8 Replies Latest reply on Nov 13, 2018 1:25 PM by Erik Kostson

    Nonlinear study elastoplastic - reaction force to high, plasticizes to early

    Juan Rodríguez

      Hello Community,

       

      I am having troubles with the following (spring) geometry:

      As you can see it´s basically a spring element which is locked in position at the bottom surface and gets loaded from the top. My goal is to calculate the stiffness (N/mm) of this structure in the elastic region and the elastic limit force before permanent deformation.

      My results show way too high reaction forces when compressing the spring compared to my experimental data.

      I am conducting a non linear simulation (Plasticity von Mises Model) which shows the following Force <-> Deflection behaviour.

       

      Simulation ResultExperimental Data

      E12.PNG

      Time 1s equals 0.1 mm (edited) of Displacement which gives me a stiffness k of approx. 30000 N/mm in the elastic region.

      The experiment shows the same degressive trend of the force - but with a much lower stiffness in the elastic region and a higher maximum force.
      (Stiffness 10x lower, maxForce 3x higher)

      Solver:
      Big Deformations
      Big Displacement
      Initial Time Step: 0,01
      Intel Direct Sparse

      My thoughts:
      Reason for that behaviour could be for a too coarse mesh, an overconstrained model or wrong material parameters (EMod; ETan; YielStrength).
      I doublechecked all those parameters and tried different mesh refinement (AspectRatio & Jacobian seem to be ok) but i never came close to the experimental data. I dont think shear locking is an issue here ...

      - so are there any other reasons which could increase the stiffness and decrease the maximum Force?

        • Re: Nonlinear study elastoplastic - reaction force to high, plasticizes to early
          Bill McEachern

          could it be the test set up? What is the force deflection curve without the spring element in the test rig?

          • Re: Nonlinear study elastoplastic - reaction force to high, plasticizes to early
            Mohamed El Orche

            Hi Juan,

             

            Maybe the time step? What results do you have if you input the stress-strain curve instead of the tangent modulus?

            • Re: Nonlinear study elastoplastic - reaction force to high, plasticizes to early
              Juan Rodríguez

              @bill

              The stiffness of the test rig is very high and should therefore have no impact on the measurements.

              @Mohamed

              I tried a different initial step size (down to 0.01)...  never experienced any convergence issues.

              Defining the Material by the Stress-strain Curve  just slighty changed the force.

               

              Edit: 1s Equals 0.1 mm not 0.01 mm as written above.

              • Re: Nonlinear study elastoplastic - reaction force to high, plasticizes to early
                Erik Kostson

                Also to point out that the calculation you do for the stiffness is not for the elastic region. At 300 N and at 1 s the spring is already way into hardening. Your stiffness in the elastic region should be checked for up to 0.1 sec. not beyond that.

                 

                If you can attach an iges or step file of the geometry with a small explanation of boundary conditions (is it full fixed at the bottom face, and you apply an incremental displacement of the top and no other restraint to the different directions there; material properties,..), I can have a look with the software I use (I would expect to agree at least for the linear part-if that is the case then there is a difference between test and FEA).

                  • Re: Nonlinear study elastoplastic - reaction force to high, plasticizes to early
                    Juan Rodríguez

                    Yes you are right, at 300N we are far into hardening - i made an error in my first post. (1 second = 0.1mm displacement).
                    I attached the geometry to my first post. The part is full fixed at the bottom surface!

                    Than u!

                      • Re: Nonlinear study elastoplastic - reaction force to high, plasticizes to early
                        Erik Kostson

                        So 2 independent software, one of them being Strand7, give for elastic perfectly plastic stainless steel (E=200 GPa) yielding at 300 MPa and zero tangent modulus, a maximum reaction force (until it goes fully plastic and curve goes flat with zero tangent stiffness) of about 430 N. Linear spring equivalent stiffness is about (3.3E7) (N/m) You should get the same value in SW (I got it), and then you know your FEA set up is OK at least. Bare in mind that I have not done a mesh convergence study.

                          • Re: Nonlinear study elastoplastic - reaction force to high, plasticizes to early
                            Jay Seaglar

                            My guess here is that the post-manufacture material properties of the component in the real world test could be the most significant unknown. Even if the manufacturer performed a simple tension test to record the stress-strain curve of the same batch of stock material that was then used to manufacture the spring, there is a lot that could happen to the material properties between that tension test and the finished product (in an unloaded state). For example, plasticity, hardening, high temperatures, surface effects, etc.

                             

                            One interesting experiment could be to alter material property input in SOLIDWORKS Simulation until software results match the real world test. This would answer the question "all other things remaining unchanged, what initial material properties would provide a matching response?" You could then investigate what could explain those material properties by looking closely at the manufacturing process, decide whether it is feasible or unlikely, and take that into consideration when making any analysis conclusions.

                              • Re: Nonlinear study elastoplastic - reaction force to high, plasticizes to early
                                Bill McEachern

                                If there was wagering to be done here I would still be betting on experimental set up (or modelling error - the FEA set up does not match the experimental set up) or measurement/calibration error. While the material property variation from manufacturing may affect the yield slightly,  the stiffness is almost impossible to change by much though any process. This seems to be a stiffness error at least for the initial elastic portion of the response if I understand the posts so far. Reaction forces nor displacement are going to be affected much by mesh density with 2nd order elements as they are direct outputs. Which brings up a point - if first order elements are being used they can affect the stiffness but it is very unlikely that anybody used them and even more unlikely that they both (Eric and Juan) would have used them.

                                @ Juan - has the test rig been calibrated? Is it a purchased tensile or compressive test machine by Instron or MTS or was it something that was put together for your work here? FEA lines up with reality in my experience if it is done properly. This, at least for me, just screams modeling error or test error in my humble opinion.