8 Replies Latest reply on Mar 5, 2018 1:43 PM by Joe Galliera

    Best meshing on boundary layers between fluid/solid and treatment of partial cells

    Jannis Wachter

      Hello all,

      i need to simulate some inlet geometries designed as different diffusors. The water flow stream will then impinge on a stack of plates, which will be 300um thick and also with a 300um spacing between them. In order to check these inlet geometries and flow distribution on the flow channels i will first need to do only stationary flow simulation without any heat transferring, just flow. In addition, i will be in very laminar flow range about reynolds numbers of ~50 to 200.

       

      So in case of meshing: Right now i have manually set planes to get exactly two fluid cells in between of two solid material plates. As SW FS is solving with cartesian cuboid cells, i thought initially it is better to have better have "pure" cells as partial cells if possible. Please see img_001 for the meshing and img_008 for the results. As for the diffusor this question does not appear, but for the channels and at the inlet into the channels i only have complete fluid cells. As for example in:

      https://forum.solidworks.com/message/324294?q=Meshing%20on%20boundary%20layers%20between%20fluid/solid%20and%20treatment%20of%20partial%20cells

      Bill McEachern

       

        • Re: Best meshing on boundary layers between fluid/solid and treatment of partial cells
          Bill McEachern

          the cells across he channels is pretty rough at 2. However, that does mean you won't get a decent approximation. Flow has empirical corrections when not enough cells go across a channel. Technically you would need like 10 or more to resolve the profile decently. I would suggest you run tests to see how it goes - maybe even use the 2D approximation for comparison runs and then use the most appropriate for the full 3D run.

          In your image you do not have a "body fitted" mesh even though it might look like it. The program automatically puts in partial cells at the boundary as that is the way the program works - immersed boundary or cut cell approximation upon which the code is based requires it. The solid cells are not plotted when you don't have heat transfer enabled - well at least they are not shown in your plots.

            • Re: Best meshing on boundary layers between fluid/solid and treatment of partial cells
              Jannis Wachter

              Thanks for you answer, Bill. So for my visible mesh i have created and as i have attached, will it actually make no real difference between what i have done and with just visible partial cells or does it make maybe a difference in computational effort if the program needs to extra create these partial cells at my body fitted borders?

               

              On the mesh i have some extra question i couldn't find any real answer: As in the picture i have more like "longer" cells than real cubic. When thinking about that, they should be more cubic especially on the roundings, i guess. But in general, does it make a huge difference? The reason i did this and also with the coarse mesh, because i have limited computational resources and to save mesh cells.

               

              Thanks for the suggestion on 2D validation: I have done this a while ago, but when i started to mesh now and create in 2D on X-Z-Plane, i realized that it also creates a lot of cells in Y-direction?! On the basic mesh, there i have only 1 cell in Y direction. I couldnt find any reasonable explanation for this.

               

              Thanks again in advance.Bill McEachern @

            • Re: Best meshing on boundary layers between fluid/solid and treatment of partial cells
              Siavash Khajehhasani

              Hi Jannis, The detailed technical answer to clarify your points can be found in 'flow simulation technical reference' notes. This PDF comes with your SW installation (you can find it under help menu) or simply google it. As a side note, I, personally, am a bit fan of performing the best controlled mesh refinement method using a dummy solid body.

              PS. sorry for the short response, as I trust you'll do a great research.

              Just for your reference:

              Screen Shot 2018-02-23 at 8.18.51 PM.png

              • Re: Best meshing on boundary layers between fluid/solid and treatment of partial cells
                Siavash Khajehhasani

                The straight answer to why it generates partial cells is: SW flow Sim only has hexahedral elements which are mapped along the global coordinate system and there is no way it can get mapped along curves. Once again using local meshes with the aid of a dummy solid body could be a game changer for you and give you power over outoma options from software which sometime are limitations for power users.

                • Re: Best meshing on boundary layers between fluid/solid and treatment of partial cells
                  Joe Galliera

                  I would carry out the calculation like this:

                  1. Possibly 2D as trials, but for 3D utilize Symmetry in the Computational Domain, at least half symmetry and without seeing rest of the model possibly quarter symmetry.  Remember that you can mirror the results (by clicking the option in the plot dialog) to show both sides of the model when using symmetry.  Now you are able to increase the fidelity of the mesh with fewer total cells than would be done for the entire model.  (Note that some BC's such as volume or mass flow rate would need to be proportionally reduced to the section simulated.)

                  2. Try first using Global mesh with the option for refining narrow channels, but sometimes this will increase the mesh greatly.  Another option is to create a local mesh, where now you can define primitive volumes on the fly in the local mesh definition.  Some things there could be to additionally have more control over the gap size to define a "narrow channel" and how many cells to put in that gap.  What types of cells (fluid or solid only, partial cells) and how many levels to reduce them by.

                  3. Change unit of Geometric Characteristic for Angle to degrees (°) instead of radians, unless you understand that better, and define mesh controls so that the cells will be split when a cell covers an arc larger than the angle than you specified.

                  4. Know that with the tools available to you in Flow Simulation that you are able to create any mesh that you want with the controls that you define.