dolfin team mailing list archive

[Marie Rognes <question105900@xxxxxxxxxxxxxxxxxxxxx>]

Question #105900 on DOLFIN changed:
https://answers.launchpad.net/dolfin/+question/105900

    Status: Open => Answered

Marie Rognes proposed the following answer:
Chris Richardson wrote:
> Question #105900 on DOLFIN changed:
> https://answers.launchpad.net/dolfin/+question/105900
>
>     Status: Answered => Open
>
> Chris Richardson is still having a problem:
> OK, thanks that clears up one thing...
>
> My question now becomes how to set the flux BC on the other boundaries when they are not so simple and 
> straight.
>
> e.g. I thought this would work, but is giving real problems at the
> corners:
>
>   

I've attached a revised version. Take a look and see if it makes sense.

Note that with (the previous):

    def boundary0(x,on_bound):
    (*)    return x[0]>DOLFIN_EPS and x[0]<1.0-DOLFIN_EPS and on_bound
    bcs = DirichletBC(V.sub(0), u0, boundary0)
      

only the facets where both endpoints (and the midpoint) satisfy (*) will 
be marked. (Somebody correct me if I'm wrong here.) In particular, the 
left-most "bottom" edge will not be marked. 

Further, then v*n will not vanish on this edge, and hence you will get 
contributions from this edge in:

        L = dot(v,f*n)*ds

Then, the definition of f on the edges starts mattering etc.

The moral is that one should be really careful with the boundary 
conditions ;)

--
Marie

-- 
You received this question notification because you are a member of
DOLFIN Team, which is an answer contact for DOLFIN.