your assumptions (6)-(8) are correct.
I would not mix the new shape derivative term you propose with the
existing linear elastic term, but you are right that term it is a good
The key point is how you want to parametrize your domain, i.e. how the
derivatives dB/da and dJ/da will look like.
vg->bfGM are just gradients of the base functions w.r.t. the space
coordinates for each element - it's an "array" with shape (n_el, n_qp,
dim, n_ep), where n_el is the number of elements (of a group), n_qp number
of quadrature points, dim the space dimension and n_ep the number of
Sorry for not very deep answers, I am now finishing some work so I am
swamped with other things...
Feel free to ask more.
On Fri, 10 Sep 2010, Peter M. Clausen wrote:
> I've searched and found some stuff, but also raised a lot of new
> questions regarding Shape optimization with SfePy. I wrote some
> formulas and questions which are easier to read in the attached
> PDF-file. I've also pasted the tex-file here so people can search the
> text /mailing list.