nice getdp results and some questions

Christophe Geuzaine Christophe.Geuzaine at
Wed Sep 6 15:36:27 CEST 2000

Samuel Kvasnica wrote:
> Hi Christophe,
> >
> > I use 'Unit[{d a}]'. But you could also write something like
> > 'Vector[CompX[{d a}]/Norm[{d a}]], CompY[{d a}]/Norm[{d a}], CompZ[{d
> > a}]/Norm[{d a}]]]'.
> Aha, fine. But would this work also on points where B is almost 0 ?

There is an absolute tolerance of 1.e-30 in Unit[] (Unit[] returns
(0,0,0) if the norm is below 1.e-30). If you need another tolerance,
just define another function with a conditional statement.

> > > Btw, could you give me an advice for some good source of material property data - e.g.
> > > magnetization curves for standard materials ?
> >
> > That is a good question. In fact, we are quite often confronted to the
> > same problem, and I don't know any good compilation of such
> > characteristics. So we often interpolate on measurement points.
> I've decided to do the same 2 days ago. Lot of work but it works.
> > > Do you think it's a bug ?
> > >

I think the very small oscillations are due to the 2nd order shape
functions. The big ones are not a bug: when you specify a cut *exactly*
on y=-6e-3, the cutting line is on the boundary of the elements paving
different regions (the chosen element is selected as the first which is
found in the rapid search algorithm, but there is no reason to prefer
one element to the other). And since only the normal component of the
induction is continuous across the interface, the result you get is the
correct result, but evaluated either in the upper part, or in the lower
part... To be sure to evaluate in the air, just add a small epsilon,
e.g. y=-6e-3 - 1.e-8.

My best,

Christophe Geuzaine

Tel: +32-(0)4-366.37.10    mailto:Christophe.Geuzaine at
Fax: +32-(0)4-366.29.10