8 Replies Latest reply on May 9, 2013 11:59 AM by Jared Conway

    Simulation - Meshing a family of parts

    Michael Mueller

      I am trying to run simulations on a single SolidWorks model where the only variable is the height.  I need to determine the worst case size in order to narrow the number of designs that I need to do mechanical testing on.  This is a medical device and will be evaluated by the FDA.  The heights range from 24mm - 70mm in 2mm increments.  I am struggling with getting results that are even close to linear across the range of sizes.  The FDA dictates that you mechanically test the tallest device, but they also state that it needs to be the worst case.  Is it best to use a consistent mesh size for each configuration?  Or is it best to have the number of elements be close to equal for each configuration regardless of mesh size?  If I use the automesh, it seems to produce unexpected results (i.e. 24mm showing higher Von Mises stress than the 70mm).

        • Re: Simulation - Meshing a family of parts
          Bill McEachern

          try specifying the element size. If you hit reset every time the element size changes depending onthe physical dimensions of the model. A picture would help or mock somehting up that is similar and show the effect. If it is hapening at a restrain it probably isn't real.

            • Re: Simulation - Meshing a family of parts
              Michael Mueller

              Is there an easy way to specify the element size rather than the mesh size?  BTW... I have tried using both standard and curvature based mesh settings.  I just really need to be able to trust the results of the FEA across multiple sizes.  I thought it would be simple and reliable to run the simulation a single piece model considering the only thing that changes is the height.  At this point, I just don't trust the results.

               

              I've attached a few images... hopefully they'll help explain what I'm talking about.80400-XX M6-24.JPG80400-XX M6-70.JPG

                • Re: Simulation - Meshing a family of parts
                  Bill McEachern

                  Just put in the element size in the dialog. I would use the curvature based mesher myself just because it is faster.

                    • Re: Simulation - Meshing a family of parts
                      Michael Mueller

                      Do you think it would be best to use one consistent element size for each configuration?  In other words... 0.03000000in for all

                      configurations (heights)?

                       

                      Capture.PNG

                        • Re: Simulation - Meshing a family of parts
                          Bill McEachern

                          yes. then the results will be comparable - well at least with respect to mesh size.

                          • Re: Simulation - Meshing a family of parts
                            Jared Conway

                            Michael, Bills got your back here. Think about it this way, if you try to get the same number of elements, you'll have to have bigger elements to fill the same volume. Those elements will be deformed and potentially give you bad results. When you go with the same size element, you'll fill the volumes with the same size elements and you can go with one that you know will give good results. A good start would be half the thickness of the smallest wall for a trend study. (2 elements across the thickness)

                              • Re: Simulation - Meshing a family of parts
                                Michael Mueller

                                Thanks Bill and Jared for your input.  I am trying to find the balance between a mesh size that is best suited for the geometry and what my computer will run in a reasonable time frame.  My computer is a Dell T7500 workstation, 64 bit quadcore xeon processor, 6 GB ram, and Nvidia Quadro FX580 graphics card.  I can't tell you how frustrating it is to wait 2 hours or more for it to run... only to find out that the solution either failed or the results are skewed.  I probably should have also mentioned that this is a non-linear material (PEEK).  Any suggestions for running non-linear static analysis?  It is just a simple static axial compression study, but I have several configurations to run in order to determine the worst case configuration.

                                  • Re: Simulation - Meshing a family of parts
                                    Jared Conway

                                    without dimensions, number of elements or an idea of the setup, i'm not sure if 2hrs is reasonable or not. if you have a simple static analysis, that seems like too long.

                                     

                                    here's my suggestions for elements:

                                    draft > get the problem running

                                    trend studies > use a coarser mesh

                                    final studies > use the fine mesh (2 elements across the thickness)

                                     

                                    that should give you a good blend.

                                     

                                    regarding nonlinear, before you go down that path, are you sure you need nonlinear? are you going into the nonlinear range of the material? do you have perm set? do you have large deformation? will the stiffness of the structure change significantly with loading?

                                     

                                    if not, stay in linear