19 Replies Latest reply on Oct 21, 2014 1:56 AM by Jared Conway

    Nonlinear modeling of plastics in SW

    Dermott McHugh

      I am trying to simulate a press fit plastic retaining ring in Solidworks Simulation using a Nonlinear study. I am using the 2d axisymmetric simplification.

       

      Study Setup.png

      For boundary conditions, all edges of items 1 2 and 3 are constrained such that the parts are essentially rigid. Body #2 is completely fixed. Body #3 is fixed in the horizontal direction and a downward displacement is applied from time = 0 to time = .5 so that it contacts Body #2. Body #1 is fixed in the horizontal direction and an upward displacement is applied from time = .5 to time = 1 so the that it also contacts Body #3. Body #4 is the press fit ring of interest. I am using no penetration contact sets with no friction.

       

      The press fit ring is made of UHMWPE. A representative true stress/strain curve is shown here:

      Material Model.png

      To try and simplify the material model, I have been using a bilinear plasticity simplification. Ex = 850 MPa and Etan = 104 MPa. Poisson's ratio is 0.46.

       

      For a first run, I ran with a linear elastic model to test out the contact and boundary conditions. Shown is the stress plot at time = 1. Maximum stress is 122 MPa.

      Elastic Solution.png

       

      When I run the plastic model, I get failure at Step > 1 with the classic error message referencing a load limit or buckling:

      Failure Message.png

      Here is the stress state when the solver kicks the bucket: Hard to see I know, but it has basically just come into contact.Plastic Failed.png

       

      Does anyone have any ideas about what might be going wrong here? It seems like the contact is failing for some reason with the plastic material condition. One thing I have noticed helps the solution get further is to increase my ETan, making the material stiffer after yield, but obviously this doesn't represent the material I will be using.

        • Re: Nonlinear modeling of plastics in SW
          Jared Conway

          did you follow all the recommendations from the dialog?

           

          do you need a nonlinear material? generally recommend starting with linear to see if you can get the problem to run then you can add complexity as you go.

            • Re: Nonlinear modeling of plastics in SW
              Dermott McHugh

              Yes I ran it as a nonlinear static analysis with a linear elastic material to test out boundary conditions. The stress state resulting from that analysis is above.

               

              I tried reducing the singularity elimination factor, but it didn't seem to have an effect. My Etan is > Ex/100, and changing it changes the material properties, so that doesn't seem like a great option either.

                • Re: Nonlinear modeling of plastics in SW
                  Jared Conway

                  what about just linear with linear material

                    • Re: Nonlinear modeling of plastics in SW
                      Dermott McHugh

                      I have run the study in linear static with a linear elastic material model, but I used a shrink fit contact condition. For the study using displacements, it gives very unrealistic results because the large displacement option is not available in 2D simplifications.

                       

                      It seems that the failure may be due to the material right around the contact buckling for some reason, possibly because the tangent modulus is so low. Does that make sense? If so, how could I go about limiting that behavior?

                        • Re: Nonlinear modeling of plastics in SW
                          Jared Conway

                          if your material is correct, you will need to try changing the solver parameters. changing and restarting can help.

                           

                          if shrink works in linear, why not use shrink in nonlinear?

                            • Re: Nonlinear modeling of plastics in SW
                              Dermott McHugh

                              Interesting. Which solver parameters would you recommend altering? I have tried quite a few of them without luck so far.

                               

                              Here are some pictures showing the contact right before the solver fails. The first image is the second to last step and the second is the final step before failure. Sorry I'm having trouble using the insert picture button right now so they're attached.

                                • Re: Nonlinear modeling of plastics in SW
                                  Jared Conway

                                  there is no optimal combination of them, you end up having to try modifying within their parameters. once you have exceeded your options, i would recommend contacting your reseller and having them pass this to the developers. it may be a limit of what the solver can do or require significant digging to determine if a setup change can allow it to complete.

                                    • Re: Nonlinear modeling of plastics in SW
                                      Dermott McHugh

                                      Have you ever had success running simulations with contact and soft, nonlinear materials? What material model did you use?

                                       

                                      Since modifying Etan helps the solution, and a linear elastic model works well, I suspect that the material model is what is affecting the solver. Has anyone had experience running nonlinear simulations with contact using soft materials which soften considerably after yield, or at least have similar stress v strain curves?

                                       

                                      Currently I am trying to run the simulation with a maximum time step of .001, which seems to make it past the critical point at which the solutions with a large maximum time step failed. We'll see if it makes it to completion.

                                        • Re: Nonlinear modeling of plastics in SW
                                          Bill McEachern

                                          Non-linear elastic or VM Plasticity would be my recommendations.

                                          Further I think you ought to try lowering your mesh density on the target bit - make it coarser.

                                          I have to say I don't get your BC's. Should they ot be out near the boundarys on this one as it is axis symmetric - could be I just don't have a grasp on what you are after here but it strikes me as odd where they are.

                                          In my expereince the shrink fit in NL studies can be problematic.

                                          You should be using surface to surface contact and not node to surface contact if you wan the thing to get very far.

                                          Hope this helps.

                                            • Re: Nonlinear modeling of plastics in SW
                                              Dermott McHugh

                                              Hi Bill,

                                               

                                              I'd be happy to clarify, thanks for your help here.

                                               

                                              The appearance of the fixtures is a little bit of a graphics glitch. See the attached figure for the full idea of it. Basically the outside edges of each body are completely constrained, so that all the parts except the plastic ring are rigid (the option to make a part rigid is not available in axisymmetric). I'm also using surface to surface contact, defined explicitly through 3 contact sets. The nonlinear simulation with a linear elastic material had no trouble solving, so I feel pretty good about my contact and boundary conditions.

                                               

                                              I will try lowering the mesh density and see how that goes. Just curious: any idea why that would help the solver? I had made it coarser because the failures I was getting earlier seemed to be due to buckling of the edges of individual elements near the contact.

                                              • Re: Nonlinear modeling of plastics in SW
                                                Dermott McHugh

                                                It seems like coarsening the mesh did help quite a bit. I have finally gotten a study to solve. Unfortunately, the contact conditions seem to have some issues, because at the final solution, the plastic part was interfering with the other metal parts significantly. The contact pressure plot does not look unrealistic, but the final position of the plastic part clearly has some issues. I am pretty baffled by this result... any ideas?

                                                 

                                                Also, Jared, I have gotten in touch with goengineer as well to see where that takes me.    

                                                  • Re: Nonlinear modeling of plastics in SW
                                                    Bill McEachern

                                                    Hi Dermott,

                                                    I would just make the other bits steel or unobtainium (use whatever finite stiffness you would like) and put the BC's not at the contacts as it seems to me to be very unrealistic and as you probably know with NL analysis unrealistic doesn't really bode well for success. Steel seems pretty stiff relative to UHMWPE.  However, it may not make any difference - your call.

                                                    I doubt that buckling is an issue. I would suspect the solution is failing due to contact convergence issues. I would coasren the mesh further on the target bit and try to get as uniform a mesh as you can get on both bits in contact:Try the standard mesher it will be more uniform it will make the solution faster and may make it more robust.

                                                    I would turn off friction: friction is non-conservative and Cosmos's friction handling is not exactly world renowned. If it work without friction then try it with friction (start small and work your way up).

                                                    It does seem odd that with contact failure as eveidence by the plot you still get a decent contact pressure plot. Is the default plot maginifaction set to 1? The other end looks as if it has contact issues as well so maybe the plot mag is not set to 1?

                                                • Re: Nonlinear modeling of plastics in SW
                                                  Jared Conway

                                                  we have performed nonlinear simulations with contact on rubber seals for example. i don't have the details of the rubber off the top of my head but i think they are probably softer than what you are using. in some situations we were not get them to complelely fold over on themselves due to numerical instabilities but we usually got pretty close. some required a lot of work on time step/solver parameters, coarsening the mesh like bill said and sometimes only ran with draft. we've also had to modify the setup of the problem to get it to solve. usualy under the guidance of the developers.

                                                    • Re: Nonlinear modeling of plastics in SW
                                                      Dermott McHugh

                                                      Okay that's good to know that other people have been successful with solidworks at least. I am sorely missing my old seat of abaqus right now.

                                                       

                                                      Am I correct in thinking that having no friction should simplify the solution?

                                                       

                                                      Just to clarify a little further, which solver parameters are you referring to, Jared? I understand you can't tell me exactly how to change which one, but I'm just looking for some examples.

                                                       

                                                      Thanks again, everyone.

                                                        • Re: Nonlinear modeling of plastics in SW
                                                          Jared Conway

                                                          all of them

                                                          singularity elimination factor

                                                          min time step

                                                          ...etc

                                                           

                                                          removing friction can help sometimes

                                                          but also adding friction can help sometimes

                                                           

                                                          its a delicate process when it comes to including all nonlinearities

                                                           

                                                          i should also note that sometimes i've had more success with 3d than 2d

                                                           

                                                          if you haven't engaged your VAR yet, I would definitely do that to make sure you aren't going down a rabbit hole. this assumes you've confirmed that your setup is valid overall. (bills comments)

                                                  • Re: Nonlinear modeling of plastics in SW
                                                    Jaspreet Hothi

                                                    There are couple of things I do when I suspect buckling. First, change the solver to Arc length, however you cannot give diplacement BC for arc length, you will have to apply force. You can also loosen the convergence tolerance. I generally start the solution with the default tolerance and when the solver kicks the bucket I lower the convergence tolerance a little and restart the solution. In my case it is enough to take me through the problamatic areas. Then I stop and continue the solution at default convergence tolerance.

                                                    Looking at your before and after pictures it appears that the contacts are doing their job properly before failure but at failure the contacts are behaving erratically. The highest stress apears to be at a location barely in contact. Another way to find if it is a contact issue is to isolate part 4 and look at the deformations for part 4 at higher scaling factor. You may see some unrealistic edge deformation at failure. Playing with mesh densities may help.

                                        • Re: Nonlinear modeling of plastics in SW
                                          Dermott McHugh

                                          Here is the state at failure with an ETan of 500 MPa (5x of actual).