dolfin team mailing list archive

["Johan Hoffman" <jhoffman@xxxxxxxxxx>]

Yes, the Unicorn solver(s) is based on a fixed point (or non-exact Newton)
iteration.

I do not know what the problem of the dolfin non-linear Newton solver is.

/Johan

> <p>Hi,</p><p>Yes, You are right. I am interested in a problem similar to
> Navier-Stokes but not exactly that.</p><p>I took a look at Unicorn,
> Although it was not updated, I figured out how the solver was written
> based on Fixed Point iterations.</p><p>My question referes to dolfin
> VariationalProblem.cpp at line 75:</p><p>&quot;void
> VariationalProblem::solve(Function&amp; u0, Function&amp;
> u1)&quot;</p><p>This means that one should be able to solve the nonlinear
> problems with more than one variable by just augmenting the state-space
> vector and solving the whole state with Newton's iteration. The only
> difference with the linear case should be the fact that instead of solving
> for the solution (U) dirrectly, the nonlinear problem should be sovled for
> an increment (\delta U) around a guessed solution (say U0) and correcting
> the U0 in every Newton's iteration.&nbsp;</p><p>To me it should be
> possible to solve a nonlinear problem with Newton's iteration based on
> what I can see in&nbsp; the afforementioned file in (dolfin/fem) .Please
> correct me if I am wrong. However the following peice of code gives an
> error like &quot;Missing eval() for user-defined function (must be
> overloaded)&quot;.  and in my usr-define functions such as Source, etc
> eval() is overloaded. I do not see where the problem arises. Any
> opinion?</p><p>{VariationalProblem nlpde(a,L,bcs,true);<br />&nbsp;
> nlpde.solve(U,P);}</p><p>&nbsp;</p><p>Thanks</p><p>Hatef</p><p>&nbsp;</p><p>&gt;Hi
> Hatef,<br /><br />&gt;It sounds as if you are interested in the
> Navier-Stokes equations, or</p><p>&gt;similar? If so, you may want to
> check out the Unicorn solver:</p><p><a
> href="http://www.fenics.org/wiki/Unicorn";
> rel="nofollow">&gt;http://www.fenics.org/wiki/Unicorn</a><br /><br
> />&gt;Direct any Unicorn questions to: unicorn-...@xxxxxxxxxx<br />&gt;<br
> />&gt;Best,<br />&gt;Johan</p>
> _______________________________________________
> DOLFIN-dev mailing list
> DOLFIN-dev@xxxxxxxxxx
> http://www.fenics.org/mailman/listinfo/dolfin-dev
>