4 Replies Latest reply on May 22, 2008 10:34 AM by Alessandro Tornincasa

    Simulating Pins with shell elements

    Alessandro Tornincasa
      I have a "desperate" question. I need to simulate rivets (they could be considered as pins) on a large structure (the frame of a train toilet). The structure is made of such long and thin parts that you should mesh it only with shell elements if you want to avoid running out of RAM (even on X64 system).

      My need is simulating reactions on pins in order to find out what are the most loaded pins.
      The great problem is that pin connectors are available only for solid elements, and I can't find fast or convenient workarounds.
      I've been suggested a couple, but they are really time consuming:
      - creating a cylindrical surface connecting the two holes and then evaluate tension in one direction (for example sigma Y) at the intersection with the hole edge. If you multiply tension by the thickness of the part you can get an idea of the value of force acting on the pin
      - Creating a small solid ring in the holes in order to be able to create a mixed mesh and define a pin connector on the solid body

      In other FEM's you'd put a beam element in the holes, bond it, and then you'd directly get axial or bending forces on the beam. The problem is that it's very hard to do it in COSMOSWorks because if beams are too short ball pins are not created.

      Do you have any better suggestions ? Any are really welcome.
        • Simulating Pins with shell elements
          Fernando Cuenca
          You can do the same in Cosmos 2008. Mixed meshing includes beam elements.
          • Simulating Pins with shell elements
            Anthony Botting
            Hi Allesandro: This may sound like a strange way to do it, but for getting reactions, I have done it with "joints" in COSMOSMotion 2007 (or 2006 will work). The setup in almost trivial, and it runs and provides an answer almost instantly. If you have it, just open the entire assembly in CM 2007 and turn-on gravity (and/or forces). Put in joints where pins are located, manually (I do this because I do not trust some of the automatic mate-to-joint mapping in 2007/2006 since the algorithm is known to arbitrarily remove redundancies, which may include the very connection you desire). Run the simulation for 1 second, and voila, you have the reactions at the joints. There is obviously no 'motion' in the assembly, but it works. I have done this for a structure hanging from multiple hinges and other supports, only to obtain the reaction distribution on each support component, then use FEA to do the elastic analysis on each subassembly separately. If you have CM 2008, I am unsure how to do this, because the developers removed functionality and completely did-away with the concept of joints.
              • Simulating Pins with shell elements
                Alessandro Tornincasa
                Hi Antony,
                thanks for your interesting suggestion, it's an interesting approach and I will give it a try.
                For what concerns COSMOS Motion 2008, a pure revolute joint is given by a concetric mate and a linear motor with zero displacement.
                A concentric mate plus a coincident or a distant mate can also be considered as a revolute joint because COSMOS Motion removes extra constraints correctly.

                When you have created the mate you can create a result graph on the concetric mate.
                  • Simulating Pins with shell elements
                    Alessandro Tornincasa
                    I've tought about a workaround to solve the issue (see attached pciture).
                    You build two cylindrical surfaces to accomodate the pin, then create a study with mixed meshing (which will allow you to select the pin connector).
                    In this study you define just shells on all surfaces, and you are able to create a pin connector on the two cylindrical shell surfaces.
                    By comparing this study with a similar one created with solid elements the results looks almost the same.