8 Replies Latest reply on Nov 15, 2013 11:43 AM by Jared Conway

    Long narrow gap

    Bryan Nathan

      I'm trying to calculate the pressure drop to various ports through a gap between a piston and cylinder. The gap is only 0.002mm wide (radially), but is 30+mm long (on a 6mm diameter piston). Using the standard refinement with 0.002mm as the smallest gap does not work (it doesn't push any flow through the gaps). I've tried narrow channel refinement along the surface of interest, but it never builds a mesh (RAM peaks at 24GB and progress essentially stops). Are there tricks that can be played with the cell aspect ratio or other things that can help? I'd prefer to have 3 cells accross the gap but have a suspicion that the software won't be able to help me with this particular problem.

       

      On a similar note, is there a rough Cell/Memory ratio for figuring out what the largest problem size can be run given a certain amount of RAM?

        • Re: Long narrow gap
          Bill McEachern

          That's an aspect ratio of 15,000:1 - non trivially large as they say.

          • Re: Long narrow gap
            Jared Conway

            On a similar note, is there a rough Cell/Memory ratio for figuring out what the largest problem size can be run given a certain amount of RAM?

             

            > there is an older KB article about this in the solidworks KB.

             

            it doesn't push any flow through the gaps

             

            > i suspect you have another problem with your model. even with really coarse meshing, the software should still be able to handle this. have you tested a really simple version that you could post?

             

            and a comment about getting 3 cells across the channel, i'd probably suggest getting it to run with a coarse mesh first then take a look at the results and see if you need that refinement. if you need that refinement, a porous medium would likely be better.

              • Re: Long narrow gap
                Bryan Nathan

                Sorry for the slow response, my machine was tied up for 2 days working on a different problem. I've done as you suggested and started increasing the mesh resolution and watching how it behaves. The picture below is from a run with 27,000 Fluid Cells and 46,000 Partial Cells (had to push the refinement level of the critical surface to 7). The behavior is still unreliable in my opinion, but I'm continuing to increase the refinement without turning on Narrow Channel Refinement (which kills my machine). The inlet flow does not match the combined outlet flows (as it should).

                Run 5 Result Summary.jpgRun 5 Mesh.jpg

                Is there a way to change the allowed aspect ratio? If I could go to 3:1 it may make the problem possible.

                 

                I'm not familiar with "KB". Can you explain what you meant?

                  • Re: Long narrow gap
                    Jared Conway

                    kb: customerportal.solidworks.com > knowledgebase

                     

                    "The behavior is still unreliable in my opinion," > what does this mean exactly?

                     

                    If I could go to 3:1 it may make the problem possible. > take a look at the basic mesh settings. the tutorial on mesh optimization should cover some of this but also more of it is covered in the technical reference and solving engineering problems documents in the install directory.

                     

                    what does the gap do in this analysis?

                     

                    and regarding the flow not going anywhere, note that the length that flow trajectories follows is finite. so you may just need to extend them or use a cutplot to see the flow characteristics.

                      • Re: Long narrow gap
                        Bryan Nathan

                        It isn't reliable means that the Inlet does not match the sum of the two outlets (not even close). The thin gap is to prevent leakage when the piston is turned at a certain angle. The analysis is to estimate how much is still leaking through.

                         

                        Increasing the trajectory lengths by a factor of 1000 did nothing. I've tried looking at cut plots but they aren't very helpful with such an extremely thin gap. I may try again after the next analysis.

                         

                        I'm going to run a more refined problem overnight to see if the results change significantly.

                          • Re: Long narrow gap
                            Jared Conway

                            sounds like a setup problem vs a mesh refinement problem. i'd suggest removing the part that rotates, make sure that problem works, then make a thinner roating part. see if that works...etc.

                             

                            or you're running out of iterations and the solution is completing before converging.

                             

                            probably worth posting the model or building a simplified one. (the simplified one will help you identify if you have a setup problem or a software problem)

                              • Re: Long narrow gap
                                Bryan Nathan

                                I did as you suggested and opened up the gap so that I could mesh it (3 cells accross). It did expose one problem with my setup (I had goals for volume flow, but as I've switched to air for this problem I need to use mass flow. Rookie mistake). The increased gap ran fine, with the combined outlets matching the inlet mass flow rate. The trajectory plot also looked as one might expect.

                                Run 9 Mesh.jpgRun 9 Trajectories.jpg

                                 

                                I then reran the original gap (0.00175mm) using mass flow as my goal. After converging, the combined outlets did not match the inlet mass flow. I'm going to slowly increase the mesh refinement to see if the answers start changing.