Re: finite volume two dimensional steady-state diffusion in unstructured triangular mesh

On Jan 9, 4:54 pm, spellu...@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx (Peter
Spellucci) wrote:
In article <1168351573.837266.101...@xxxxxxxxxxxxxxxxxxxxxxxxxxx>, "=?iso-8859-1?q?Erik_Wikstr=F6m?=" <eri...@xxxxxxxxxxxxxxxxxxx> writes: >I have a mesh consisting of triangles on which I'm to calculate heat
>conduction. I understand how to calculate when the mesh consists of
>structured square cells and I have an idea of how to do it with
>triangles, but I'm wondering if someone have any good documentation
>just to be sure.
>
>My problem is this, how to select the centre of the triangle and how to
>calculate the distance between nodes. My current idea is to use the
>centroid as centre and to calculate the distance between nodes by
>taking the shortest distance from the centre to shared edge for both
>the triangles and adding them up. Does anyone know how this is usually
>done?
>
>--
>Erik Wikstr=F6m
>

in the subject line you had "finite volume" : then you must take the
center of the circumscribed circle in order to get connections which
cut the edges orthogonally. you compute the distance as the euclidean distance
and apply standard finite differences in order to express the normal derivatives

Thank you for your answer, a question though. Can't the fact that the
circumcenter does not always lie in the triangle cause trouble. For
some triangles the circumcenter can lie inside the neighbouring
triangle and the distance can then become very short.

Another question, if two neighbouring triangles does not lie in the
same plane, but at an angel. Assuming small depth compared to the area
and that the angle between the normals of the triangles is small.
Should I still use the euclidean distance or would the distance along
the "surface" of the triangles be more correct?

--
Erik Wikström

.

Relevant Pages
Loading