[Paraview] Re: ParaView Digest, Vol 35, Issue 30

[Yvan Fournier]

Hello,

You can read cells as VTK convex point sets (I have a Patch
for ParaView to read EnSight nfaced cells as such element types,
submitted a while ago to PV's bugtracker). This is quite
memory intensive: a test on a 500 000 polyhedral cell
mesh led to some 12 million internal tetrahdra for
ParaView 2.6, using some 2 Gb memory (not leaving enough
memory for subsequent operations on a 32bit Linux system),
while the same mesh handled much better under Ensight
(but even EnSight tends to be less stable using
meshes wih general polygons or polyhedra than with
regular meshes.

With smaller meshes, ParaView + patch works quite well,
though you may observe some "false" internal boundaries
between polyhedra and regular elements, which dissapear
when the mesh is tetrahedralized.

The most faces I have encountered in practice for polyhedra
is around 50, resulting from merging non-conforming meshes,
splitting faces along their intersections. Our code
is EDF's Code_Saturne CFD code (now under GPL).

I remember reading on this list that CD Adapco's Star-CCM uses
VTK as a base for their viewer (and use polyhedral meshes) but
that they keep their modifications to VTK for polyhedra to
themselves.

Best regards,

  Yvan Fournier

  Research Engineer
  EDF R&D
  Fluid Mechanics, Power Generation and Environment Department
  6 quai Watier
  78400 Chatou
  France

  yvan.fournier at edf.fr

  (secondary address: yvan.fournier at free.fr)


> Message: 1
> Date: Mon, 26 Mar 2007 15:29:41 +0100
> From: "Parker, Andrew (UK)" <Andrew.Parker2 at baesystems.com>
> Subject: [Paraview] Arbitrary Polyhedral
> To: <paraview at paraview.org>
> Message-ID: <0DB084BE537971419DC376EF0CD07840024786 at glkms2122>
> Content-Type: text/plain; charset="us-ascii"
>
>
> Hi,
>
> Bet this has been covered/done to death, but I need to ask, sorry.  We
> have a cut-cell code, our solver can handle, and has been able to for a
> number of years, arbitrary polyhedral, i.e. cells with arbitrary numbers
> of faces, and are not known cells types.  Recently we meshed something
> where one of the cells had 108 faces defining it.  Now, whether this is
> desirable is for another time, but can paraview read in meshes and
> solution data which have arbitrary numbers of faces defining the cells??
> If it can, what file format is recommended??
>
> If not, then is this on the horizon (version III), or will it never be?
>
> Apologies, I know this will cover old ground for most people,
>
> Andy
> - - - - - - - - - - - - - - - - -
> Dr Andrew Parker
> Research Engineer
>
> BAE Systems
> Advanced Technology Centre (ATC) - Sowerby
> Mathematical Modelling Department
> Filton
> PO Box 5,
> Bristol
> BS34 7QW
> UK
>
> e:   andrew.parker2 at baesystems.com
> <blocked::mailto:andrew.parker2 at baesystems.com>
> t:   0117 302 8248
> f:   0117 302 8007
> m:
>