Maintaining a uniform thickness of the inflation layer on a wall with narrow face.

    • vladr


      I am having a difficulty to create a uniform inflation on a wall that has a narrow face near the corner.

      Problem description:

      An (isolated) geometry consists of two regions: a thin wall cylinder ("gap") that is connected to a larger "volume" on one of its faces

       Geometry Geometry Crossection

      The model will be a heat transfer with natural convection in Fluent. Both regions will be the same material (air).

      The convection will be modeled in the "volume" only.

      The "gap" will have heat conduction only, however will need to resolve a sharp gradient (between a heater and a body being heated), so few elements are needed across it.

      I mesh the gap first with hexahedral elements (multizone or sweep).

      Then the  "volume" with automatic or tetrahedral algorithm, with prescribed inflation (Total thickness) on the walls and also the same inflation on the imaginary surface between the gap and the volume:

       Mesh cross-section Mesh cross-section, view from the right side



    • vladr

      The problem is that in the corner where the gap is connected to the "volume", the inflation layer gets compressed:

      Problematic region

      I guess that this is because the inflation forces elements to have their thickness (normal to the "wall") to be not greater than its length (parallel to the "wall"). However, I could not find if there is any parameter that can "relax" that requirement. 

      If there are less elements across that face, then the problem becomes less severe:

      Problematic region with coarser face mesh

      However I con not use this workaround because I need to maintain the number of the elements in the gap on that face which is a critical region for the model. And in general would need to maintain some independence between the mesh in the "gap" and the thickness of the inflation in the "volume", at least to some extent.

      Is there a workaround to have the thickness of the inflation layer unaffected by the surface mesh on that narrow face?

      Thank you in advance.

    • vladr

      AN UPDATE:

      I have found the solution to some extent: the permissible ratio of the height of the element in the inflation layer over is width is controlled with a "Maximum height over base" parameter in the global mesh inflation group under the advanced settings.

      The maximum is 5 and it didn't solve it completely. Increasing the Total Thickness of the inflation on the problematic face also did not help to completely equalize the thickness of the layers near the corner. Only decreasing the thickness elsewhere helped to achieve what I wanted (here I also used twice less elements across the gap than in the initial post):

      Corrected inflation layer thickness in the corner. Skeweness colormap.


      However now I have a problem which I left out in my initial question: the tetrahedral mesh in that corner has few elements of a very poor quality (the colormap in the pic is a Skeweness). I have no clear vision how to improve it and just tweak global sizing parameters as well as playing with element size on the adjacent faces. Can anyone suggest any ideas which way to go, which parameters to focus on to improve these few tetrahedrons in the corner?

      Note that they are also quite long in the direction into the page which is dictated by the mesh on that small face. Should I create a separate discussion for this question?

      I would still welcome any suggestions, If there are, on if there is a better way to control the layer thickness, than I did.

      Thank you

    • Rob
      Ansys Employee

      I'd look at decomposing the model to let you sweep the thin walled section through the larger volume. You can pick multiple volumes for a named selection later on so it'll not effect the overall model as seen by the solver. 

    • vladr

      Thank you for your suggestion, I will try it and post an update on how it worked.

      Does this thin wall subvolume of the large "volume" (i.e. flow region) necessarily has to have an inflation (decreasing element size when approaching a solid wall) in it for a CFD simulation, of just a several elements with certain constant size in it (then followed by actual inflation layer in the bulk "volume") would do equally well?  

    • Rob
      Ansys Employee

      If the thin volume is a solid it doesn't need any inflation.  Technically nothing needs inflation, but it's a very efficient way to get the near wall flow correct without huge numbers of cells. 

Viewing 5 reply threads
  • You must be logged in to reply to this topic.