11 Replies Latest reply on Dec 10, 2014 11:54 PM by Jared Conway

    Simulation mass model produces different results than force model

    Ian Soukup


      I have two models variations:

      • a mass model in which I generate the my load from a block and apply gravitational acceleration
      • force model in which I apply a force of the same magnitude as the mass from the mass model

       

      I have applied mesh control to specific regions on a support pin and measured strain in those regions.  The support pin and mesh control on the pin is identical in both model varients.  I get a 10% difference in the measurements between the two models even though the contact force on the support pin is identical between the two model varients when measured after the simulation is complete.  The mass model is quite larger, so is it possible that the additional mesh elements outside of my mesh controlled regions are causing that big of a deviation?

       

      Thanks,

      Ian

        • Re: Simulation mass model produces different results than force model
          Shaun Densberger

          Can you post some pictures of the model so that we can get a better idea of what we're dealing with?

           

          One key point here is that you're measuring and comparing the strain, which is determined by differentiating the displacement field, so very small changes in the displacement solution can cause a large change in the strain solution.

           

          Now, between the two models, does one have this block while the other doesn't? If so, how is the block interfaced to the model? These specifics can have a large impact on the results depending on the characteristics of the model. For example, imagine you have a very thin plate (support on its edges) that you're resting a block on (where the block is about 50% the size of the plate). If you were to compare the results between a model where you represented the block with a force vector and one where you modeled the block (both models have gravity), then you'd see different results; the magnitude of the difference depends on how you modeled the block in. Think of a force load as water weight; it has absolutely zero stiffness to it and gets uniformly distributed over the surface it's applied to. The block does not do this. If you bond the block to the plate, then you're adding a significant amount of stiffness to the model (the bonded interface between the block and the plate increases the local stiffness). If you apply a contact interface, then you'll end up biasing the weight of the block around it's edges due to how the plate deforms.

           

          Finally, in your force model, did you keep the gravity load active? Key in mind that a gravity load is for the entire model, and that if you excluded it from your force model then you're essentially excluding the weight of your other components from the model. How much of an impact this has depends on the specifics of the model.

            • Re: Simulation mass model produces different results than force model
              Ian Soukup

              Shaun,

               

              Thank you for your reply.  Yes I am very familiar with stiffness and constraint differences between a force load and a load applied by an object with elastic properties.  I am also certain that the force on the pins is the same in magnitude between the mass and force model. I have accounted for weight of all components below the pins in the force model and do not have gravity active.  This is also verified by checking the contact forces on the pins after the simulation is complete and both models show the same forces.  The pins themselves have the same mesh density applied through mesh control but the mass models is much larger so other components have a courser mesh.

               

              The strain is measured on specified patches defined by split lines and averaged over the area of the patch along a principle axis.  I have performed a few sensitivity studies such as varying the width of the force applied to the U-shaped beam and it doesn’t change the results within a fairly large variation.  I have also increased the mesh size in the force model to the surrounding geometry to a similar size as the mass model and again the results were not affected.  See the attached images.

              Models.JPG

              Thanks,

              Ian

            • Re: Simulation mass model produces different results than force model
              Mike Pogue

              If I understand the second model correctly, you have explicitly modeled the entire assembly and applied a no penetration condition at that joint. In this case, a 10% difference between that and the first model would be pretty impressive. The first model can't get at the contact stress at all, it can only reproduce some approximation of the average contact stress. Hertz contact between the two components, which the no-penetration condition is trying to reproduce, will not be evenly distributed. So the max stress in that model has to be higher than the average

               

              Given a 10% difference, and all of the sensitivity studies you've done, I'd have a high confidence that you are within 10% or so of the "actual" answer.

               

              That seems pretty good to me.

                • Re: Simulation mass model produces different results than force model
                  Ian Soukup

                  Mike,

                   

                  Thank you for your answer, the mass model does show a higher strain response.  I have applied the same no penetration contact conditions at the pins, where I am making my strain measurement, the difference is on the Ushaped beam attached to the pins where I have another no penetration contact for the mass below the Ushaped beam (mass model) verses a force load on a small area defined by a split line (force model).  See image above.

                   

                  Is it possible that the higher contact stresses on that beam would result in a higher response within the pin?

                   

                  I will also try re-evaluating my convergence as suggested by Shaun.