Wrong SaveScalars convective flux

Clearly defined bug reports and their fixes
Post Reply
Posts: 1
Joined: 08 Dec 2023, 17:02
Antispam: Yes

Wrong SaveScalars convective flux

Post by cedric.fournier »


First of all I want to thanks all contributors for the great software that elmer is !

I would like to report here a wrong calculation of the convective flux accross a surface boundary obtained with the SaveSclars model. Here you can found a zip file containing the mesh, the sif file, log file and also the vtu results :
https://drive.google.com/file/d/1ZoH264 ... sp=sharing
  • elmer version : Version: 9.0 (Rev: Release, Compiled: 2023-11-30)
  • Convective flux over boundary 3 is correct : 1.273442854802E-003 (third column of _air_flow.result file), 1.2686e-3 obtained with ParaView SurfaceFlow
  • Convective flux over boundary 1 and 2 are incorrect : 6.764948413966E-017 and 1.249834425623E-016 (first and second column of _air_flow.result file or _air_flow1.result of _air_flow2.result), 6.31e-4 and 6.38e-4 obtained with ParaView SurfaceFlow
So, what I could say is
  • The velocity field obtained is correct, but the computed convective flux over boundary 1 and 2 is not (~0)
  • The computed convective flux over boundary 3 is correct
  • I have tired severals variants (Solver 6, 7 and 8) (adding a solver only for bc1 and bc2, using a solver for bc1, bc2 and bc3)

  • A coarse mesh was used to reduce computing time and size
  • I know that the heat equation does not converge (due to coarse mesh), but this does not matter for the convective flux computed with velocity
Thank you for you support.

Site Admin
Posts: 4843
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland

Re: Wrong SaveScalars convective flux

Post by raback »

Hi Cedric,

Maybe you could check to remove the SaveScalars instances that give correct solution and check whether the problematic instances still give wrong answers. Or do the result depend on which order the instances are accessed.

The module should not have any memory that it was called with different parameter set, but there might still be some leftovers e.g. in allocation.

Post Reply