analyzing a NonLinear Time Domain
> Problem ( Inrush current in a power transformer ), but
> I have had many problems to define a precise value for
> the Relaxation Factor of the Nonlinear iteracion.
> Initially, an unitary value is adapted, but as the
> saturation of the core increase, the relaxation factor
> value should be smaller. But, if I correct this for
> high flux density value, the resolution for low values
> it worsens (I improve the solution on one hand, but
> for other I worsen it) How I could improve this?
> Can I solve the problem with variable step algorithms?
> Can I define funcional dependence between the
> relaxation factor and the field variables ( or
> magnetization current for example)? How do I make
> this?
> 

Hi Alejandro,

You can define the time step, the relaxation value, etc. as classical
GetDP expressions. For example, you can use "RelaxationFactor RelFact[]
;" in a resolution with this defined in the Function field : 

RelFact[] = 
 ($Time < 45.) ? 1. :
 ($Time < 46.) ? 0.5 :
 ($Time < 62.) ? 0.01 : 0.8 ;

Better, you could store some values depending of the fields in the
equation terms in registers (i.e. using the "#number" mechanism), and
play with that value in the tests for the relaxation factor.

-- 
Christophe Geuzaine

Tel: +32-(0)4-366.37.10    mailto:Christophe.Geuzaine at ulg.ac.be
Fax: +32-(0)4-366.29.10    http://www.geuz.org